EC-SWE dataset

.gitattributes

@@ -0,0 +1,61 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.avro filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.lz4 filter=lfs diff=lfs merge=lfs -text
+*.mds filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text
+# Audio files - uncompressed
+*.pcm filter=lfs diff=lfs merge=lfs -text
+*.sam filter=lfs diff=lfs merge=lfs -text
+*.raw filter=lfs diff=lfs merge=lfs -text
+# Audio files - compressed
+*.aac filter=lfs diff=lfs merge=lfs -text
+*.flac filter=lfs diff=lfs merge=lfs -text
+*.mp3 filter=lfs diff=lfs merge=lfs -text
+*.ogg filter=lfs diff=lfs merge=lfs -text
+*.wav filter=lfs diff=lfs merge=lfs -text
+# Image files - uncompressed
+*.bmp filter=lfs diff=lfs merge=lfs -text
+*.gif filter=lfs diff=lfs merge=lfs -text
+*.png filter=lfs diff=lfs merge=lfs -text
+*.tiff filter=lfs diff=lfs merge=lfs -text
+# Image files - compressed
+*.jpg filter=lfs diff=lfs merge=lfs -text
+*.jpeg filter=lfs diff=lfs merge=lfs -text
+*.webp filter=lfs diff=lfs merge=lfs -text
+# Video files - compressed
+*.mp4 filter=lfs diff=lfs merge=lfs -text
+*.webm filter=lfs diff=lfs merge=lfs -text
+harbor/sources/druid-v9/task/grader/lib/budget-agent.jar filter=lfs diff=lfs merge=lfs -text

LICENSE

@@ -0,0 +1,32 @@
+This dataset is a collection of tasks. Each task includes code from an
+open-source repository (the `patch`, `test_patch`, `grader_files` and
+`dockerfile` fields may quote or modify it). That code stays under the licence
+of its upstream repository, listed per task below as reported by GitHub.
+
+The task text and grader design, where they are not upstream code, are
+copyright 2026 Toptal, LLC. Contact Toptal, LLC for licence terms.
+
+- BabylonJS__Babylon.js-babylonjs_scene_remove_owner_refs_contract-382616f8: Apache-2.0 (https://github.com/BabylonJS/Babylon.js)
+- MathNya__umya-spreadsheet-umya_structural_edit_reference_owners_contract: MIT (https://github.com/MathNya/umya-spreadsheet)
+- SickChill__SickChill-sickchill_episode_metadata_files_contract-2e2a49e4: GPL-3.0 (https://github.com/SickChill/SickChill)
+- Zettlr__Zettlr-zettlr_clickable_keywords: GPL-3.0 (https://github.com/Zettlr/Zettlr)
+- agronholm__apscheduler-apscheduler-expired-job-lease: MIT (https://github.com/agronholm/apscheduler)
+- aio-libs__aioftp-aioftp-stream-abort-desynchronization: Apache-2.0 (https://github.com/aio-libs/aioftp)
+- alibaba__druid-druid_lenient_script: Apache-2.0 (https://github.com/alibaba/druid)
+- alibaba__hooks-ahooks_request_cache_races: MIT (https://github.com/alibaba/hooks)
+- apache__geode-geode_oql_index_null_semantics_contract-1c066303: Apache-2.0 (https://github.com/apache/geode)
+- bashrc2__epicyon-epicyon_deleted_post_announce_cache_contract-1c066303: AGPL-3.0 (https://github.com/bashrc2/epicyon)
+- buchgr__bazel-remote-bazel_remote_disk_eviction_accounting_contract-afe5ea56: Apache-2.0 (https://github.com/buchgr/bazel-remote)
+- celery__celery-celery_retry_expires_countdown_contract: NOASSERTION (https://github.com/celery/celery)
+- colinhacks__zod-zod-json-schema-target-legality: MIT (https://github.com/colinhacks/zod)
+- coredns__coredns-coredns__4097: Apache-2.0 (https://github.com/coredns/coredns)
+- dgraph-io__badger-journal_recovery_after_unclean_stop: Apache-2.0 (https://github.com/dgraph-io/badger)
+- dotnet__Open-XML-SDK-open_xml_sdk_deletepart_refs_contract-1c066303: MIT (https://github.com/dotnet/Open-XML-SDK)
+- igniterealtime__Openfire-openfire_userdeleting_last_owner_contract-0163de9c: Apache-2.0 (https://github.com/igniterealtime/Openfire)
+- mikro-orm__mikro-orm-mikro-orm-pessimistic-locking: MIT (https://github.com/mikro-orm/mikro-orm)
+- ory__kratos-kratos_cleanup_stale_flows_courier_contract: Apache-2.0 (https://github.com/ory/kratos)
+- pallets__click-click-help-defaults: BSD-3-Clause (https://github.com/pallets/click)
+- python-rope__rope-rope_rename_all_reexport_owners_contract-1c066303: LGPL-3.0 (https://github.com/python-rope/rope)
+- sequelize__sequelize-sequelize-bulkcreate-ignore-duplicates: MIT (https://github.com/sequelize/sequelize)
+- terser__terser-retained_top_level_names_keep_their_values: NOASSERTION (https://github.com/terser/terser)
+- tsileo__microblog.pub-microblogpub_deleted_object_resurrect_lookup_contract-382616f8: AGPL-3.0 (https://github.com/tsileo/microblog.pub)

README.md

@@ -0,0 +1,417 @@
+---
+# Dataset card metadata. Spec: https://github.com/huggingface/hub-docs/blob/main/datasetcard.md?plain=1
+pretty_name: EC SWE
+language:
+- en
+language_details: en-US
+license: other
+license_name: mixed-upstream
+license_link: LICENSE
+license_details: >-
+  Each task carries code from an open-source upstream repository (patch,
+  test_patch, grader_files, dockerfile), under that repository's licence; the
+  repo_license column names it per row. The task text and grader design are
+  copyright 2026 Toptal, LLC; contact Toptal, LLC for licence terms.
+tags:
+- code
+- software-engineering
+- swe-bench
+- agents
+- coding-agents
+- benchmark
+- evaluation
+- bug-fixing
+- program-repair
+- multi-language
+annotations_creators:
+- expert-generated
+language_creators:
+- expert-generated
+multilinguality:
+- monolingual
+size_categories:
+- n<1K
+source_datasets:
+- original
+task_categories:
+- text-generation
+task_ids:
+- language-modeling
+configs:
+- config_name: default
+  data_files:
+  - split: test
+    path: data/test.parquet
+dataset_info:
+  config_name: default
+  features:
+  - name: instance_id
+    dtype: string
+  - name: repo
+    dtype: string
+  - name: base_commit
+    dtype: string
+  - name: setup_patch
+    dtype: string
+  - name: problem_statement
+    dtype: string
+  - name: patch
+    dtype: string
+  - name: test_patch
+    dtype: string
+  - name: patch_files
+    list: string
+  - name: grader
+    list:
+    - name: criterion
+      dtype: string
+    - name: weight
+      dtype: float64
+  - name: grader_files
+    list:
+    - name: path
+      dtype: string
+    - name: encoding
+      dtype: string
+    - name: content
+      dtype: string
+  - name: dockerfile
+    dtype: string
+  - name: language
+    dtype: string
+  - name: repo_license
+    dtype: string
+  - name: hud_task
+    dtype: string
+  - name: difficulty
+    dtype: string
+  - name: agent_runs
+    dtype: int64
+  - name: agent_rewards
+    list: float64
+  - name: agent_mean_reward
+    dtype: float64
+  - name: agent_solve_rate
+    dtype: float64
+  - name: agent_models
+    list: string
+  - name: gold_reward
+    dtype: float64
+  - name: base_reward
+    dtype: float64
+  splits:
+  - name: test
+    num_bytes: 16944456
+    num_examples: 24
+  download_size: 7713543
+  dataset_size: 16944456
+---
+
+# Dataset Card for ExpertCrowd SWE (EC-SWE)
+
+This dataset contains real software engineering tasks for evaluating coding
+agents. Each one is a bug or missing behaviour in a real open-source repository, pinned to a
+commit, with a written problem statement, a reference fix, and a weighted
+grader backed by hidden tests. The tasks were chosen to sit at the typical
+difficulty of the pool of reviewed tasks they were drawn from, measured by how
+well current frontier agents scored on them: no outliers at either end. Mean
+agent reward runs from 0.34 to 0.43, and 1.6% of agent runs scored a full 1.0.
+
+## Dataset Details
+
+### Dataset Description
+
+Every task was written by a software engineer, run on the HUD evaluation
+platform, and accepted by a human reviewer (review status `verified`). Before a
+task can be accepted, its reference fix has to score 1.0 on every grader
+criterion, the unchanged repository has to fail, and at least 5 agent runs on a
+current frontier model have to be graded. Automated checks also look at the
+prompt (it must state the symptom, not the fix), the grader (weights add up to
+100, each criterion is one observable requirement), and whether the task
+overlaps a public benchmark.
+
+This dataset keeps the tasks closest to the typical difficulty of all verified tasks with agent scores, chosen from those that met every export
+check (see [Curation Rationale](#curation-rationale)).
+
+## Uses
+
+### Direct Use
+
+- Evaluating coding agents on realistic repository-level tasks: give the agent
+  `problem_statement` and a checkout of `repo` at `base_commit` with
+  `setup_patch` applied, then score the resulting diff with the task's grader.
+- Studying difficulty: every task carries the per-run rewards of 128 graded
+  agent runs, so results can be compared against a known baseline.
+- Building or validating graders and task-quality tooling, using the reference
+  fix, the hidden tests and the weighted criteria together.
+
+## Dataset Structure
+
+One split, `test`, with one row per task.
+
+| Field | Type | What it holds |
+| --- | --- | --- |
+| `instance_id` | string | `<owner>__<repo>-<task>`, unique |
+| `repo` | string | upstream GitHub repository, `owner/name` |
+| `base_commit` | string | the 40-character upstream commit the task is built on |
+| `setup_patch` | string | a patch the environment applies to `base_commit` before the agent starts; empty for most tasks. Apply it first: `patch` and `test_patch` are written against the result |
+| `problem_statement` | string | the exact prompt the agent was sent, including a short preamble about the workspace (`/workspace/repo`) and resource limits |
+| `patch` | string | the reference fix, non-test files only |
+| `test_patch` | string | test files the reference fix adds or changes (may be empty) |
+| `patch_files` | list of string | every file the reference fix touches |
+| `grader` | list of (criterion, weight) | the grader's criteria and their weights in percent; weights add up to 100 |
+| `grader_files` | list of (path, encoding, content) | the grader as built into the task environment (`grader/`, `grading/` or `protected/`): hidden tests, the repository's regression suite where the grader reruns it, and the script that scores a run. `encoding` is `utf-8`, or `base64` for binary fixtures. The author's calibration notes, control diffs and adversarial manifests are left out |
+| `dockerfile` | string | the environment the agent ran in |
+| `language`, `repo_license` | string | from the GitHub API for `repo` |
+| `hud_task` | string | the task's name on HUD, for tracing a row back |
+| `difficulty` | string | bucket of `agent_mean_reward`: very hard (under 0.2), hard, medium, easy, very easy (0.8 and up) |
+| `agent_runs`, `agent_rewards` | int, list of float | graded agent runs on current models and each run's reward (0 to 1) |
+| `agent_mean_reward`, `agent_solve_rate` | float | mean reward, and the share of runs that scored a full 1.0 |
+| `agent_models` | list of string | the models those runs used |
+| `gold_reward`, `base_reward` | float | reward of the reference fix (1.0) and of the unchanged repository |
+
+```python
+from datasets import load_dataset
+
+ds = load_dataset("ExpertCrowd/ec-swe", split="test")
+hard = ds.filter(lambda r: r["difficulty"] in ("very hard", "hard"))
+```
+
+### At a glance
+
+- **Tasks:** one per repository
+- **Graded agent runs behind the difficulty numbers:** 128
+- **Models those runs used:** claude-fable-5, claude-fable-5-1, claude-opus-5, gemini-3.8-flash, gpt-6-astra
+
+| Language | Tasks |
+| --- | --- |
+| Python | 8 |
+| TypeScript | 6 |
+| Go | 4 |
+| Java | 3 |
+| JavaScript | 1 |
+| Rust | 1 |
+| C# | 1 |
+
+### Difficulty
+
+Difficulty is the mean reward of graded agent runs on current frontier models.
+The tasks sit around the middle of the distribution of every verified task with
+5 or more graded agent runs, so the mean and median stay close to the pool's
+while the very hard and easy tails are left out:
+
+| Mean agent reward | Difficulty | All verified with agent scores (124) | This dataset |
+| --- | --- | --- | --- |
+| 0.0 to 0.2 | very hard | 5 (4%) | 0% |
+| 0.2 to 0.4 | hard | 61 (49%) | 62% |
+| 0.4 to 0.6 | medium | 55 (44%) | 38% |
+| 0.6 to 0.8 | easy | 3 (2%) | 0% |
+| 0.8 to 1.0 | very easy | 0 (0%) | 0% |
+| **mean** | | **0.377** | **0.380** |
+| **median** | | **0.379** | **0.373** |
+
+### Tasks
+
+| Instance | Repository | Language | Difficulty | Mean reward |
+| --- | --- | --- | --- | --- |
+| `bashrc2__epicyon-epicyon_deleted_post_announce_cache_contract-1c066303` | bashrc2/epicyon | Python | hard | 0.34 |
+| `apache__geode-geode_oql_index_null_semantics_contract-1c066303` | apache/geode | Java | hard | 0.34 |
+| `coredns__coredns-coredns__4097` | coredns/coredns | Go | hard | 0.34 |
+| `sequelize__sequelize-sequelize-bulkcreate-ignore-duplicates` | sequelize/sequelize | TypeScript | hard | 0.34 |
+| `MathNya__umya-spreadsheet-umya_structural_edit_reference_owners_contract` | MathNya/umya-spreadsheet | Rust | hard | 0.35 |
+| `alibaba__druid-druid_lenient_script` | alibaba/druid | Java | hard | 0.35 |
+| `aio-libs__aioftp-aioftp-stream-abort-desynchronization` | aio-libs/aioftp | Python | hard | 0.35 |
+| `python-rope__rope-rope_rename_all_reexport_owners_contract-1c066303` | python-rope/rope | Python | hard | 0.35 |
+| `SickChill__SickChill-sickchill_episode_metadata_files_contract-2e2a49e4` | SickChill/SickChill | Python | hard | 0.35 |
+| `igniterealtime__Openfire-openfire_userdeleting_last_owner_contract-0163de9c` | igniterealtime/Openfire | Java | hard | 0.36 |
+| `Zettlr__Zettlr-zettlr_clickable_keywords` | Zettlr/Zettlr | TypeScript | hard | 0.36 |
+| `buchgr__bazel-remote-bazel_remote_disk_eviction_accounting_contract-afe5ea56` | buchgr/bazel-remote | Go | hard | 0.37 |
+| `terser__terser-retained_top_level_names_keep_their_values` | terser/terser | JavaScript | hard | 0.38 |
+| `ory__kratos-kratos_cleanup_stale_flows_courier_contract` | ory/kratos | Go | hard | 0.38 |
+| `dotnet__Open-XML-SDK-open_xml_sdk_deletepart_refs_contract-1c066303` | dotnet/Open-XML-SDK | C# | hard | 0.39 |
+| `dgraph-io__badger-journal_recovery_after_unclean_stop` | dgraph-io/badger | Go | medium | 0.40 |
+| `BabylonJS__Babylon.js-babylonjs_scene_remove_owner_refs_contract-382616f8` | BabylonJS/Babylon.js | TypeScript | medium | 0.40 |
+| `alibaba__hooks-ahooks_request_cache_races` | alibaba/hooks | TypeScript | medium | 0.41 |
+| `celery__celery-celery_retry_expires_countdown_contract` | celery/celery | Python | medium | 0.43 |
+| `colinhacks__zod-zod-json-schema-target-legality` | colinhacks/zod | TypeScript | medium | 0.43 |
+| `pallets__click-click-help-defaults` | pallets/click | Python | medium | 0.43 |
+| `tsileo__microblog.pub-microblogpub_deleted_object_resurrect_lookup_contract-382616f8` | tsileo/microblog.pub | Python | medium | 0.43 |
+| `mikro-orm__mikro-orm-mikro-orm-pessimistic-locking` | mikro-orm/mikro-orm | TypeScript | medium | 0.43 |
+| `agronholm__apscheduler-apscheduler-expired-job-lease` | agronholm/apscheduler | Python | medium | 0.43 |
+
+## Running the tasks
+
+There are two ways to run a task: on HUD, the platform the tasks were written for, or with
+Harbor on your own machine.
+
+### On HUD
+
+Every task runs on HUD, including the nine that have no Harbor package. The tasks are
+published there as the taskset [`ec-swe`](https://www.hud.ai/shared/tasksets/979e6e02-d726-46f6-a066-914afc5c9f44),
+which anyone can view; it is still being filled as its environments are built.
+
+To run a task in a HUD workspace of your own: each folder under `harbor/sources/` is the build
+source of one task's HUD environment. With a HUD account, on Linux, macOS or WSL:
+
+```bash
+pip install hud                       # the hud command
+hud set HUD_API_KEY=<your key>        # a key from hud.ai/project/api-keys
+hf download ExpertCrowd/ec-swe --repo-type dataset --include "harbor/sources/**" --local-dir ec-swe
+cd ec-swe/harbor/sources/<task folder>
+hud deploy                            # builds the environment in your HUD workspace
+hud sync tasks my-ec-swe tasks.py     # puts this folder's task into a taskset of yours
+hud eval my-ec-swe claude --full      # runs it on the platform with an agent
+```
+
+Results appear as jobs and traces on hud.ai (`hud jobs`). On HUD the agent and the grader
+share one container, kept apart by file permissions, and the four tasks with a criterion
+judged by a model get that judge from HUD's gateway. The commands follow HUD's
+documentation. We rehearsed `hud deploy` and `hud sync tasks` for every folder in their
+dry-run mode and ran them in full for the terser folder; we have not run an agent with
+`hud eval` from this folder.
+
+### With Harbor
+
+The `harbor/` folder of this repository makes tasks runnable end
+to end with [Harbor](https://github.com/harbor-framework/harbor) 0.23.0 and
+Docker, and nothing else. Each task is packaged as two containers: one for the
+agent, holding the repository at its base commit and nothing from the grading
+side, and a separate one that grades the patch the agent leaves behind with the
+task's original grader. Both are offline by default. The folder holds each
+task's complete build context, so its environment can be rebuilt from this
+repository alone.
+
+```bash
+hf download ExpertCrowd/ec-swe --repo-type dataset --include "harbor/**" --local-dir ec-swe
+cd ec-swe/harbor
+pip install harbor==0.23.0
+python build_images.py --task <task folder>
+harbor run -p tasks/<task folder> -a oracle     # the reference fix: reward 1.0
+harbor run -p tasks/<task folder> -a nop        # no change: reward 0.0
+harbor run -p tasks/<task folder> -a <agent> -m <model> --allow-agent-host <model API host>
+```
+
+`harbor/README.md` has the requirements, a table that maps each task folder to
+its `instance_id` here, the scripted checks, and what was recorded for each
+task. Every packaged task was run through Harbor on the files in the folder: an
+empty change scores 0, and the reference fix scores 1.0 with every criterion
+equal to the task's original grading. The folder contains the reference fixes
+and the hidden graders, as the rows of this dataset do.
+
+11 more (badger, bazel-remote, celery, geode, kratos, microblogpub, open-xml-sdk, openfire, terser, umya, zettlr) are packaged and are being re-checked; they are added
+as each one passes.
+
+Nine tasks have no Harbor package yet: click, sequelize, druid and hooks (one criterion each
+is judged by a model), SickChill, Babylon.js, mikro-orm, aioftp and apscheduler. The folder
+holds the complete build context of each, and `harbor/README.md` says for each how it is
+built and graded and what it needs.
+
+## Dataset Creation
+
+### Curation Rationale
+
+The goal was a public set at the typical difficulty of the full pool: hard
+enough that no current agent solves the tasks reliably, without tasks so hard
+that every agent scores near zero, or so easy that they no longer separate
+models.
+
+1. **Pool.** Every task whose current review status on HUD is `verified'. Validation copies of a task are not tasks and were
+   left out.
+2. **Export checks.** A task was kept only if HUD gave a prompt, a GitHub
+   repository and a pinned 40-character commit, a non-empty reference fix,
+   grader weights adding up to 100, a reference run on the current task version
+   scoring 1.0, and at least 5 graded agent runs on a current model. The
+   repository had to be an established upstream project, not one created for
+   the task, and the environment could not depend on a task author's personal
+   images. Two tasks with the same reference fix, or the same repository and
+   scenario, count once.
+3. **Stratified sample.** All verified tasks with 5 or more
+   graded agent runs were ranked by mean agent reward and cut into bins of
+   equal size. Each bin contributed one task that passed the export checks,
+   preferring a repository not yet in the sample. Matching against the whole
+   scored pool, not only the exportable part, keeps the export checks from
+   skewing the difficulty curve.
+4. **Typical difficulty.** From that sample, the tasks whose mean agent
+   reward was closest to the sample mean (0.374) were kept: every task within
+   0.059 of it. The nearest task left out was 0.064 away, so the cut does not
+   split a tie.
+
+### Source Data
+
+#### Data Collection and Processing
+
+Task authors chose a bug or missing behaviour in an open-source repository,
+wrote the problem statement, built a container that checks out the repository at
+a pinned commit, wrote a reference fix and a hidden-test grader, and ran the
+task on HUD: a reference run, a run with no fix, and at least 5 agent runs. The
+export reads the prompt from the agent's own trace, the criteria and weights
+from the reference run's grading, and the repository, commit, reference fix and
+grader from the environment build.
+
+#### Who are the source data producers?
+
+The code comes from the upstream open-source projects named in `repo`, written
+by their contributors. The problem statements, reference fixes, hidden tests and
+graders were written by professional software engineers contracted to build the
+tasks.
+
+### Annotations
+
+#### Annotation process
+
+Each task was reviewed on HUD by a human reviewer, who could send it back for
+revision or reject it; only tasks marked `verified` are included. Automated
+checks run before review on the prompt, the grader, the reference and baseline
+runs, and overlap with public benchmarks. The difficulty fields come from graded
+agent runs, not from people.
+
+#### Who are the annotators?
+
+Experienced software engineers acting as reviewers, separate from each task's
+author.
+
+#### Personal and Sensitive Information
+
+The tasks are about source code. Author and reviewer names were left out of
+the export. Upstream code can contain contributor names in comments, licence
+headers or commit metadata, as it does in the upstream repositories.
+
+## Recommendations
+
+Report `agent_solve_rate`-style full solves alongside mean reward, report
+results per difficulty bucket as well as overall, and give the model's training
+cutoff relative to this dataset's release.
+
+## Citation
+
+**BibTeX:**
+
+```bibtex
+@misc{toptal2026ecswe,
+  author       = {{Toptal, LLC}},
+  title        = {{EC SWE}},
+  year         = {2026},
+  publisher    = {Hugging Face},
+  howpublished = {\url{https://huggingface.co/datasets/ExpertCrowd/ec-swe}}
+}
+```
+
+**APA:**
+
+Toptal, LLC. (2026). *EC SWE* [Data set]. Hugging Face.
+https://huggingface.co/datasets/ExpertCrowd/ec-swe
+
+## Glossary
+
+- **Gold run:** the task run with the reference fix applied. It must score 1.0.
+- **Base run:** the task run on the unchanged repository. It should score low.
+- **Reward:** the grader's weighted score for one run, 0 to 1.
+- **Verified:** a task a human reviewer accepted on HUD.
+
+## Dataset Card Authors
+
+Toptal, LLC
+
+## Dataset Card Contact
+
+Toptal, LLC, through the ExpertCrowd organization on Hugging Face.

data/test.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8d46fa9d12cfc210336016075e357f6230b5aaa42fdc4276a3a30671ef97d62c
+size 7713543

harbor/README.md

@@ -0,0 +1,317 @@
+# ExpertCrowd SWE tasks: runnable Harbor packages
+
+This folder makes 4 ExpertCrowd software-engineering tasks runnable end
+to end with [Harbor](https://github.com/harbor-framework/harbor) and Docker, and
+nothing else. Each task is a real repository at a fixed commit, a written
+request, and the original hidden rubric that scores a change to that repository.
+An agent works in one container; a second container, which the agent never
+sees, grades the patch it leaves behind. The table under "Recorded
+qualification" maps each task folder to its row in `data/test.parquet`.
+
+The folder contains the reference solutions and the hidden graders. A model you
+are evaluating must see only the task prompt and its own sandbox, which is what
+Harbor gives it. Do not put the other files where that model, or its training
+data, can reach them.
+
+## Two ways to run a task
+
+- **With Harbor, on your own machine.** 4 tasks are packaged for
+  this under `tasks/`. Everything on this page up to "Tasks without a Harbor
+  package" is about this route.
+- **On HUD, the platform the tasks were written for.** All 24 tasks run there,
+  and the agent results in the dataset come from such runs. The tasks are
+  published there as the taskset
+  [`ec-swe`](https://www.hud.ai/shared/tasksets/979e6e02-d726-46f6-a066-914afc5c9f44),
+  which anyone can view; it is still being filled as its environments are
+  built. To run a task in a workspace of your own: each folder under
+  `sources/` is the build source of a task's HUD environment. With a HUD
+  account and the `hud` command, on Linux, macOS or WSL:
+
+  ```bash
+  pip install hud                       # the hud command
+  hud set HUD_API_KEY=<your key>        # a key from hud.ai/project/api-keys
+  cd sources/<folder>
+  hud deploy                            # builds the environment in your HUD workspace
+  hud sync tasks my-ec-swe tasks.py     # puts this folder's task into a taskset of yours
+  hud eval my-ec-swe claude --full      # runs it on the platform with an agent
+  ```
+
+  Results appear as jobs and traces on hud.ai (`hud jobs`). This is also the
+  route for the nine tasks without a Harbor package, including the four with a
+  criterion judged by a model, which HUD's gateway serves. The commands follow
+  HUD's documentation; we rehearsed `hud deploy` and `hud sync tasks` for every
+  folder in their dry-run mode and ran them in full for the terser folder; we
+  have not run an agent with `hud eval` from this folder. The build
+  sources are the ones the platform built the recorded environments from, with
+  the changes listed under "Contents".
+
+## What you need
+
+- Docker with Compose and BuildKit, on Linux, macOS or Windows with WSL.
+- Python 3.12 and Harbor 0.23.0: `pip install harbor==0.23.0`.
+- Network access while images are built. Tasks themselves run offline.
+- Disk: allow 5 to 10 GB per task for its images. Each task container is
+  limited to 2 CPUs and 8 GiB.
+
+No GPU, model account or API key is needed for the checks below.
+
+## Quick start
+
+From this folder:
+
+```bash
+python verify_bundle.py
+python build_images.py --task coredns_rewrite_svcb_opt
+harbor run -p tasks/coredns_rewrite_svcb_opt -a oracle    # the reference fix: reward 1.0
+harbor run -p tasks/coredns_rewrite_svcb_opt -a nop       # no change: reward 0.0
+python run_checks.py --task coredns_rewrite_svcb_opt --output checks-001
+```
+
+1. `verify_bundle.py` checks every file against `checksums.json`.
+2. `build_images.py` builds the task's source image from `sources/`. Leave out
+   `--task` to build every task. A build takes ten to forty minutes: most tasks
+   finish their build by scoring their own reference fix and the untouched
+   repository, and refuse to produce an image unless those score 1.0 and 0.0.
+3. `harbor run` with Harbor's `oracle` agent applies the reference fix and with
+   its `nop` agent changes nothing. These two need no script from this folder.
+4. `run_checks.py` runs five scripted checks per task and exits 0 only if each
+   gives what is expected:
+
+| Check | What is submitted | Expected |
+| --- | --- | --- |
+| `noop` | nothing | reward 0 |
+| `reference` | the reference fix | reward 1, and the agent's container is isolated |
+| `broken` | the reference fix plus a syntax error | reward 0, graded normally |
+| `infra` | the reference fix, with one grader file removed from the verifier | no reward: the attempt is invalid |
+| `tamper` | nothing, while a process keeps writing a full score to the result file | reward 0 |
+
+## Running an agent
+
+```bash
+harbor run -p tasks/coredns_rewrite_svcb_opt -a <agent> -m <model> --allow-agent-host <model API host>
+```
+
+Harbor's agents run inside the task's container, so they need to reach their
+model. Every task here declares no network for both phases;
+`--allow-agent-host` opens the agent phase to the hosts you name and nothing
+else, and the verifier stays offline. `harbor run --help` lists the agents.
+An agent that installs itself into the container may need further hosts for
+that step (`--allow-environment-host`).
+
+Only Harbor's `oracle` and `nop` agents were run for the qualification below.
+No run with a model-backed Harbor agent is part of this folder yet.
+
+## Recorded qualification
+
+Every task below was run with stock Harbor 0.23.0 on the files shipped here,
+with these results.
+
+| Task folder | `instance_id` in the dataset | Language | Criteria | Empty change | Reference fix | Agent image scan |
+| --- | --- | --- | --- | --- | --- | --- |
+| `coredns_rewrite_svcb_opt` | `coredns__coredns-coredns__4097` | Go | 7 | 0 | 1 | clean |
+| `epicyon_deleted_post_announce_cache_contract` | `bashrc2__epicyon-epicyon_deleted_post_announce_cache_contract-1c066303` | Python | 11 | 0 | 1 | clean |
+| `rope_rename_all_reexport_owners_contract` | `python-rope__rope-rope_rename_all_reexport_owners_contract-1c066303` | Python | 8 | 0 | 1 | clean |
+| `zod_json_schema_target_legality` | `colinhacks__zod-zod-json-schema-target-legality` | TypeScript | 12 | 0 | 1 | clean |
+
+The `broken`, `infra` and `tamper` checks, and the author's calibration patches where a task has them, have also been run on these exact files for 1 of the tasks: `coredns_rewrite_svcb_opt`. For the other tasks the `broken` and `infra` checks and the calibration patches last passed on an earlier build of the package, and all of these checks are being repeated on the files here. `manifest.json` records per task which checks ran on which build (`checks_on_shipped_files`, `further_checks`).
+
+- **Empty change** and **Reference fix** are the rewards of Harbor's `nop` and
+  `oracle` runs.
+- **Reference fix** also means that every criterion's value and weight equals
+  what the task's own grading code gives the reference fix inside its
+  unmodified source image, and that the isolation check passed: the agent runs
+  as an unprivileged user, sees none of the grading files and gets no traffic
+  out.
+- **Agent image scan** means no line of the reference patch, the hidden graders
+  or the criterion names is in the agent's image, other than lines the public
+  repository already contained.
+- **Author's calibration patches** are partial, alternative and cheating
+  solutions the task's author shipped with the task. Each patch is submitted
+  through the package and has to reproduce the score the author's own build
+  recorded.
+
+The per-task records are in `evidence/<task>/`: a summary, the native grading
+of the reference, the author's calibration output, each verifier report and the
+agent image scan.
+
+## How a score is produced
+
+- The agent's container holds the repository at its base commit with a fresh
+  one-commit history, the toolchain, and nothing from the grading side. The
+  agent runs as an unprivileged user.
+- When the agent finishes, a hook in the task captures its change as a patch
+  against a protected copy of the base commit. Only that patch reaches the
+  verifier.
+- The verifier is a fresh container. It checks its own grader files against
+  recorded hashes, applies the patch, and runs the task's original grading code
+  unchanged. Grader commands, and with them the submitted code, run as an
+  unprivileged user; the report, the rubric files and the repository's original
+  history belong to root.
+- The reward is the task's original reward: weighted criteria between 0 and 1,
+  with partial credit. Most tasks mark some criteria as blockers and cap the
+  reward at 0.6 when one fails. A few tasks apply a rule of their own instead
+  (`reward_rule` is `authored` in `manifest.json`); their reward is kept as
+  computed and cross-checked against the criteria.
+- An attempt is **invalid**, and gets no reward, when grading itself could not
+  be completed: a missing grader file or dependency, a grader command that
+  timed out or ended outside its pass/fail contract, missing criteria, or a
+  report that does not add up. Harbor reports it as an error, not as a score.
+  Re-run or exclude it; do not average it in as a zero.
+- A submission that does not compile is not an invalid attempt. Several tasks'
+  graders report "could not judge" when the build breaks. The verifier then
+  resets the repository and puts the same check to the untouched code. If the
+  grader reaches a verdict there, the submission is what broke it: the
+  criterion counts as unmet and the attempt is scored. If it cannot judge the
+  untouched code either, the environment is at fault and the attempt is invalid.
+
+## Contents
+
+- `tasks/<task id>/`: one Harbor task per folder: `instruction.md` (the prompt
+  the agent receives), `task.toml`, `environment/` (agent image recipe),
+  `tests/` (verifier recipe, verifier and its pinned configuration) and
+  `solution/` (reference fix).
+- `sources/`: each task's build context as downloaded from HUD. In eighteen of
+  them the environment name was changed to a neutral one (`ec-swe-...`, at most
+  three lines in `env.py`, `pyproject.toml` and `tasks.py`), and druid's had
+  comments shortened and three unused files removed; `manifest.json` lists the
+  changed files per task (`source_edited`). Every packaged task was built and
+  checked from the files here. `sources/archives/` holds the original archive
+  of each context that was not changed.
+  `build_images.py` builds from these.
+- `build_images.py`, `run_checks.py`, `verify_bundle.py`: standard-library
+  scripts; `run_checks.py` calls the `harbor` command.
+- `evidence/`: the recorded qualification.
+- `manifest.json`, `checksums.json`: versions, image names and file hashes.
+
+## Limits of what is shown
+
+- The checks are scripted. They show that each task can be built, attempted,
+  graded and scored as its author intended. They do not measure any model.
+- Everything was qualified on one machine (Docker Desktop on Windows, through
+  WSL). It has not been repeated on a second machine. As a test of the download
+  itself, this folder was fetched from the Hub into an empty Linux directory on
+  that machine, and one task (`retained_top_level_names_keep_their_values`) was
+  built and put through all five checks from there.
+- Task images are built on your machine, so their image IDs differ from the
+  ones recorded here. What is pinned is the content of the grader files, which
+  the verifier checks before every grading.
+- Builds download base images and dependencies. A recipe that builds today can
+  stop building when an upstream package is withdrawn.
+- A download gives every file mode 644. The images were qualified from build
+  contexts in which every file had mode 777, and some images keep the modes of
+  files they copy. `build_images.py` builds from a temporary copy with mode
+  777 for that reason. If you build a `Dockerfile.hud` by hand, give its
+  context the same modes first.
+- Run one task at a time on a machine that is otherwise quiet when a score
+  matters: some repositories' own test suites are timing-sensitive, and a busy
+  machine has turned a correct fix's gate red here.
+- The graders execute the submitted code. The isolation described above is a
+  concrete, tested boundary, not a security certification.
+
+## Tasks without a Harbor package
+
+Nine of the dataset's 24 tasks have no package under `tasks/` yet. The complete
+build context of each is under `sources/` (unmodified, except druid's: see its
+entry), so you can
+build the task's image and run it in a setup of your own. This section says how
+each is meant to run, what it needs, and what was seen when it was tried here.
+None of these routes gives the isolation of the packaged tasks: the agent and
+the grader share one container, separated by file permissions.
+
+**Building.** `python build_images.py --task <name>` builds the image
+`ec-swe-<name>:source-v<n>` from `sources/<name>-v<n>/Dockerfile.hud`, the same
+way as for the packaged tasks. `<name>` is the bold name in the list below.
+
+**How these images run a task.** Each image is an environment for the HUD agent
+harness. Its default command, `hud serve tasks.py`, serves the task; a harness
+built on the `hud` Python package (each `Dockerfile.hud` pins the version)
+connects to it, lets the agent work in the checkout as UID 1000, and then calls
+the task's grading in the same container as root. `/hud`, which holds the hidden
+tests, the grader and the reference fix, is readable by root only.
+
+**Grading without the harness.** In the standard layout (click, sequelize,
+hooks, sickchill, babylonjs, mikro-orm) two functions in `/app/tasks.py` do the
+work: `_setup()` prepares the checkout in `/workspace/repo`, and
+`_grade(mode)` returns the reward and the criteria. With `mode` `"golden"` it
+grades the reference fix; with `None` it grades the changes made to the
+checkout since setup. The results reported below were obtained by running this
+script inside the image, as root and with no network:
+
+```python
+import asyncio, json, sys
+sys.path.insert(0, "/app")
+import tasks
+
+async def main():
+    await tasks._setup()
+    result = await tasks._grade("golden")
+    print(json.dumps(result.model_dump(mode="json")))
+
+asyncio.run(main())
+```
+
+```bash
+docker run --rm -i --network none ec-swe-<name>:source-v<n> /opt/hudvenv/bin/python - < script.py
+```
+
+**The tasks**
+
+- **click** (`pallets__click-click-help-defaults`, Python, `sources/click-v9`).
+  Standard layout. One criterion, `maintainer_review` (6.25 of 100 points), is
+  decided by a language model through HUD's gateway, so grading needs
+  `HUD_API_KEY` in the container's environment and network access to that
+  gateway. When the judge cannot be reached, the task's own code drops that
+  criterion, rescales the others and records `judge_errors` in the result. Not
+  built or run here.
+- **sequelize** (`sequelize__sequelize-sequelize-bulkcreate-ignore-duplicates`,
+  TypeScript, `sources/sequelize-v27`). As click: `maintainer_review` is 6.25
+  points. Not built or run here.
+- **hooks** (`alibaba__hooks-ahooks_request_cache_races`, TypeScript,
+  `sources/hooks-v4`). As click: `maintainer_review` is 3 points. Not built or
+  run here.
+- **druid** (`alibaba__druid-druid_lenient_script`, Java, `sources/druid-v9`).
+  As click: `maintainer_review` is 2 points. The layout differs a little: there
+  is no `coding/` folder and no `tasks._grade`, the scenario function in
+  `tasks.py` does the grading, and the HUD runtime is in `/opt/hudenv`, so the
+  script above does not apply. This build context is not byte-identical to the
+  original: comments that recorded internal review history were shortened in
+  `tasks.py` and `preflight.sh`, and three files the image build does not read
+  were removed. `tasks.py` is the same program. Not built or run here.
+- **sickchill**
+  (`SickChill__SickChill-sickchill_episode_metadata_files_contract-2e2a49e4`,
+  Python, `sources/sickchill-v7`). Builds here; the build, which has network
+  access, ends by scoring the reference fix 1.0 and the untouched repository
+  0.0. Graded afterwards with no network, the reference fix scored 0.6: the
+  criteria of the fix itself passed (0.97 before the cap), and the gate failed
+  because the repository suite it runs includes tests that fetch from
+  pypi.org. Grade it with network access, or add those tests to
+  `task/grader/suite_deselect.txt`, the task's own list of network-dependent
+  tests, and rebuild.
+- **mikro-orm** (`mikro-orm__mikro-orm-mikro-orm-pessimistic-locking`,
+  TypeScript, `sources/mikro-orm-v7`, image about 10 GB). Builds here. Graded
+  with the script above, the reference fix scored 1.0. Its grader starts
+  database servers (PostgreSQL, MySQL, SQL Server), some of them under their
+  own accounts with `runuser`, which needs root. The packaged verifier runs
+  grader commands as an unprivileged user, which is why there is no package
+  yet.
+- **babylonjs**
+  (`BabylonJS__Babylon.js-babylonjs_scene_remove_owner_refs_contract-382616f8`,
+  TypeScript, `sources/babylonjs-v8`, image about 10 GB). Builds here. Grading
+  the reference fix with the script above did not complete on the
+  qualification machine (16 CPUs, 31 GB of memory, Docker Desktop on Windows):
+  setup changes the owner of the whole dependency tree and ran into its
+  600-second limit twice, and on a third attempt grading had not finished after
+  two hours. It may behave on a Linux host with a fast local disk; that has
+  not been tried.
+- **aioftp** (`aio-libs__aioftp-aioftp-stream-abort-desynchronization`, Python,
+  `sources/aioftp-v11`) and **apscheduler**
+  (`agronholm__apscheduler-apscheduler-expired-job-lease`, Python,
+  `sources/apscheduler-v17`). A different layout: the checkout is at
+  `/home/ubuntu/<name>`, the HUD runtime is in `/hud/.venv`, and grading is the
+  function `grading.runner.grade_repository` in `/hud/grading`. The scenarios
+  (the task itself, `reference_gold` and `reference_base`) are defined in
+  `env.py`; there is no `tasks._grade`, so the script above does not apply.
+  Not built or run here.
+
+**Packaged, not yet in `tasks/`.** `badger`, `bazel-remote`, `celery`, `geode`, `kratos`, `microblogpub`, `open-xml-sdk`, `openfire`, `terser`, `umya`, `zettlr`: its build files were given a neutral environment name; the package is being rebuilt and re-checked on them.
+

harbor/build_images.py

@@ -0,0 +1,79 @@
+"""Build the task source images from this folder.
+
+    python build_images.py                    every packaged task
+    python build_images.py --task <task id>   one task (repeatable)
+
+A task without a package is built only when it is named: its id is the short
+name under "unpackaged" in manifest.json (README, "Tasks without a Harbor
+package").
+
+Each image is built from the task's original build context under sources/ and
+tagged with the name its Harbor task expects. A build downloads its base image
+and the repository's dependencies, so it needs network access; nothing is
+downloaded when tasks run. Most builds end with the task's own calibration:
+the reference fix must score 1.0 and the untouched repository 0.0 or the image
+is not produced. Expect ten to forty minutes and several gigabytes per task.
+
+The images were qualified from build contexts in which every file had mode 777
+(a Windows disk seen from Linux), and some images keep the modes of files they
+copy. A download gives files mode 644. So that your images match the qualified
+ones, each build runs from a temporary copy of the context with mode 777; the
+files under sources/ are left as they are.
+"""
+
+import argparse
+import json
+import os
+import shutil
+import subprocess
+import tempfile
+import time
+from pathlib import Path
+
+parser = argparse.ArgumentParser(description=__doc__, formatter_class=argparse.RawDescriptionHelpFormatter)
+parser.add_argument("--task", action="append", default=[])
+parser.add_argument("--docker-context")
+args = parser.parse_args()
+root = Path(__file__).resolve().parent
+manifest = json.loads((root / "manifest.json").read_text())
+context = args.docker_context or ("desktop-linux" if os.name == "nt" else None)
+docker = ["docker", *(["--context", context] if context else [])]
+known = {task["task_id"]: task for task in manifest["tasks"]}
+extra = {task["task_id"]: task for task in manifest.get("unpackaged", []) if task["source_folder"]}
+unknown = [name for name in args.task if name not in known and name not in extra]
+if unknown:
+    raise SystemExit("Unknown task: " + ", ".join(unknown) + ". See manifest.json.")
+known.update({name: extra[name] for name in args.task if name in extra})
+logs = root / "build-logs"
+logs.mkdir(exist_ok=True)
+built = {}
+for task_id, task in known.items():
+    if args.task and task_id not in args.task:
+        continue
+    source = root / "sources" / task["source_folder"]
+    scratch = None
+    if os.name != "nt":
+        scratch = Path(tempfile.mkdtemp(prefix="ec-swe-build-"))
+        source = Path(shutil.copytree(source, scratch / source.name))
+        for path in [source, *source.rglob("*")]:
+            path.chmod(0o777)
+    log = logs / f"{task_id}.log"
+    started = time.time()
+    print(f"building {task['source_image']} (log: {log.relative_to(root)})", flush=True)
+    with log.open("wb") as handle:
+        code = subprocess.run(
+            [*docker, "build", "--progress=plain", "-f", str(source / task["dockerfile"]),
+             "-t", task["source_image"], str(source)],
+            stdout=handle, stderr=subprocess.STDOUT,
+        ).returncode
+    if scratch:
+        shutil.rmtree(scratch, ignore_errors=True)
+    if code:
+        raise SystemExit(f"Build failed for {task_id}; see {log}")
+    image_id = subprocess.check_output(
+        [*docker, "image", "inspect", task["source_image"], "--format", "{{.Id}}"], text=True
+    ).strip()
+    built[task_id] = {"image": task["source_image"], "id": image_id, "seconds": round(time.time() - started)}
+    print(f"  done in {built[task_id]['seconds']}s: {image_id}", flush=True)
+(root / "built-images.json").write_text(json.dumps(built, indent=2) + "\n")
+print(f"Built {len(built)} task image(s)")

harbor/evidence/coredns/agent-image-scan.txt

@@ -0,0 +1,14 @@
+needles: 71
+1 matching line(s): /workspace/repo/plugin/rewrite/ttl_test.go
+scan finished
+absent: /hud
+absent: /app
+absent: /tests
+present: /opt/task-tools/baseline.git
+reference-patch lines in the agent's tracked files: 0, of which the pristine checkout lacks: 0
+grader lines in the agent's tracked files: 1, of which the pristine checkout lacks: 0
+reference-patch lines in total: 9, grader lines in total: 60
+private lines in 0 other files outside the checkout: 0, of which upstream lacks: 0
+--- private lines only the agent's checkout holds (expect none)
+--- the agent's checkout: uncommitted changes (expect none), then its history
+b21e6b7 baseline

harbor/evidence/coredns/native-reference.json

@@ -0,0 +1,121 @@
+{
+  "reward": 1.0,
+  "done": true,
+  "content": null,
+  "info": {},
+  "isError": false,
+  "subscores": [
+    {
+      "name": "svcb_target_rewrite",
+      "weight": 0.25,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "CRITERION svcb_target_rewrite: PASS\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "bash /hud/grader/run_grading.sh svcb_target_rewrite",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 90
+        }
+      }
+    },
+    {
+      "name": "https_target_rewrite",
+      "weight": 0.2,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "CRITERION https_target_rewrite: PASS\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "bash /hud/grader/run_grading.sh https_target_rewrite",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 90
+        }
+      }
+    },
+    {
+      "name": "soa_mailbox_rewrite",
+      "weight": 0.3,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "CRITERION soa_mailbox_rewrite: PASS\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "bash /hud/grader/run_grading.sh soa_mailbox_rewrite",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 90
+        }
+      }
+    },
+    {
+      "name": "opt_ttl_preserved",
+      "weight": 0.25,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "CRITERION opt_ttl_preserved: PASS\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "bash /hud/grader/run_grading.sh opt_ttl_preserved",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 90
+        }
+      }
+    },
+    {
+      "name": "existing_name_value_rewrite_unaffected",
+      "weight": 0.0,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "CRITERION existing_name_value_rewrite_unaffected: PASS\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "bash /hud/grader/run_grading.sh existing_name_value_rewrite_unaffected",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 90
+        }
+      }
+    },
+    {
+      "name": "ordinary_ttl_clamp_unaffected",
+      "weight": 0.0,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "CRITERION ordinary_ttl_clamp_unaffected: PASS\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "bash /hud/grader/run_grading.sh ordinary_ttl_clamp_unaffected",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 90
+        }
+      }
+    },
+    {
+      "name": "full_suite_regression",
+      "weight": 0.0,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "CRITERION full_suite_regression: PASS\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "bash /hud/grader/run_grading.sh full_suite_regression",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 180
+        }
+      }
+    }
+  ]
+}

harbor/evidence/coredns/reports/broken.json

@@ -0,0 +1,242 @@
+{
+  "revision": "coding-rubric-v2",
+  "task_id": "coredns_rewrite_svcb_opt",
+  "grader_version": "0.6.18",
+  "commands": [
+    {
+      "name": "svcb_target_rewrite",
+      "weight": 0.25,
+      "value": 0.0,
+      "children": null,
+      "info": {
+        "exit_code": 1,
+        "stdout": "  ----- go test -run ^TestSVCBTargetRewrittenBySuffixAutoAnswerRule$ -v ./plugin/rewrite/... (last 2000 bytes) -----\n  | # github.com/coredns/coredns/plugin/rewrite [github.com/coredns/coredns/plugin/rewrite.test]\n  | plugin/rewrite/name.go:466:1: syntax error: non-declaration statement outside function body\n  | FAIL\tgithub.com/coredns/coredns/plugin/rewrite [build failed]\n  | FAIL\n  -----\nCRITERION svcb_target_rewrite: FAIL\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh svcb_target_rewrite'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 90
+        }
+      }
+    },
+    {
+      "name": "https_target_rewrite",
+      "weight": 0.2,
+      "value": 0.0,
+      "children": null,
+      "info": {
+        "exit_code": 1,
+        "stdout": "  ----- go test -run ^TestHTTPSTargetRewrittenBySuffixAutoAnswerRule$ -v ./plugin/rewrite/... (last 2000 bytes) -----\n  | # github.com/coredns/coredns/plugin/rewrite [github.com/coredns/coredns/plugin/rewrite.test]\n  | plugin/rewrite/name.go:466:1: syntax error: non-declaration statement outside function body\n  | FAIL\tgithub.com/coredns/coredns/plugin/rewrite [build failed]\n  | FAIL\n  -----\nCRITERION https_target_rewrite: FAIL\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh https_target_rewrite'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 90
+        }
+      }
+    },
+    {
+      "name": "soa_mailbox_rewrite",
+      "weight": 0.3,
+      "value": 0.0,
+      "children": null,
+      "info": {
+        "exit_code": 1,
+        "stdout": "  ----- go test -run ^TestSOAMailboxRewrittenBySuffixAutoAnswerRule$ -v ./plugin/rewrite/... (last 2000 bytes) -----\n  | # github.com/coredns/coredns/plugin/rewrite [github.com/coredns/coredns/plugin/rewrite.test]\n  | plugin/rewrite/name.go:466:1: syntax error: non-declaration statement outside function body\n  | FAIL\tgithub.com/coredns/coredns/plugin/rewrite [build failed]\n  | FAIL\n  -----\nCRITERION soa_mailbox_rewrite: FAIL\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh soa_mailbox_rewrite'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 90
+        }
+      }
+    },
+    {
+      "name": "opt_ttl_preserved",
+      "weight": 0.25,
+      "value": 0.0,
+      "children": null,
+      "info": {
+        "exit_code": 1,
+        "stdout": "  ----- go test -run ^TestTTLRuleDoesNotClampOPTPseudoRecord$ -v ./plugin/rewrite/... (last 2000 bytes) -----\n  | # github.com/coredns/coredns/plugin/rewrite [github.com/coredns/coredns/plugin/rewrite.test]\n  | plugin/rewrite/name.go:466:1: syntax error: non-declaration statement outside function body\n  | FAIL\tgithub.com/coredns/coredns/plugin/rewrite [build failed]\n  | FAIL\n  -----\nCRITERION opt_ttl_preserved: FAIL\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh opt_ttl_preserved'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 90
+        }
+      }
+    },
+    {
+      "name": "existing_name_value_rewrite_unaffected",
+      "weight": 0.0,
+      "value": 0.0,
+      "children": null,
+      "info": {
+        "exit_code": 1,
+        "stdout": "  ----- go test -run ^TestValueResponseReverter$ -v ./plugin/rewrite/... (last 2000 bytes) -----\n  | # github.com/coredns/coredns/plugin/rewrite [github.com/coredns/coredns/plugin/rewrite.test]\n  | plugin/rewrite/name.go:466:1: syntax error: non-declaration statement outside function body\n  | FAIL\tgithub.com/coredns/coredns/plugin/rewrite [build failed]\n  | FAIL\n  -----\nCRITERION existing_name_value_rewrite_unaffected: FAIL\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh existing_name_value_rewrite_unaffected'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 90
+        }
+      }
+    },
+    {
+      "name": "ordinary_ttl_clamp_unaffected",
+      "weight": 0.0,
+      "value": 0.0,
+      "children": null,
+      "info": {
+        "exit_code": 1,
+        "stdout": "  ----- go test -run ^TestTtlRewrite$ -v ./plugin/rewrite/... (last 2000 bytes) -----\n  | # github.com/coredns/coredns/plugin/rewrite [github.com/coredns/coredns/plugin/rewrite.test]\n  | plugin/rewrite/name.go:466:1: syntax error: non-declaration statement outside function body\n  | FAIL\tgithub.com/coredns/coredns/plugin/rewrite [build failed]\n  | FAIL\n  -----\nCRITERION ordinary_ttl_clamp_unaffected: FAIL\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh ordinary_ttl_clamp_unaffected'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 90
+        }
+      }
+    },
+    {
+      "name": "full_suite_regression",
+      "weight": 0.0,
+      "value": 0.0,
+      "children": null,
+      "info": {
+        "exit_code": 1,
+        "stdout": "  ----- go test ./plugin/rewrite/... (last 2000 bytes) -----\n  | # github.com/coredns/coredns/plugin/rewrite [github.com/coredns/coredns/plugin/rewrite.test]\n  | plugin/rewrite/name.go:466:1: syntax error: non-declaration statement outside function body\n  | FAIL\tgithub.com/coredns/coredns/plugin/rewrite [build failed]\n  | FAIL\n  -----\nCRITERION full_suite_regression: FAIL\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh full_suite_regression'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 180
+        }
+      }
+    }
+  ],
+  "status": "graded",
+  "patch_sha256": "aa144710c3f317d2e0d59de1514c2759ce4f896e3838d8bd45b9d21200843de8",
+  "result": {
+    "reward": 0.0,
+    "done": true,
+    "content": null,
+    "info": {},
+    "isError": false,
+    "subscores": [
+      {
+        "name": "svcb_target_rewrite",
+        "weight": 0.25,
+        "value": 0.0,
+        "children": null,
+        "info": {
+          "exit_code": 1,
+          "stdout": "  ----- go test -run ^TestSVCBTargetRewrittenBySuffixAutoAnswerRule$ -v ./plugin/rewrite/... (last 2000 bytes) -----\n  | # github.com/coredns/coredns/plugin/rewrite [github.com/coredns/coredns/plugin/rewrite.test]\n  | plugin/rewrite/name.go:466:1: syntax error: non-declaration statement outside function body\n  | FAIL\tgithub.com/coredns/coredns/plugin/rewrite [build failed]\n  | FAIL\n  -----\nCRITERION svcb_target_rewrite: FAIL\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh svcb_target_rewrite'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 90
+          }
+        }
+      },
+      {
+        "name": "https_target_rewrite",
+        "weight": 0.2,
+        "value": 0.0,
+        "children": null,
+        "info": {
+          "exit_code": 1,
+          "stdout": "  ----- go test -run ^TestHTTPSTargetRewrittenBySuffixAutoAnswerRule$ -v ./plugin/rewrite/... (last 2000 bytes) -----\n  | # github.com/coredns/coredns/plugin/rewrite [github.com/coredns/coredns/plugin/rewrite.test]\n  | plugin/rewrite/name.go:466:1: syntax error: non-declaration statement outside function body\n  | FAIL\tgithub.com/coredns/coredns/plugin/rewrite [build failed]\n  | FAIL\n  -----\nCRITERION https_target_rewrite: FAIL\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh https_target_rewrite'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 90
+          }
+        }
+      },
+      {
+        "name": "soa_mailbox_rewrite",
+        "weight": 0.3,
+        "value": 0.0,
+        "children": null,
+        "info": {
+          "exit_code": 1,
+          "stdout": "  ----- go test -run ^TestSOAMailboxRewrittenBySuffixAutoAnswerRule$ -v ./plugin/rewrite/... (last 2000 bytes) -----\n  | # github.com/coredns/coredns/plugin/rewrite [github.com/coredns/coredns/plugin/rewrite.test]\n  | plugin/rewrite/name.go:466:1: syntax error: non-declaration statement outside function body\n  | FAIL\tgithub.com/coredns/coredns/plugin/rewrite [build failed]\n  | FAIL\n  -----\nCRITERION soa_mailbox_rewrite: FAIL\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh soa_mailbox_rewrite'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 90
+          }
+        }
+      },
+      {
+        "name": "opt_ttl_preserved",
+        "weight": 0.25,
+        "value": 0.0,
+        "children": null,
+        "info": {
+          "exit_code": 1,
+          "stdout": "  ----- go test -run ^TestTTLRuleDoesNotClampOPTPseudoRecord$ -v ./plugin/rewrite/... (last 2000 bytes) -----\n  | # github.com/coredns/coredns/plugin/rewrite [github.com/coredns/coredns/plugin/rewrite.test]\n  | plugin/rewrite/name.go:466:1: syntax error: non-declaration statement outside function body\n  | FAIL\tgithub.com/coredns/coredns/plugin/rewrite [build failed]\n  | FAIL\n  -----\nCRITERION opt_ttl_preserved: FAIL\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh opt_ttl_preserved'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 90
+          }
+        }
+      },
+      {
+        "name": "existing_name_value_rewrite_unaffected",
+        "weight": 0.0,
+        "value": 0.0,
+        "children": null,
+        "info": {
+          "exit_code": 1,
+          "stdout": "  ----- go test -run ^TestValueResponseReverter$ -v ./plugin/rewrite/... (last 2000 bytes) -----\n  | # github.com/coredns/coredns/plugin/rewrite [github.com/coredns/coredns/plugin/rewrite.test]\n  | plugin/rewrite/name.go:466:1: syntax error: non-declaration statement outside function body\n  | FAIL\tgithub.com/coredns/coredns/plugin/rewrite [build failed]\n  | FAIL\n  -----\nCRITERION existing_name_value_rewrite_unaffected: FAIL\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh existing_name_value_rewrite_unaffected'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 90
+          }
+        }
+      },
+      {
+        "name": "ordinary_ttl_clamp_unaffected",
+        "weight": 0.0,
+        "value": 0.0,
+        "children": null,
+        "info": {
+          "exit_code": 1,
+          "stdout": "  ----- go test -run ^TestTtlRewrite$ -v ./plugin/rewrite/... (last 2000 bytes) -----\n  | # github.com/coredns/coredns/plugin/rewrite [github.com/coredns/coredns/plugin/rewrite.test]\n  | plugin/rewrite/name.go:466:1: syntax error: non-declaration statement outside function body\n  | FAIL\tgithub.com/coredns/coredns/plugin/rewrite [build failed]\n  | FAIL\n  -----\nCRITERION ordinary_ttl_clamp_unaffected: FAIL\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh ordinary_ttl_clamp_unaffected'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 90
+          }
+        }
+      },
+      {
+        "name": "full_suite_regression",
+        "weight": 0.0,
+        "value": 0.0,
+        "children": null,
+        "info": {
+          "exit_code": 1,
+          "stdout": "  ----- go test ./plugin/rewrite/... (last 2000 bytes) -----\n  | # github.com/coredns/coredns/plugin/rewrite [github.com/coredns/coredns/plugin/rewrite.test]\n  | plugin/rewrite/name.go:466:1: syntax error: non-declaration statement outside function body\n  | FAIL\tgithub.com/coredns/coredns/plugin/rewrite [build failed]\n  | FAIL\n  -----\nCRITERION full_suite_regression: FAIL\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh full_suite_regression'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 180
+          }
+        }
+      }
+    ]
+  }
+}

harbor/evidence/coredns/reports/infra.json

@@ -0,0 +1,12 @@
+{
+  "revision": "coding-rubric-v2",
+  "task_id": "coredns_rewrite_svcb_opt",
+  "grader_version": "0.6.18",
+  "commands": [],
+  "status": "error",
+  "patch_sha256": "51f9168417f7ef7d6b8793dcf746d6bf88df29ce28a191190f08f39260727ae7",
+  "error": {
+    "type": "FileNotFoundError",
+    "message": "[Errno 2] No such file or directory: '/hud/grader/run_grading.sh'"
+  }
+}

harbor/evidence/coredns/reports/noop.json

@@ -0,0 +1,16 @@
+{
+  "revision": "coding-rubric-v2",
+  "task_id": "coredns_rewrite_svcb_opt",
+  "grader_version": "0.6.18",
+  "commands": [],
+  "status": "empty_patch",
+  "patch_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
+  "result": {
+    "reward": 0.0,
+    "done": true,
+    "content": "Empty diff: no changes were made.",
+    "info": {},
+    "isError": false,
+    "subscores": null
+  }
+}

harbor/evidence/coredns/reports/reference.json

@@ -0,0 +1,242 @@
+{
+  "revision": "coding-rubric-v2",
+  "task_id": "coredns_rewrite_svcb_opt",
+  "grader_version": "0.6.18",
+  "commands": [
+    {
+      "name": "svcb_target_rewrite",
+      "weight": 0.25,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "CRITERION svcb_target_rewrite: PASS\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh svcb_target_rewrite'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 90
+        }
+      }
+    },
+    {
+      "name": "https_target_rewrite",
+      "weight": 0.2,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "CRITERION https_target_rewrite: PASS\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh https_target_rewrite'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 90
+        }
+      }
+    },
+    {
+      "name": "soa_mailbox_rewrite",
+      "weight": 0.3,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "CRITERION soa_mailbox_rewrite: PASS\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh soa_mailbox_rewrite'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 90
+        }
+      }
+    },
+    {
+      "name": "opt_ttl_preserved",
+      "weight": 0.25,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "CRITERION opt_ttl_preserved: PASS\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh opt_ttl_preserved'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 90
+        }
+      }
+    },
+    {
+      "name": "existing_name_value_rewrite_unaffected",
+      "weight": 0.0,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "CRITERION existing_name_value_rewrite_unaffected: PASS\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh existing_name_value_rewrite_unaffected'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 90
+        }
+      }
+    },
+    {
+      "name": "ordinary_ttl_clamp_unaffected",
+      "weight": 0.0,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "CRITERION ordinary_ttl_clamp_unaffected: PASS\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh ordinary_ttl_clamp_unaffected'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 90
+        }
+      }
+    },
+    {
+      "name": "full_suite_regression",
+      "weight": 0.0,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "CRITERION full_suite_regression: PASS\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh full_suite_regression'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 180
+        }
+      }
+    }
+  ],
+  "status": "graded",
+  "patch_sha256": "51f9168417f7ef7d6b8793dcf746d6bf88df29ce28a191190f08f39260727ae7",
+  "result": {
+    "reward": 1.0,
+    "done": true,
+    "content": null,
+    "info": {},
+    "isError": false,
+    "subscores": [
+      {
+        "name": "svcb_target_rewrite",
+        "weight": 0.25,
+        "value": 1.0,
+        "children": null,
+        "info": {
+          "exit_code": 0,
+          "stdout": "CRITERION svcb_target_rewrite: PASS\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh svcb_target_rewrite'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 90
+          }
+        }
+      },
+      {
+        "name": "https_target_rewrite",
+        "weight": 0.2,
+        "value": 1.0,
+        "children": null,
+        "info": {
+          "exit_code": 0,
+          "stdout": "CRITERION https_target_rewrite: PASS\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh https_target_rewrite'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 90
+          }
+        }
+      },
+      {
+        "name": "soa_mailbox_rewrite",
+        "weight": 0.3,
+        "value": 1.0,
+        "children": null,
+        "info": {
+          "exit_code": 0,
+          "stdout": "CRITERION soa_mailbox_rewrite: PASS\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh soa_mailbox_rewrite'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 90
+          }
+        }
+      },
+      {
+        "name": "opt_ttl_preserved",
+        "weight": 0.25,
+        "value": 1.0,
+        "children": null,
+        "info": {
+          "exit_code": 0,
+          "stdout": "CRITERION opt_ttl_preserved: PASS\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh opt_ttl_preserved'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 90
+          }
+        }
+      },
+      {
+        "name": "existing_name_value_rewrite_unaffected",
+        "weight": 0.0,
+        "value": 1.0,
+        "children": null,
+        "info": {
+          "exit_code": 0,
+          "stdout": "CRITERION existing_name_value_rewrite_unaffected: PASS\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh existing_name_value_rewrite_unaffected'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 90
+          }
+        }
+      },
+      {
+        "name": "ordinary_ttl_clamp_unaffected",
+        "weight": 0.0,
+        "value": 1.0,
+        "children": null,
+        "info": {
+          "exit_code": 0,
+          "stdout": "CRITERION ordinary_ttl_clamp_unaffected: PASS\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh ordinary_ttl_clamp_unaffected'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 90
+          }
+        }
+      },
+      {
+        "name": "full_suite_regression",
+        "weight": 0.0,
+        "value": 1.0,
+        "children": null,
+        "info": {
+          "exit_code": 0,
+          "stdout": "CRITERION full_suite_regression: PASS\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh full_suite_regression'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 180
+          }
+        }
+      }
+    ]
+  }
+}

harbor/evidence/coredns/reports/tamper.json

@@ -0,0 +1,16 @@
+{
+  "revision": "coding-rubric-v2",
+  "task_id": "coredns_rewrite_svcb_opt",
+  "grader_version": "0.6.18",
+  "commands": [],
+  "status": "empty_patch",
+  "patch_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
+  "result": {
+    "reward": 0.0,
+    "done": true,
+    "content": "Empty diff: no changes were made.",
+    "info": {},
+    "isError": false,
+    "subscores": null
+  }
+}

harbor/evidence/coredns/summary.json

@@ -0,0 +1,172 @@
+{
+  "name": "coredns",
+  "task_id": "coredns_rewrite_svcb_opt",
+  "image": "ec-swe-coredns:source-v4",
+  "build": {
+    "ok": true,
+    "image_id": "sha256:5f122f58a039aefe65ccf2fd057eceb288c09278edde20f2ff0f2598f52c22e9",
+    "ladder": {},
+    "seconds": 0
+  },
+  "qualified": true,
+  "measure": {
+    "ok": true,
+    "hud": "0.6.18",
+    "path": "/usr/local/bin:/usr/bin:/bin:/usr/local/games:/usr/games",
+    "handed": [],
+    "repo_dir": "/workspace/repo",
+    "python_usr_local": false,
+    "seconds": 14,
+    "reused": true
+  },
+  "parity": {
+    "ok": true,
+    "reward": 1.0,
+    "criteria": 7,
+    "unmet": [],
+    "info": null,
+    "seconds": 25,
+    "reused": true
+  },
+  "prepare": {
+    "ok": true,
+    "bundle": "coredns-r4",
+    "converter": "b8bba68767da888f",
+    "pinned_files": 9,
+    "criteria": 7,
+    "seconds": 2
+  },
+  "controls": {
+    "ok": true,
+    "checks": {
+      "noop_is_zero": true,
+      "reference_is_one": true,
+      "reference_equals_native": true,
+      "broken_is_graded_zero": true,
+      "missing_grader_is_invalid": true,
+      "agent_is_isolated": true,
+      "planted_reward_is_ignored": true
+    },
+    "results": {
+      "noop": {
+        "run": "coredns-r4-control-noop-1791885502",
+        "reward": 0.0,
+        "stop": "completed",
+        "error": null,
+        "seconds": 175,
+        "probe": false,
+        "status": "empty_patch",
+        "verifier_error": null,
+        "exit_codes": [],
+        "attempts": 1
+      },
+      "reference": {
+        "run": "coredns-r4-control-reference-1791885502",
+        "reward": 1.0,
+        "stop": "completed",
+        "error": null,
+        "seconds": 185,
+        "probe": true,
+        "status": "graded",
+        "verifier_error": null,
+        "exit_codes": [
+          0
+        ],
+        "subscores": {
+          "svcb_target_rewrite": [
+            1.0,
+            0.25
+          ],
+          "https_target_rewrite": [
+            1.0,
+            0.2
+          ],
+          "soa_mailbox_rewrite": [
+            1.0,
+            0.3
+          ],
+          "opt_ttl_preserved": [
+            1.0,
+            0.25
+          ],
+          "existing_name_value_rewrite_unaffected": [
+            1.0,
+            0.0
+          ],
+          "ordinary_ttl_clamp_unaffected": [
+            1.0,
+            0.0
+          ],
+          "full_suite_regression": [
+            1.0,
+            0.0
+          ]
+        },
+        "attempts": 1
+      },
+      "broken": {
+        "run": "coredns-r4-control-broken-1791885678",
+        "reward": 0.0,
+        "stop": "completed",
+        "error": null,
+        "seconds": 111,
+        "probe": false,
+        "status": "graded",
+        "verifier_error": null,
+        "exit_codes": [
+          1
+        ],
+        "attempts": 1
+      },
+      "infra": {
+        "run": "coredns-r4-control-infra-1791885688",
+        "reward": null,
+        "stop": "error",
+        "error": "RewardFileNotFoundError: No reward file found at <jobs>/jobs/coredns-r4-control-infra-1791885688/coredns_rewrite_svcb_opt__8gHWTyF/verifier/reward.txt or <jobs>/jobs/coredns-r4-control-infra-1791885688/coredns_rewrite_svcb_opt__8gHWTyF/verifier/reward.json",
+        "seconds": 92,
+        "probe": false,
+        "status": "error",
+        "verifier_error": "[Errno 2] No such file or directory: '/hud/grader/run_grading.sh'",
+        "exit_codes": [],
+        "attempts": 1
+      },
+      "tamper": {
+        "run": "coredns-r4-control-tamper-1791885780",
+        "reward": 0.0,
+        "stop": "completed",
+        "error": null,
+        "seconds": 104,
+        "probe": false,
+        "status": "empty_patch",
+        "verifier_error": null,
+        "exit_codes": [],
+        "attempts": 1
+      }
+    },
+    "bundle": "coredns-r4",
+    "seconds": 383
+  },
+  "patches": {
+    "ok": true,
+    "skipped": "no calibration patches in the source",
+    "seconds": 0
+  },
+  "leak": {
+    "scan_finished": true,
+    "hits_outside_checkout": [],
+    "other_files_holding_only_upstream_lines": 0,
+    "library_files_sharing_a_line": 0,
+    "most_lines_shared_by_one_library_file": 0,
+    "private_paths_absent": true,
+    "public_lines_shared_with_patch_and_graders": [
+      0,
+      1
+    ],
+    "private_lines_only_in_agent_checkout": [
+      0,
+      0
+    ],
+    "ok": true,
+    "seconds": 31
+  }
+}

harbor/evidence/epicyon/agent-image-scan.txt

@@ -0,0 +1,26 @@
+needles: 391
+1 matching line(s): /workspace/repo/src/webapp_hashtagswarm.py
+5 matching line(s): /workspace/repo/src/webapp_utils.py
+1 matching line(s): /workspace/repo/src/gemini.py
+1 matching line(s): /workspace/repo/src/webapp_column_right.py
+8 matching line(s): /workspace/repo/src/utils.py
+1 matching line(s): /workspace/repo/src/webapp_post.py
+2 matching line(s): /workspace/repo/src/person.py
+2 matching line(s): /workspace/repo/src/markdown.py
+1 matching line(s): /workspace/repo/src/webapp_timeline.py
+3 matching line(s): /workspace/repo/src/webapp_search.py
+2 matching line(s): /workspace/repo/src/webapp_profile.py
+1 matching line(s): /workspace/repo/src/daemon_get_collections.py
+1 matching line(s): /workspace/repo/src/follow.py
+scan finished
+absent: /hud
+absent: /app
+absent: /tests
+present: /opt/task-tools/baseline.git
+reference-patch lines in the agent's tracked files: 6, of which the pristine checkout lacks: 0
+grader lines in the agent's tracked files: 0, of which the pristine checkout lacks: 0
+reference-patch lines in total: 129, grader lines in total: 251
+private lines in 0 other files outside the checkout: 0, of which upstream lacks: 0
+--- private lines only the agent's checkout holds (expect none)
+--- the agent's checkout: uncommitted changes (expect none), then its history
+54dfc93 baseline

harbor/evidence/epicyon/author-calibration.txt

@@ -0,0 +1,188 @@
+=== gold ===
+  PASS  media_file_of_a_deleted_post_is_removed                w=0.0550 blocker    8.9s
+        PASS: media_file_of_a_deleted_post_is_removed — 2 fresh runs agree
+  PASS  hashtag_index_forgets_a_deleted_post                   w=0.0550            0.0s
+        PASS: hashtag_index_forgets_a_deleted_post — 2 fresh runs agree
+  PASS  a_deleted_blog_post_leaves_no_rendition                w=0.0550            0.0s
+        PASS: a_deleted_blog_post_leaves_no_rendition — 2 fresh runs agree
+  PASS  deleted_post_is_delisted_from_the_media_log            w=0.0550            0.0s
+        PASS: deleted_post_is_delisted_from_the_media_log — 2 fresh runs agree
+  PASS  a_deleted_blog_post_leaves_no_json_or_attachment       w=0.0550            0.0s
+        PASS: a_deleted_blog_post_leaves_no_json_or_attachment — 2 fresh runs agree
+  PASS  timeline_indexes_forget_a_deleted_post                 w=0.0550            0.0s
+        PASS: timeline_indexes_forget_a_deleted_post — 2 fresh runs agree
+  PASS  features_copy_of_a_deleted_article_is_removed          w=0.1600            0.0s
+        PASS: features_copy_of_a_deleted_article_is_removed — 2 fresh runs agree
+  PASS  features_index_forgets_a_deleted_article               w=0.1600            0.0s
+        PASS: features_index_forgets_a_deleted_article — 2 fresh runs agree
+  PASS  event_category_index_forgets_a_deleted_event           w=0.1600            0.0s
+        PASS: event_category_index_forgets_a_deleted_event — 2 fresh runs agree
+  PASS  split_event_category_indexes_forget_a_deleted_event    w=0.1600            0.0s
+        PASS: split_event_category_indexes_forget_a_deleted_event — 2 fresh runs agree
+  ... calibration still running (12:40:09) ...
+  PASS  gate_repo_suite_green_and_tests_untouched              w=0.0300 blocker   99.5s CORE_GREEN=1
+        PASS: the repository's own checks are all there and its suite is green
+  -> uncapped 1.000, failed blockers none, reward 1.000
+
+=== base ===
+  ... calibration still running (12:41:09) ...
+  FAIL  media_file_of_a_deleted_post_is_removed                w=0.0550 blocker    9.1s
+        FAIL: the deleted post's media is still on disk: /media/2962/PmShgwJNlHh9vzi0oYe2DvamtaXhy4NK.png
+  FAIL  hashtag_index_forgets_a_deleted_post                   w=0.0550            0.0s
+        FAIL: the hashtag index still points at the deleted post
+  FAIL  a_deleted_blog_post_leaves_no_rendition                w=0.0550            0.0s
+        FAIL: the deleted blog post is still published as markdown, micron
+  FAIL  deleted_post_is_delisted_from_the_media_log            w=0.0550            0.0s
+        FAIL: the media log still lists the deleted post's media: /media/2962/VYXCsOVxGn9aXtE6hXCQSEOAtXlGFMtq.png
+  FAIL  a_deleted_blog_post_leaves_no_json_or_attachment       w=0.0550            0.0s
+        FAIL: the deleted blog post's own json is still there
+  FAIL  timeline_indexes_forget_a_deleted_post                 w=0.0550            0.0s
+        FAIL: the outbox index still lists the deleted post
+  FAIL  features_copy_of_a_deleted_article_is_removed          w=0.1600            0.0s
+        FAIL: the news actor still holds a copy of the deleted article
+  FAIL  features_index_forgets_a_deleted_article               w=0.1600            0.0s
+        FAIL: the news actor's index still lists the deleted article
+  FAIL  event_category_index_forgets_a_deleted_event           w=0.1600            0.0s
+        FAIL: the Music tag index still lists the deleted event
+  FAIL  split_event_category_indexes_forget_a_deleted_event    w=0.1600            0.0s
+        FAIL: the Outdoors tag index still lists the deleted event
+  ... calibration still running (12:42:09) ...
+  FAIL  gate_repo_suite_green_and_tests_untouched              w=0.0300 blocker   98.2s CORE_GREEN=0
+        FAIL: no scored behaviour was fixed — this gate does not pay for an empty change
+  (conditional: gate_repo_suite_green_and_tests_untouched was told the core is dead and refused its weight itself; it reports FAIL and exits non-zero)
+  -> uncapped 0.000, failed blockers ['media_file_of_a_deleted_post_is_removed', 'gate_repo_suite_green_and_tests_untouched'], reward 0.000
+
+=== noop ===
+  FAIL  media_file_of_a_deleted_post_is_removed                w=0.0550 blocker    9.0s
+        FAIL: the deleted post's media is still on disk: /media/2962/nf7xpIWDBCvhJk9tvhysMdZZXQyBebMZ.png
+  FAIL  hashtag_index_forgets_a_deleted_post                   w=0.0550            0.0s
+        FAIL: the hashtag index still points at the deleted post
+  FAIL  a_deleted_blog_post_leaves_no_rendition                w=0.0550            0.0s
+        FAIL: the deleted blog post is still published as markdown, micron
+  FAIL  deleted_post_is_delisted_from_the_media_log            w=0.0550            0.0s
+        FAIL: the media log still lists the deleted post's media: /media/2962/OXkJVwcseZBktiqdtA1YCSPU3Vb2Q6x4.png
+  FAIL  a_deleted_blog_post_leaves_no_json_or_attachment       w=0.0550            0.0s
+        FAIL: the deleted blog post's own json is still there
+  FAIL  timeline_indexes_forget_a_deleted_post                 w=0.0550            0.0s
+        FAIL: the outbox index still lists the deleted post
+  FAIL  features_copy_of_a_deleted_article_is_removed          w=0.1600            0.0s
+        FAIL: the news actor still holds a copy of the deleted article
+  FAIL  features_index_forgets_a_deleted_article               w=0.1600            0.0s
+        FAIL: the news actor's index still lists the deleted article
+  FAIL  event_category_index_forgets_a_deleted_event           w=0.1600            0.0s
+        FAIL: the Music tag index still lists the deleted event
+  FAIL  split_event_category_indexes_forget_a_deleted_event    w=0.1600            0.0s
+        FAIL: the Outdoors tag index still lists the deleted event
+  ... calibration still running (12:43:09) ...
+  ... calibration still running (12:44:09) ...
+  FAIL  gate_repo_suite_green_and_tests_untouched              w=0.0300 blocker   97.7s CORE_GREEN=0
+        FAIL: no scored behaviour was fixed — this gate does not pay for an empty change
+  (conditional: gate_repo_suite_green_and_tests_untouched was told the core is dead and refused its weight itself; it reports FAIL and exits non-zero)
+  -> uncapped 0.000, failed blockers ['media_file_of_a_deleted_post_is_removed', 'gate_repo_suite_green_and_tests_untouched'], reward 0.000
+
+=== lazy ===
+  FAIL  media_file_of_a_deleted_post_is_removed                w=0.0550 blocker    8.8s
+        FAIL: another post's media was removed too: /media/2962/S0ERe3NbiHgjFwDs9XA1aDIqAHKSiduC.png
+  FAIL  hashtag_index_forgets_a_deleted_post                   w=0.0550            0.0s
+        FAIL: the hashtag index still points at the deleted post
+  FAIL  a_deleted_blog_post_leaves_no_rendition                w=0.0550            0.0s
+        FAIL: the deleted blog post is still published as markdown, micron
+  FAIL  deleted_post_is_delisted_from_the_media_log            w=0.0550            0.0s
+        FAIL: the media log still lists the deleted post's media: /media/2962/HlzRTOE90XNJdTWuocCt6eTJKgu2ZqRq.png
+  FAIL  a_deleted_blog_post_leaves_no_json_or_attachment       w=0.0550            0.0s
+        FAIL: the deleted blog post's own json is still there
+  PASS  timeline_indexes_forget_a_deleted_post                 w=0.0550            0.0s
+        PASS: timeline_indexes_forget_a_deleted_post — 2 fresh runs agree
+  FAIL  features_copy_of_a_deleted_article_is_removed          w=0.1600            0.0s
+        FAIL: the news actor still holds a copy of the deleted article
+  FAIL  features_index_forgets_a_deleted_article               w=0.1600            0.0s
+        FAIL: the news actor's index still lists the deleted article
+  FAIL  event_category_index_forgets_a_deleted_event           w=0.1600            0.0s
+        FAIL: the Music tag index still lists the deleted event
+  FAIL  split_event_category_indexes_forget_a_deleted_event    w=0.1600            0.0s
+        FAIL: the Outdoors tag index still lists the deleted event
+  ... calibration still running (12:45:09) ...
+  ... calibration still running (12:46:09) ...
+  PASS  gate_repo_suite_green_and_tests_untouched              w=0.0300 blocker   96.1s CORE_GREEN=1
+        PASS: the repository's own checks are all there and its suite is green
+  -> uncapped 0.085, failed blockers ['media_file_of_a_deleted_post_is_removed'], reward 0.085
+
+=== alt ===
+  PASS  media_file_of_a_deleted_post_is_removed                w=0.0550 blocker    9.4s
+        PASS: media_file_of_a_deleted_post_is_removed — 2 fresh runs agree
+  PASS  hashtag_index_forgets_a_deleted_post                   w=0.0550            0.0s
+        PASS: hashtag_index_forgets_a_deleted_post — 2 fresh runs agree
+  PASS  a_deleted_blog_post_leaves_no_rendition                w=0.0550            0.0s
+        PASS: a_deleted_blog_post_leaves_no_rendition — 2 fresh runs agree
+  PASS  deleted_post_is_delisted_from_the_media_log            w=0.0550            0.0s
+        PASS: deleted_post_is_delisted_from_the_media_log — 2 fresh runs agree
+  PASS  a_deleted_blog_post_leaves_no_json_or_attachment       w=0.0550            0.0s
+        PASS: a_deleted_blog_post_leaves_no_json_or_attachment — 2 fresh runs agree
+  PASS  timeline_indexes_forget_a_deleted_post                 w=0.0550            0.0s
+        PASS: timeline_indexes_forget_a_deleted_post — 2 fresh runs agree
+  PASS  features_copy_of_a_deleted_article_is_removed          w=0.1600            0.0s
+        PASS: features_copy_of_a_deleted_article_is_removed — 2 fresh runs agree
+  PASS  features_index_forgets_a_deleted_article               w=0.1600            0.0s
+        PASS: features_index_forgets_a_deleted_article — 2 fresh runs agree
+  PASS  event_category_index_forgets_a_deleted_event           w=0.1600            0.0s
+        PASS: event_category_index_forgets_a_deleted_event — 2 fresh runs agree
+  PASS  split_event_category_indexes_forget_a_deleted_event    w=0.1600            0.0s
+        PASS: split_event_category_indexes_forget_a_deleted_event — 2 fresh runs agree
+  ... calibration still running (12:47:09) ...
+  ... calibration still running (12:48:09) ...
+  PASS  gate_repo_suite_green_and_tests_untouched              w=0.0300 blocker   99.1s CORE_GREEN=1
+        PASS: the repository's own checks are all there and its suite is green
+  -> uncapped 1.000, failed blockers none, reward 1.000
+
+=== cheat ===
+  PASS  media_file_of_a_deleted_post_is_removed                w=0.0550 blocker    9.3s
+        PASS: media_file_of_a_deleted_post_is_removed — 2 fresh runs agree
+  PASS  hashtag_index_forgets_a_deleted_post                   w=0.0550            0.0s
+        PASS: hashtag_index_forgets_a_deleted_post — 2 fresh runs agree
+  PASS  a_deleted_blog_post_leaves_no_rendition                w=0.0550            0.0s
+        PASS: a_deleted_blog_post_leaves_no_rendition — 2 fresh runs agree
+  PASS  deleted_post_is_delisted_from_the_media_log            w=0.0550            0.0s
+        PASS: deleted_post_is_delisted_from_the_media_log — 2 fresh runs agree
+  PASS  a_deleted_blog_post_leaves_no_json_or_attachment       w=0.0550            0.0s
+        PASS: a_deleted_blog_post_leaves_no_json_or_attachment — 2 fresh runs agree
+  PASS  timeline_indexes_forget_a_deleted_post                 w=0.0550            0.0s
+        PASS: timeline_indexes_forget_a_deleted_post — 2 fresh runs agree
+  PASS  features_copy_of_a_deleted_article_is_removed          w=0.1600            0.0s
+        PASS: features_copy_of_a_deleted_article_is_removed — 2 fresh runs agree
+  PASS  features_index_forgets_a_deleted_article               w=0.1600            0.0s
+        PASS: features_index_forgets_a_deleted_article — 2 fresh runs agree
+  PASS  event_category_index_forgets_a_deleted_event           w=0.1600            0.0s
+        PASS: event_category_index_forgets_a_deleted_event — 2 fresh runs agree
+  PASS  split_event_category_indexes_forget_a_deleted_event    w=0.1600            0.0s
+        PASS: split_event_category_indexes_forget_a_deleted_event — 2 fresh runs agree
+  FAIL  gate_repo_suite_green_and_tests_untouched              w=0.0300 blocker    2.4s CORE_GREEN=1
+        FAIL: the repository's own checks were changed or removed
+  -> uncapped 0.970, failed blockers ['gate_repo_suite_green_and_tests_untouched'], reward 0.600
+
+=== calibration ===
+REWARD gold 1.000
+REWARD base 0.000
+REWARD noop 0.000
+REWARD lazy 0.085
+REWARD alt 1.000
+REWARD cheat 0.600
+  gold: **1.00** (expected 1.00) OK
+  base: **0.00** (expected 0.00) OK
+  gold: all criteria green, one by one
+  base: all cells red one by one; the gate's own suite is green on the untouched checkout (G-4) and the gate refuses to be paid for it (G-5')
+  noop: 0.000 OK
+  lazy: 0.085 (< 0.50) OK
+  alt: 1.000 (>= 0.80) OK
+  cheat: 0.600 (uncapped 0.970, capped by gate_repo_suite_green_and_tests_untouched) OK
+LADDER gold=1.000 base=0.000 noop=0.000 lazy=0.085 alt=1.000 cheat=0.600
+
+CALIBRATION OK; written to /hud/.calib/calibration.json
+  gold: **1.00** (expected 1.00) OK
+  base: **0.00** (expected 0.00) OK
+  gold: all criteria green, one by one
+  base: all cells red one by one; the gate's own suite is green on the untouched checkout (G-4) and the gate refuses to be paid for it (G-5')
+  noop: 0.000 OK
+  lazy: 0.085 (< 0.50) OK
+  alt: 1.000 (>= 0.80) OK
+  cheat: 0.600 (uncapped 0.970, capped by gate_repo_suite_green_and_tests_untouched) OK
+LADDER gold=1.000 base=0.000 noop=0.000 lazy=0.085 alt=1.000 cheat=0.600

harbor/evidence/epicyon/native-reference.json

@@ -0,0 +1,185 @@
+{
+  "reward": 1.0,
+  "done": true,
+  "content": null,
+  "info": {},
+  "isError": false,
+  "subscores": [
+    {
+      "name": "media_file_of_a_deleted_post_is_removed",
+      "weight": 0.055,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "PASS: media_file_of_a_deleted_post_is_removed \u2014 2 fresh runs agree\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "CORE_GREEN=1 bash /hud/grader/run_grading.sh media_file_of_a_deleted_post_is_removed 600",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "hashtag_index_forgets_a_deleted_post",
+      "weight": 0.055,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "PASS: hashtag_index_forgets_a_deleted_post \u2014 2 fresh runs agree\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "CORE_GREEN=1 bash /hud/grader/run_grading.sh hashtag_index_forgets_a_deleted_post 600",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "a_deleted_blog_post_leaves_no_rendition",
+      "weight": 0.055,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "PASS: a_deleted_blog_post_leaves_no_rendition \u2014 2 fresh runs agree\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "CORE_GREEN=1 bash /hud/grader/run_grading.sh a_deleted_blog_post_leaves_no_rendition 600",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "deleted_post_is_delisted_from_the_media_log",
+      "weight": 0.055,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "PASS: deleted_post_is_delisted_from_the_media_log \u2014 2 fresh runs agree\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "CORE_GREEN=1 bash /hud/grader/run_grading.sh deleted_post_is_delisted_from_the_media_log 600",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "a_deleted_blog_post_leaves_no_json_or_attachment",
+      "weight": 0.055,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "PASS: a_deleted_blog_post_leaves_no_json_or_attachment \u2014 2 fresh runs agree\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "CORE_GREEN=1 bash /hud/grader/run_grading.sh a_deleted_blog_post_leaves_no_json_or_attachment 600",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "timeline_indexes_forget_a_deleted_post",
+      "weight": 0.055,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "PASS: timeline_indexes_forget_a_deleted_post \u2014 2 fresh runs agree\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "CORE_GREEN=1 bash /hud/grader/run_grading.sh timeline_indexes_forget_a_deleted_post 600",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "features_copy_of_a_deleted_article_is_removed",
+      "weight": 0.16,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "PASS: features_copy_of_a_deleted_article_is_removed \u2014 2 fresh runs agree\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "CORE_GREEN=1 bash /hud/grader/run_grading.sh features_copy_of_a_deleted_article_is_removed 600",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "features_index_forgets_a_deleted_article",
+      "weight": 0.16,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "PASS: features_index_forgets_a_deleted_article \u2014 2 fresh runs agree\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "CORE_GREEN=1 bash /hud/grader/run_grading.sh features_index_forgets_a_deleted_article 600",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "event_category_index_forgets_a_deleted_event",
+      "weight": 0.16,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "PASS: event_category_index_forgets_a_deleted_event \u2014 2 fresh runs agree\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "CORE_GREEN=1 bash /hud/grader/run_grading.sh event_category_index_forgets_a_deleted_event 600",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "split_event_category_indexes_forget_a_deleted_event",
+      "weight": 0.16,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "PASS: split_event_category_indexes_forget_a_deleted_event \u2014 2 fresh runs agree\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "CORE_GREEN=1 bash /hud/grader/run_grading.sh split_event_category_indexes_forget_a_deleted_event 600",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "gate_repo_suite_green_and_tests_untouched",
+      "weight": 0.03,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "PASS: the repository's own checks are all there and its suite is green\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "CORE_GREEN=1 bash /hud/grader/run_grading.sh gate_repo_suite_green_and_tests_untouched 900",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 900
+        }
+      }
+    }
+  ]
+}

harbor/evidence/epicyon/reports/noop.json

@@ -0,0 +1,377 @@
+{
+  "revision": "coding-rubric-v2",
+  "task_id": "epicyon_deleted_post_announce_cache_contract",
+  "grader_version": "0.6.15",
+  "commands": [
+    {
+      "name": "media_file_of_a_deleted_post_is_removed",
+      "weight": 0.055,
+      "value": 0.0,
+      "children": null,
+      "info": {
+        "exit_code": 1,
+        "stdout": "FAIL: the deleted post's media is still on disk: /media/2962/2E0Haokz35sNz4rtdcPOAD8fim5pp00C.png\n      CELL=media_file_of_a_deleted_post_is_removed FAIL the deleted post's media is still on disk: /media/2962/2E0Haokz35sNz4rtdcPOAD8fim5pp00C.png\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=1 bash /hud/grader/run_grading.sh media_file_of_a_deleted_post_is_removed 600'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "hashtag_index_forgets_a_deleted_post",
+      "weight": 0.055,
+      "value": 0.0,
+      "children": null,
+      "info": {
+        "exit_code": 1,
+        "stdout": "FAIL: the hashtag index still points at the deleted post\n      CELL=hashtag_index_forgets_a_deleted_post FAIL the hashtag index still points at the deleted post\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=1 bash /hud/grader/run_grading.sh hashtag_index_forgets_a_deleted_post 600'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "a_deleted_blog_post_leaves_no_rendition",
+      "weight": 0.055,
+      "value": 0.0,
+      "children": null,
+      "info": {
+        "exit_code": 1,
+        "stdout": "FAIL: the deleted blog post is still published as markdown, micron\n      CELL=a_deleted_blog_post_leaves_no_rendition FAIL the deleted blog post is still published as markdown, micron\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=1 bash /hud/grader/run_grading.sh a_deleted_blog_post_leaves_no_rendition 600'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "deleted_post_is_delisted_from_the_media_log",
+      "weight": 0.055,
+      "value": 0.0,
+      "children": null,
+      "info": {
+        "exit_code": 1,
+        "stdout": "FAIL: the media log still lists the deleted post's media: /media/2962/MupdsseYdqxwX1nmyj2fRxETTraXMtfU.png\n      CELL=deleted_post_is_delisted_from_the_media_log FAIL the media log still lists the deleted post's media: /media/2962/MupdsseYdqxwX1nmyj2fRxETTraXMtfU.png\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=1 bash /hud/grader/run_grading.sh deleted_post_is_delisted_from_the_media_log 600'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "a_deleted_blog_post_leaves_no_json_or_attachment",
+      "weight": 0.055,
+      "value": 0.0,
+      "children": null,
+      "info": {
+        "exit_code": 1,
+        "stdout": "FAIL: the deleted blog post's own json is still there\n      CELL=a_deleted_blog_post_leaves_no_json_or_attachment FAIL the deleted blog post's own json is still there\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=1 bash /hud/grader/run_grading.sh a_deleted_blog_post_leaves_no_json_or_attachment 600'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "timeline_indexes_forget_a_deleted_post",
+      "weight": 0.055,
+      "value": 0.0,
+      "children": null,
+      "info": {
+        "exit_code": 1,
+        "stdout": "FAIL: the outbox index still lists the deleted post\n      CELL=timeline_indexes_forget_a_deleted_post FAIL the outbox index still lists the deleted post\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=1 bash /hud/grader/run_grading.sh timeline_indexes_forget_a_deleted_post 600'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "features_copy_of_a_deleted_article_is_removed",
+      "weight": 0.16,
+      "value": 0.0,
+      "children": null,
+      "info": {
+        "exit_code": 1,
+        "stdout": "FAIL: the news actor still holds a copy of the deleted article\n      CELL=features_copy_of_a_deleted_article_is_removed FAIL the news actor still holds a copy of the deleted article\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=1 bash /hud/grader/run_grading.sh features_copy_of_a_deleted_article_is_removed 600'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "features_index_forgets_a_deleted_article",
+      "weight": 0.16,
+      "value": 0.0,
+      "children": null,
+      "info": {
+        "exit_code": 1,
+        "stdout": "FAIL: the news actor's index still lists the deleted article\n      CELL=features_index_forgets_a_deleted_article FAIL the news actor's index still lists the deleted article\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=1 bash /hud/grader/run_grading.sh features_index_forgets_a_deleted_article 600'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "event_category_index_forgets_a_deleted_event",
+      "weight": 0.16,
+      "value": 0.0,
+      "children": null,
+      "info": {
+        "exit_code": 1,
+        "stdout": "FAIL: the Music tag index still lists the deleted event\n      CELL=event_category_index_forgets_a_deleted_event FAIL the Music tag index still lists the deleted event\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=1 bash /hud/grader/run_grading.sh event_category_index_forgets_a_deleted_event 600'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "split_event_category_indexes_forget_a_deleted_event",
+      "weight": 0.16,
+      "value": 0.0,
+      "children": null,
+      "info": {
+        "exit_code": 1,
+        "stdout": "FAIL: the Outdoors tag index still lists the deleted event\n      CELL=split_event_category_indexes_forget_a_deleted_event FAIL the Outdoors tag index still lists the deleted event\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=1 bash /hud/grader/run_grading.sh split_event_category_indexes_forget_a_deleted_event 600'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "gate_repo_suite_green_and_tests_untouched",
+      "weight": 0.03,
+      "value": 0.0,
+      "children": null,
+      "info": {
+        "exit_code": 1,
+        "stdout": "the repository's own tests: green \u2014 the repository's own checks are all there and its suite is green\nFAIL: no scored behaviour was fixed \u2014 this gate does not pay for an empty change\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=0 bash /hud/grader/run_grading.sh gate_repo_suite_green_and_tests_untouched 900'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 900
+        }
+      }
+    }
+  ],
+  "status": "graded",
+  "patch_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
+  "result": {
+    "reward": 0.0,
+    "done": true,
+    "content": "Blocker criteria failed (gate_repo_suite_green_and_tests_untouched, media_file_of_a_deleted_post_is_removed); cap 0.60, reward 0.000 (uncapped 0.000).",
+    "info": {
+      "uncapped_reward": 0.0,
+      "failed_blockers": [
+        "gate_repo_suite_green_and_tests_untouched",
+        "media_file_of_a_deleted_post_is_removed"
+      ],
+      "blocker_cap": 0.6
+    },
+    "isError": false,
+    "subscores": [
+      {
+        "name": "media_file_of_a_deleted_post_is_removed",
+        "weight": 0.055,
+        "value": 0.0,
+        "children": null,
+        "info": {
+          "exit_code": 1,
+          "stdout": "FAIL: the deleted post's media is still on disk: /media/2962/2E0Haokz35sNz4rtdcPOAD8fim5pp00C.png\n      CELL=media_file_of_a_deleted_post_is_removed FAIL the deleted post's media is still on disk: /media/2962/2E0Haokz35sNz4rtdcPOAD8fim5pp00C.png\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=1 bash /hud/grader/run_grading.sh media_file_of_a_deleted_post_is_removed 600'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 600
+          }
+        }
+      },
+      {
+        "name": "hashtag_index_forgets_a_deleted_post",
+        "weight": 0.055,
+        "value": 0.0,
+        "children": null,
+        "info": {
+          "exit_code": 1,
+          "stdout": "FAIL: the hashtag index still points at the deleted post\n      CELL=hashtag_index_forgets_a_deleted_post FAIL the hashtag index still points at the deleted post\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=1 bash /hud/grader/run_grading.sh hashtag_index_forgets_a_deleted_post 600'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 600
+          }
+        }
+      },
+      {
+        "name": "a_deleted_blog_post_leaves_no_rendition",
+        "weight": 0.055,
+        "value": 0.0,
+        "children": null,
+        "info": {
+          "exit_code": 1,
+          "stdout": "FAIL: the deleted blog post is still published as markdown, micron\n      CELL=a_deleted_blog_post_leaves_no_rendition FAIL the deleted blog post is still published as markdown, micron\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=1 bash /hud/grader/run_grading.sh a_deleted_blog_post_leaves_no_rendition 600'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 600
+          }
+        }
+      },
+      {
+        "name": "deleted_post_is_delisted_from_the_media_log",
+        "weight": 0.055,
+        "value": 0.0,
+        "children": null,
+        "info": {
+          "exit_code": 1,
+          "stdout": "FAIL: the media log still lists the deleted post's media: /media/2962/MupdsseYdqxwX1nmyj2fRxETTraXMtfU.png\n      CELL=deleted_post_is_delisted_from_the_media_log FAIL the media log still lists the deleted post's media: /media/2962/MupdsseYdqxwX1nmyj2fRxETTraXMtfU.png\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=1 bash /hud/grader/run_grading.sh deleted_post_is_delisted_from_the_media_log 600'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 600
+          }
+        }
+      },
+      {
+        "name": "a_deleted_blog_post_leaves_no_json_or_attachment",
+        "weight": 0.055,
+        "value": 0.0,
+        "children": null,
+        "info": {
+          "exit_code": 1,
+          "stdout": "FAIL: the deleted blog post's own json is still there\n      CELL=a_deleted_blog_post_leaves_no_json_or_attachment FAIL the deleted blog post's own json is still there\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=1 bash /hud/grader/run_grading.sh a_deleted_blog_post_leaves_no_json_or_attachment 600'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 600
+          }
+        }
+      },
+      {
+        "name": "timeline_indexes_forget_a_deleted_post",
+        "weight": 0.055,
+        "value": 0.0,
+        "children": null,
+        "info": {
+          "exit_code": 1,
+          "stdout": "FAIL: the outbox index still lists the deleted post\n      CELL=timeline_indexes_forget_a_deleted_post FAIL the outbox index still lists the deleted post\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=1 bash /hud/grader/run_grading.sh timeline_indexes_forget_a_deleted_post 600'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 600
+          }
+        }
+      },
+      {
+        "name": "features_copy_of_a_deleted_article_is_removed",
+        "weight": 0.16,
+        "value": 0.0,
+        "children": null,
+        "info": {
+          "exit_code": 1,
+          "stdout": "FAIL: the news actor still holds a copy of the deleted article\n      CELL=features_copy_of_a_deleted_article_is_removed FAIL the news actor still holds a copy of the deleted article\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=1 bash /hud/grader/run_grading.sh features_copy_of_a_deleted_article_is_removed 600'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 600
+          }
+        }
+      },
+      {
+        "name": "features_index_forgets_a_deleted_article",
+        "weight": 0.16,
+        "value": 0.0,
+        "children": null,
+        "info": {
+          "exit_code": 1,
+          "stdout": "FAIL: the news actor's index still lists the deleted article\n      CELL=features_index_forgets_a_deleted_article FAIL the news actor's index still lists the deleted article\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=1 bash /hud/grader/run_grading.sh features_index_forgets_a_deleted_article 600'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 600
+          }
+        }
+      },
+      {
+        "name": "event_category_index_forgets_a_deleted_event",
+        "weight": 0.16,
+        "value": 0.0,
+        "children": null,
+        "info": {
+          "exit_code": 1,
+          "stdout": "FAIL: the Music tag index still lists the deleted event\n      CELL=event_category_index_forgets_a_deleted_event FAIL the Music tag index still lists the deleted event\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=1 bash /hud/grader/run_grading.sh event_category_index_forgets_a_deleted_event 600'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 600
+          }
+        }
+      },
+      {
+        "name": "split_event_category_indexes_forget_a_deleted_event",
+        "weight": 0.16,
+        "value": 0.0,
+        "children": null,
+        "info": {
+          "exit_code": 1,
+          "stdout": "FAIL: the Outdoors tag index still lists the deleted event\n      CELL=split_event_category_indexes_forget_a_deleted_event FAIL the Outdoors tag index still lists the deleted event\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=1 bash /hud/grader/run_grading.sh split_event_category_indexes_forget_a_deleted_event 600'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 600
+          }
+        }
+      },
+      {
+        "name": "gate_repo_suite_green_and_tests_untouched",
+        "weight": 0.03,
+        "value": 0.0,
+        "children": null,
+        "info": {
+          "exit_code": 1,
+          "stdout": "the repository's own tests: green \u2014 the repository's own checks are all there and its suite is green\nFAIL: no scored behaviour was fixed \u2014 this gate does not pay for an empty change\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=0 bash /hud/grader/run_grading.sh gate_repo_suite_green_and_tests_untouched 900'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 900
+          }
+        }
+      }
+    ]
+  }
+}

harbor/evidence/epicyon/reports/reference.json

@@ -0,0 +1,370 @@
+{
+  "revision": "coding-rubric-v2",
+  "task_id": "epicyon_deleted_post_announce_cache_contract",
+  "grader_version": "0.6.15",
+  "commands": [
+    {
+      "name": "media_file_of_a_deleted_post_is_removed",
+      "weight": 0.055,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "PASS: media_file_of_a_deleted_post_is_removed \u2014 2 fresh runs agree\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=1 bash /hud/grader/run_grading.sh media_file_of_a_deleted_post_is_removed 600'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "hashtag_index_forgets_a_deleted_post",
+      "weight": 0.055,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "PASS: hashtag_index_forgets_a_deleted_post \u2014 2 fresh runs agree\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=1 bash /hud/grader/run_grading.sh hashtag_index_forgets_a_deleted_post 600'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "a_deleted_blog_post_leaves_no_rendition",
+      "weight": 0.055,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "PASS: a_deleted_blog_post_leaves_no_rendition \u2014 2 fresh runs agree\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=1 bash /hud/grader/run_grading.sh a_deleted_blog_post_leaves_no_rendition 600'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "deleted_post_is_delisted_from_the_media_log",
+      "weight": 0.055,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "PASS: deleted_post_is_delisted_from_the_media_log \u2014 2 fresh runs agree\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=1 bash /hud/grader/run_grading.sh deleted_post_is_delisted_from_the_media_log 600'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "a_deleted_blog_post_leaves_no_json_or_attachment",
+      "weight": 0.055,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "PASS: a_deleted_blog_post_leaves_no_json_or_attachment \u2014 2 fresh runs agree\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=1 bash /hud/grader/run_grading.sh a_deleted_blog_post_leaves_no_json_or_attachment 600'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "timeline_indexes_forget_a_deleted_post",
+      "weight": 0.055,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "PASS: timeline_indexes_forget_a_deleted_post \u2014 2 fresh runs agree\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=1 bash /hud/grader/run_grading.sh timeline_indexes_forget_a_deleted_post 600'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "features_copy_of_a_deleted_article_is_removed",
+      "weight": 0.16,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "PASS: features_copy_of_a_deleted_article_is_removed \u2014 2 fresh runs agree\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=1 bash /hud/grader/run_grading.sh features_copy_of_a_deleted_article_is_removed 600'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "features_index_forgets_a_deleted_article",
+      "weight": 0.16,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "PASS: features_index_forgets_a_deleted_article \u2014 2 fresh runs agree\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=1 bash /hud/grader/run_grading.sh features_index_forgets_a_deleted_article 600'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "event_category_index_forgets_a_deleted_event",
+      "weight": 0.16,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "PASS: event_category_index_forgets_a_deleted_event \u2014 2 fresh runs agree\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=1 bash /hud/grader/run_grading.sh event_category_index_forgets_a_deleted_event 600'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "split_event_category_indexes_forget_a_deleted_event",
+      "weight": 0.16,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "PASS: split_event_category_indexes_forget_a_deleted_event \u2014 2 fresh runs agree\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=1 bash /hud/grader/run_grading.sh split_event_category_indexes_forget_a_deleted_event 600'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "gate_repo_suite_green_and_tests_untouched",
+      "weight": 0.03,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "PASS: the repository's own checks are all there and its suite is green\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=1 bash /hud/grader/run_grading.sh gate_repo_suite_green_and_tests_untouched 900'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 900
+        }
+      }
+    }
+  ],
+  "status": "graded",
+  "patch_sha256": "b5866d660e5151ed07222c809e6e92e4ff78f28d7a154d244b65d4cba0f847bf",
+  "result": {
+    "reward": 1.0,
+    "done": true,
+    "content": null,
+    "info": {},
+    "isError": false,
+    "subscores": [
+      {
+        "name": "media_file_of_a_deleted_post_is_removed",
+        "weight": 0.055,
+        "value": 1.0,
+        "children": null,
+        "info": {
+          "exit_code": 0,
+          "stdout": "PASS: media_file_of_a_deleted_post_is_removed \u2014 2 fresh runs agree\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=1 bash /hud/grader/run_grading.sh media_file_of_a_deleted_post_is_removed 600'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 600
+          }
+        }
+      },
+      {
+        "name": "hashtag_index_forgets_a_deleted_post",
+        "weight": 0.055,
+        "value": 1.0,
+        "children": null,
+        "info": {
+          "exit_code": 0,
+          "stdout": "PASS: hashtag_index_forgets_a_deleted_post \u2014 2 fresh runs agree\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=1 bash /hud/grader/run_grading.sh hashtag_index_forgets_a_deleted_post 600'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 600
+          }
+        }
+      },
+      {
+        "name": "a_deleted_blog_post_leaves_no_rendition",
+        "weight": 0.055,
+        "value": 1.0,
+        "children": null,
+        "info": {
+          "exit_code": 0,
+          "stdout": "PASS: a_deleted_blog_post_leaves_no_rendition \u2014 2 fresh runs agree\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=1 bash /hud/grader/run_grading.sh a_deleted_blog_post_leaves_no_rendition 600'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 600
+          }
+        }
+      },
+      {
+        "name": "deleted_post_is_delisted_from_the_media_log",
+        "weight": 0.055,
+        "value": 1.0,
+        "children": null,
+        "info": {
+          "exit_code": 0,
+          "stdout": "PASS: deleted_post_is_delisted_from_the_media_log \u2014 2 fresh runs agree\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=1 bash /hud/grader/run_grading.sh deleted_post_is_delisted_from_the_media_log 600'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 600
+          }
+        }
+      },
+      {
+        "name": "a_deleted_blog_post_leaves_no_json_or_attachment",
+        "weight": 0.055,
+        "value": 1.0,
+        "children": null,
+        "info": {
+          "exit_code": 0,
+          "stdout": "PASS: a_deleted_blog_post_leaves_no_json_or_attachment \u2014 2 fresh runs agree\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=1 bash /hud/grader/run_grading.sh a_deleted_blog_post_leaves_no_json_or_attachment 600'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 600
+          }
+        }
+      },
+      {
+        "name": "timeline_indexes_forget_a_deleted_post",
+        "weight": 0.055,
+        "value": 1.0,
+        "children": null,
+        "info": {
+          "exit_code": 0,
+          "stdout": "PASS: timeline_indexes_forget_a_deleted_post \u2014 2 fresh runs agree\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=1 bash /hud/grader/run_grading.sh timeline_indexes_forget_a_deleted_post 600'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 600
+          }
+        }
+      },
+      {
+        "name": "features_copy_of_a_deleted_article_is_removed",
+        "weight": 0.16,
+        "value": 1.0,
+        "children": null,
+        "info": {
+          "exit_code": 0,
+          "stdout": "PASS: features_copy_of_a_deleted_article_is_removed \u2014 2 fresh runs agree\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=1 bash /hud/grader/run_grading.sh features_copy_of_a_deleted_article_is_removed 600'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 600
+          }
+        }
+      },
+      {
+        "name": "features_index_forgets_a_deleted_article",
+        "weight": 0.16,
+        "value": 1.0,
+        "children": null,
+        "info": {
+          "exit_code": 0,
+          "stdout": "PASS: features_index_forgets_a_deleted_article \u2014 2 fresh runs agree\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=1 bash /hud/grader/run_grading.sh features_index_forgets_a_deleted_article 600'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 600
+          }
+        }
+      },
+      {
+        "name": "event_category_index_forgets_a_deleted_event",
+        "weight": 0.16,
+        "value": 1.0,
+        "children": null,
+        "info": {
+          "exit_code": 0,
+          "stdout": "PASS: event_category_index_forgets_a_deleted_event \u2014 2 fresh runs agree\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=1 bash /hud/grader/run_grading.sh event_category_index_forgets_a_deleted_event 600'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 600
+          }
+        }
+      },
+      {
+        "name": "split_event_category_indexes_forget_a_deleted_event",
+        "weight": 0.16,
+        "value": 1.0,
+        "children": null,
+        "info": {
+          "exit_code": 0,
+          "stdout": "PASS: split_event_category_indexes_forget_a_deleted_event \u2014 2 fresh runs agree\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=1 bash /hud/grader/run_grading.sh split_event_category_indexes_forget_a_deleted_event 600'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 600
+          }
+        }
+      },
+      {
+        "name": "gate_repo_suite_green_and_tests_untouched",
+        "weight": 0.03,
+        "value": 1.0,
+        "children": null,
+        "info": {
+          "exit_code": 0,
+          "stdout": "PASS: the repository's own checks are all there and its suite is green\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=1 bash /hud/grader/run_grading.sh gate_repo_suite_green_and_tests_untouched 900'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 900
+          }
+        }
+      }
+    ]
+  }
+}

harbor/evidence/epicyon/summary.json

@@ -0,0 +1,157 @@
+{
+  "name": "epicyon",
+  "measure": {
+    "ok": true,
+    "hud": "0.6.15",
+    "path": "/opt/agentvenv/bin:/usr/local/bin:/usr/bin:/bin:/usr/local/games:/usr/games",
+    "handed": [
+      "PYTHONDONTWRITEBYTECODE",
+      "PYTHONNOUSERSITE",
+      "PYTHONPATH",
+      "PYTHONUNBUFFERED"
+    ],
+    "repo_dir": "/workspace/repo",
+    "python_usr_local": true,
+    "seconds": 4
+  },
+  "parity": {
+    "ok": true,
+    "reward": 1.0,
+    "criteria": 11,
+    "unmet": [],
+    "info": null,
+    "seconds": 157
+  },
+  "task_id": "epicyon_deleted_post_announce_cache_contract",
+  "image": "ec-swe-epicyon:source-v7",
+  "build": {
+    "ok": true,
+    "image_id": "sha256:74752fd4e8ebf481d6f0c01fd9b7e542c539f850bf9eed8503f41aaeabd7d99d",
+    "ladder": {
+      "gold": 1.0,
+      "base": 0.0,
+      "noop": 0.0,
+      "lazy": 0.085,
+      "alt": 1.0,
+      "cheat": 0.6
+    },
+    "source_digest": "d8f491bc55257bb8",
+    "source": "edited",
+    "seconds": 0
+  },
+  "prepare": {
+    "ok": true,
+    "bundle": "epicyon-r7",
+    "converter": "b8bba68767da888f",
+    "pinned_files": 9,
+    "criteria": 11,
+    "seconds": 1
+  },
+  "qualified": false,
+  "controls": {
+    "ok": false,
+    "checks": {
+      "noop_is_zero": true,
+      "reference_is_one": true,
+      "reference_equals_native": true,
+      "agent_is_isolated": true
+    },
+    "results": {
+      "noop": {
+        "run": "epicyon-r7-control-noop-1791898762",
+        "reward": 0.0,
+        "stop": "completed",
+        "error": null,
+        "seconds": 351,
+        "probe": false,
+        "status": "graded",
+        "verifier_error": null,
+        "exit_codes": [
+          1
+        ],
+        "attempts": 1
+      },
+      "reference": {
+        "run": "epicyon-r7-control-reference-1791899113",
+        "reward": 1.0,
+        "stop": "completed",
+        "error": null,
+        "seconds": 413,
+        "probe": true,
+        "status": "graded",
+        "verifier_error": null,
+        "exit_codes": [
+          0
+        ],
+        "subscores": {
+          "media_file_of_a_deleted_post_is_removed": [
+            1.0,
+            0.055
+          ],
+          "hashtag_index_forgets_a_deleted_post": [
+            1.0,
+            0.055
+          ],
+          "a_deleted_blog_post_leaves_no_rendition": [
+            1.0,
+            0.055
+          ],
+          "deleted_post_is_delisted_from_the_media_log": [
+            1.0,
+            0.055
+          ],
+          "a_deleted_blog_post_leaves_no_json_or_attachment": [
+            1.0,
+            0.055
+          ],
+          "timeline_indexes_forget_a_deleted_post": [
+            1.0,
+            0.055
+          ],
+          "features_copy_of_a_deleted_article_is_removed": [
+            1.0,
+            0.16
+          ],
+          "features_index_forgets_a_deleted_article": [
+            1.0,
+            0.16
+          ],
+          "event_category_index_forgets_a_deleted_event": [
+            1.0,
+            0.16
+          ],
+          "split_event_category_indexes_forget_a_deleted_event": [
+            1.0,
+            0.16
+          ],
+          "gate_repo_suite_green_and_tests_untouched": [
+            1.0,
+            0.03
+          ]
+        },
+        "attempts": 1
+      }
+    },
+    "bundle": "epicyon-r7",
+    "complete": false,
+    "seconds": 765
+  },
+  "leak": {
+    "scan_finished": true,
+    "hits_outside_checkout": [],
+    "other_files_holding_only_upstream_lines": 0,
+    "library_files_sharing_a_line": 0,
+    "most_lines_shared_by_one_library_file": 0,
+    "private_paths_absent": true,
+    "public_lines_shared_with_patch_and_graders": [
+      6,
+      0
+    ],
+    "private_lines_only_in_agent_checkout": [
+      0,
+      0
+    ],
+    "ok": true,
+    "seconds": 26
+  }
+}

harbor/evidence/packages-reproduced-by-final-converter.json

@@ -0,0 +1,20 @@
+{
+ "converter": "b8bba68767da888f",
+ "packages": {
+  "badger": "differs: tests/docker-compose.yaml",
+  "bazel-remote": "differs: tests/docker-compose.yaml",
+  "celery": "differs: tests/docker-compose.yaml",
+  "coredns": "differs: tests/docker-compose.yaml",
+  "epicyon": "differs: tests/docker-compose.yaml",
+  "geode": "differs: tests/docker-compose.yaml",
+  "kratos": "differs: task.toml, environment/Dockerfile, environment/capture.py, environment/capture_submission.py, environment/docker-compose.yaml, environment/sanitize.py",
+  "microblogpub": "differs: tests/docker-compose.yaml",
+  "open-xml-sdk": "differs: task.toml, environment/Dockerfile, environment/capture.py, environment/capture_submission.py, environment/docker-compose.yaml, environment/sanitize.py",
+  "openfire": "differs: task.toml, environment/Dockerfile, environment/capture.py, environment/capture_submission.py, environment/docker-compose.yaml, environment/sanitize.py",
+  "rope": "differs: tests/docker-compose.yaml",
+  "terser": "differs: tests/docker-compose.yaml",
+  "umya": "differs: task.toml, environment/Dockerfile, environment/capture.py, environment/capture_submission.py, environment/docker-compose.yaml, environment/sanitize.py",
+  "zettlr": "differs: tests/docker-compose.yaml",
+  "zod": "differs: tests/docker-compose.yaml"
+ }
+}

harbor/evidence/rope/agent-image-scan.txt

@@ -0,0 +1,15 @@
+needles: 855
+1 matching line(s): /workspace/repo/ropetest/pycoretest.py
+4 matching line(s): /workspace/repo/ropetest/refactor/renametest.py
+scan finished
+absent: /hud
+absent: /app
+absent: /tests
+present: /opt/task-tools/baseline.git
+reference-patch lines in the agent's tracked files: 3, of which the pristine checkout lacks: 0
+grader lines in the agent's tracked files: 0, of which the pristine checkout lacks: 0
+reference-patch lines in total: 363, grader lines in total: 485
+private lines in 0 other files outside the checkout: 0, of which upstream lacks: 0
+--- private lines only the agent's checkout holds (expect none)
+--- the agent's checkout: uncommitted changes (expect none), then its history
+9c94847 baseline

harbor/evidence/rope/author-calibration.txt

@@ -0,0 +1,844 @@
+=== gold ===
+  PASS  export_list_follows_the_rename                 w=0.08            3.0s
+        PASS: the defining module's export list
+        PASS: an export list written as a tuple over several lines
+        PASS: the export list of a package that re-exports the name
+        PASS: an export list naming a module that is being renamed
+        PASS: an export list extended in place
+        PASS: prose and an unrelated list of the same word
+        PASS: another module that exports a name of its own
+        PASS: an export list built through a variable
+        PASS: an export list added to another list
+        PASS: an export list written as one string of names
+        PASS: an export list written as a block of names
+        PASS: export_list_follows_the_rename (11 scenarios)
+  PASS  type_constructor_names_follow_the_rename       w=0.08            1.7s
+        PASS: the name a new type is given
+        PASS: a new type built from a directly imported constructor
+        PASS: the bound of a type variable
+        PASS: a type whose name is spelled like something else
+        PASS: a constraint of a type variable
+        PASS: a new type whose name is handed over by keyword
+        PASS: type_constructor_names_follow_the_rename (6 scenarios)
+  PASS  declared_attribute_names_follow_the_rename     w=0.08            0.9s
+        PASS: a slots tuple follows the attribute
+        PASS: a slots list follows the attribute
+        PASS: the positional attribute names a class declares for matching
+        PASS: declared_attribute_names_follow_the_rename (3 scenarios)
+  PASS  string_annotations_follow_and_data_stays       w=0.22            1.8s
+        PASS: a forward reference in a signature
+        PASS: a forward reference inside a generic
+        PASS: a literal type beside a forward reference
+        PASS: the metadata of an annotated type
+        PASS: a class body with a literal field and a forward reference
+        PASS: string_annotations_follow_and_data_stays (5 scenarios)
+  PASS  move_keeps_both_export_contracts               w=0.08 blocker    0.8s
+        PASS: the module left behind stops exporting the name
+        PASS: the module arrived in starts exporting the name
+        PASS: a module with no export list is not given one
+        PASS: move_keeps_both_export_contracts (3 scenarios)
+  PASS  move_destination_list_however_built_gains_the_name w=0.22            1.3s
+        PASS: the list the destination publishes is built through a variable
+        PASS: the list the destination publishes is chosen by a condition
+        PASS: the list the destination publishes is a sorted copy
+        PASS: the list the destination publishes is one string of names
+        PASS: the generated import does not carry the export list across
+        PASS: move_destination_list_however_built_gains_the_name (5 scenarios)
+  PASS  move_source_list_however_built_loses_the_name  w=0.21            1.4s
+        PASS: a list written as one string of names
+        PASS: a name put on the list at its front
+        PASS: a name appended to the list
+        PASS: a tuple of two names
+        PASS: a tuple of names held in a variable
+        PASS: move_source_list_however_built_loses_the_name (5 scenarios)
+  PASS  gate_repo_suite_green_and_tests_untouched      w=0.03 blocker   35.4s
+        PASS: every shipped test file is byte for byte what it was
+        PASS: the repository's own tests pass
+  -> uncapped 1.000, failed blockers none, failed gates none, reward 1.000
+
+=== base ===
+  FAIL  export_list_follows_the_rename                 w=0.08            2.5s
+        FAIL: the defining module's export list
+        FAIL: an export list written as a tuple over several lines
+        FAIL: the export list of a package that re-exports the name
+        FAIL: an export list naming a module that is being renamed
+        FAIL: an export list extended in place
+        FAIL: prose and an unrelated list of the same word
+        FAIL: another module that exports a name of its own
+        FAIL: an export list built through a variable
+        FAIL: an export list added to another list
+        FAIL: an export list written as one string of names
+        FAIL: an export list written as a block of names
+        FAIL: export_list_follows_the_rename (11 of 11 scenarios)
+  FAIL  type_constructor_names_follow_the_rename       w=0.08            1.5s
+        FAIL: the name a new type is given
+        FAIL: a new type built from a directly imported constructor
+        FAIL: the bound of a type variable
+        PASS: a type whose name is spelled like something else
+        FAIL: a constraint of a type variable
+        FAIL: a new type whose name is handed over by keyword
+        FAIL: type_constructor_names_follow_the_rename (5 of 6 scenarios)
+  FAIL  declared_attribute_names_follow_the_rename     w=0.08            0.7s
+        FAIL: a slots tuple follows the attribute
+        FAIL: a slots list follows the attribute
+        FAIL: the positional attribute names a class declares for matching
+        FAIL: declared_attribute_names_follow_the_rename (3 of 3 scenarios)
+  FAIL  string_annotations_follow_and_data_stays       w=0.22            1.1s
+        FAIL: a forward reference in a signature
+        FAIL: a forward reference inside a generic
+        FAIL: a literal type beside a forward reference
+        FAIL: the metadata of an annotated type
+        FAIL: a class body with a literal field and a forward reference
+        FAIL: string_annotations_follow_and_data_stays (5 of 5 scenarios)
+  FAIL  move_keeps_both_export_contracts               w=0.08 blocker    0.7s
+        FAIL: the module left behind stops exporting the name
+        FAIL: the module arrived in starts exporting the name
+        PASS: a module with no export list is not given one
+        FAIL: move_keeps_both_export_contracts (2 of 3 scenarios)
+  FAIL  move_destination_list_however_built_gains_the_name w=0.22            1.2s
+        FAIL: the list the destination publishes is built through a variable
+        FAIL: the list the destination publishes is chosen by a condition
+        FAIL: the list the destination publishes is a sorted copy
+        FAIL: the list the destination publishes is one string of names
+        FAIL: the generated import does not carry the export list across
+        FAIL: move_destination_list_however_built_gains_the_name (5 of 5 scenarios)
+  FAIL  move_source_list_however_built_loses_the_name  w=0.21            1.2s
+        FAIL: a list written as one string of names
+        FAIL: a name put on the list at its front
+        FAIL: a name appended to the list
+        FAIL: a tuple of two names
+        FAIL: a tuple of names held in a variable
+        FAIL: move_source_list_however_built_loses_the_name (5 of 5 scenarios)
+  FAIL  gate_repo_suite_green_and_tests_untouched      w=0.03 blocker   19.0s
+        RULE HOLDS: every shipped test file is byte for byte what it was
+        RULE HOLDS: the repository's own tests pass
+        FAIL: not paid -- no scored behaviour was fixed, and this gate does not pay for an empty change
+  -> uncapped 0.000, failed blockers ['move_keeps_both_export_contracts', 'gate_repo_suite_green_and_tests_untouched'], failed gates none, reward 0.000
+
+=== noop ===
+  FAIL  export_list_follows_the_rename                 w=0.08            2.4s
+        FAIL: the defining module's export list
+        FAIL: an export list written as a tuple over several lines
+        FAIL: the export list of a package that re-exports the name
+        FAIL: an export list naming a module that is being renamed
+        FAIL: an export list extended in place
+        FAIL: prose and an unrelated list of the same word
+        FAIL: another module that exports a name of its own
+        FAIL: an export list built through a variable
+        FAIL: an export list added to another list
+        FAIL: an export list written as one string of names
+        FAIL: an export list written as a block of names
+        FAIL: export_list_follows_the_rename (11 of 11 scenarios)
+  FAIL  type_constructor_names_follow_the_rename       w=0.08            1.4s
+        FAIL: the name a new type is given
+        FAIL: a new type built from a directly imported constructor
+        FAIL: the bound of a type variable
+        PASS: a type whose name is spelled like something else
+        FAIL: a constraint of a type variable
+        FAIL: a new type whose name is handed over by keyword
+        FAIL: type_constructor_names_follow_the_rename (5 of 6 scenarios)
+  FAIL  declared_attribute_names_follow_the_rename     w=0.08            0.7s
+        FAIL: a slots tuple follows the attribute
+        FAIL: a slots list follows the attribute
+        FAIL: the positional attribute names a class declares for matching
+        FAIL: declared_attribute_names_follow_the_rename (3 of 3 scenarios)
+  FAIL  string_annotations_follow_and_data_stays       w=0.22            1.1s
+        FAIL: a forward reference in a signature
+        FAIL: a forward reference inside a generic
+        FAIL: a literal type beside a forward reference
+        FAIL: the metadata of an annotated type
+        FAIL: a class body with a literal field and a forward reference
+        FAIL: string_annotations_follow_and_data_stays (5 of 5 scenarios)
+  FAIL  move_keeps_both_export_contracts               w=0.08 blocker    0.7s
+        FAIL: the module left behind stops exporting the name
+        FAIL: the module arrived in starts exporting the name
+        PASS: a module with no export list is not given one
+        FAIL: move_keeps_both_export_contracts (2 of 3 scenarios)
+  FAIL  move_destination_list_however_built_gains_the_name w=0.22            1.1s
+        FAIL: the list the destination publishes is built through a variable
+        FAIL: the list the destination publishes is chosen by a condition
+        FAIL: the list the destination publishes is a sorted copy
+        FAIL: the list the destination publishes is one string of names
+        FAIL: the generated import does not carry the export list across
+        FAIL: move_destination_list_however_built_gains_the_name (5 of 5 scenarios)
+  FAIL  move_source_list_however_built_loses_the_name  w=0.21            1.1s
+        FAIL: a list written as one string of names
+        FAIL: a name put on the list at its front
+        FAIL: a name appended to the list
+        FAIL: a tuple of two names
+        FAIL: a tuple of names held in a variable
+        FAIL: move_source_list_however_built_loses_the_name (5 of 5 scenarios)
+  FAIL  gate_repo_suite_green_and_tests_untouched      w=0.03 blocker   18.4s
+        RULE HOLDS: every shipped test file is byte for byte what it was
+        RULE HOLDS: the repository's own tests pass
+        FAIL: not paid -- no scored behaviour was fixed, and this gate does not pay for an empty change
+  -> uncapped 0.000, failed blockers ['move_keeps_both_export_contracts', 'gate_repo_suite_green_and_tests_untouched'], failed gates none, reward 0.000
+
+=== lazy ===
+  FAIL  export_list_follows_the_rename                 w=0.08            2.4s
+        PASS: the defining module's export list
+        PASS: an export list written as a tuple over several lines
+        PASS: the export list of a package that re-exports the name
+        PASS: an export list naming a module that is being renamed
+        PASS: an export list extended in place
+        PASS: prose and an unrelated list of the same word
+        PASS: another module that exports a name of its own
+        FAIL: an export list built through a variable
+        FAIL: an export list added to another list
+        FAIL: an export list written as one string of names
+        FAIL: an export list written as a block of names
+        FAIL: export_list_follows_the_rename (4 of 11 scenarios)
+  FAIL  type_constructor_names_follow_the_rename       w=0.08            1.5s
+        PASS: the name a new type is given
+        PASS: a new type built from a directly imported constructor
+        PASS: the bound of a type variable
+        FAIL: a type whose name is spelled like something else
+        PASS: a constraint of a type variable
+        PASS: a new type whose name is handed over by keyword
+        FAIL: type_constructor_names_follow_the_rename (1 of 6 scenarios)
+  FAIL  declared_attribute_names_follow_the_rename     w=0.08            0.7s
+        FAIL: a slots tuple follows the attribute
+        FAIL: a slots list follows the attribute
+        FAIL: the positional attribute names a class declares for matching
+        FAIL: declared_attribute_names_follow_the_rename (3 of 3 scenarios)
+  FAIL  string_annotations_follow_and_data_stays       w=0.22            1.1s
+        FAIL: a forward reference in a signature
+        FAIL: a forward reference inside a generic
+        FAIL: a literal type beside a forward reference
+        FAIL: the metadata of an annotated type
+        FAIL: a class body with a literal field and a forward reference
+        FAIL: string_annotations_follow_and_data_stays (5 of 5 scenarios)
+  FAIL  move_keeps_both_export_contracts               w=0.08 blocker    0.7s
+        FAIL: the module left behind stops exporting the name
+        FAIL: the module arrived in starts exporting the name
+        PASS: a module with no export list is not given one
+        FAIL: move_keeps_both_export_contracts (2 of 3 scenarios)
+  FAIL  move_destination_list_however_built_gains_the_name w=0.22            1.1s
+        FAIL: the list the destination publishes is built through a variable
+        FAIL: the list the destination publishes is chosen by a condition
+        FAIL: the list the destination publishes is a sorted copy
+        FAIL: the list the destination publishes is one string of names
+        FAIL: the generated import does not carry the export list across
+        FAIL: move_destination_list_however_built_gains_the_name (5 of 5 scenarios)
+  FAIL  move_source_list_however_built_loses_the_name  w=0.21            1.1s
+        FAIL: a list written as one string of names
+        FAIL: a name put on the list at its front
+        FAIL: a name appended to the list
+        FAIL: a tuple of two names
+        FAIL: a tuple of names held in a variable
+        FAIL: move_source_list_however_built_loses_the_name (5 of 5 scenarios)
+  FAIL  gate_repo_suite_green_and_tests_untouched      w=0.03 blocker   18.6s
+        RULE HOLDS: every shipped test file is byte for byte what it was
+        RULE HOLDS: the repository's own tests pass
+        FAIL: not paid -- no scored behaviour was fixed, and this gate does not pay for an empty change
+  -> uncapped 0.000, failed blockers ['move_keeps_both_export_contracts', 'gate_repo_suite_green_and_tests_untouched'], failed gates none, reward 0.000
+
+=== alt ===
+  PASS  export_list_follows_the_rename                 w=0.08            2.4s
+        PASS: the defining module's export list
+        PASS: an export list written as a tuple over several lines
+        PASS: the export list of a package that re-exports the name
+        PASS: an export list naming a module that is being renamed
+        PASS: an export list extended in place
+        PASS: prose and an unrelated list of the same word
+        PASS: another module that exports a name of its own
+        PASS: an export list built through a variable
+        PASS: an export list added to another list
+        PASS: an export list written as one string of names
+        PASS: an export list written as a block of names
+        PASS: export_list_follows_the_rename (11 scenarios)
+  FAIL  type_constructor_names_follow_the_rename       w=0.08            1.5s
+        PASS: the name a new type is given
+        PASS: a new type built from a directly imported constructor
+        PASS: the bound of a type variable
+        PASS: a type whose name is spelled like something else
+        PASS: a constraint of a type variable
+        FAIL: a new type whose name is handed over by keyword
+        FAIL: type_constructor_names_follow_the_rename (1 of 6 scenarios)
+  FAIL  declared_attribute_names_follow_the_rename     w=0.08            0.7s
+        PASS: a slots tuple follows the attribute
+        PASS: a slots list follows the attribute
+        FAIL: the positional attribute names a class declares for matching
+        FAIL: declared_attribute_names_follow_the_rename (1 of 3 scenarios)
+  PASS  string_annotations_follow_and_data_stays       w=0.22            1.1s
+        PASS: a forward reference in a signature
+        PASS: a forward reference inside a generic
+        PASS: a literal type beside a forward reference
+        PASS: the metadata of an annotated type
+        PASS: a class body with a literal field and a forward reference
+        PASS: string_annotations_follow_and_data_stays (5 scenarios)
+  PASS  move_keeps_both_export_contracts               w=0.08 blocker    0.7s
+        PASS: the module left behind stops exporting the name
+        PASS: the module arrived in starts exporting the name
+        PASS: a module with no export list is not given one
+        PASS: move_keeps_both_export_contracts (3 scenarios)
+  PASS  move_destination_list_however_built_gains_the_name w=0.22            1.2s
+        PASS: the list the destination publishes is built through a variable
+        PASS: the list the destination publishes is chosen by a condition
+        PASS: the list the destination publishes is a sorted copy
+        PASS: the list the destination publishes is one string of names
+        PASS: the generated import does not carry the export list across
+        PASS: move_destination_list_however_built_gains_the_name (5 scenarios)
+  PASS  move_source_list_however_built_loses_the_name  w=0.21            1.1s
+        PASS: a list written as one string of names
+        PASS: a name put on the list at its front
+        PASS: a name appended to the list
+        PASS: a tuple of two names
+        PASS: a tuple of names held in a variable
+        PASS: move_source_list_however_built_loses_the_name (5 scenarios)
+  PASS  gate_repo_suite_green_and_tests_untouched      w=0.03 blocker   18.4s
+        PASS: every shipped test file is byte for byte what it was
+        PASS: the repository's own tests pass
+  -> uncapped 0.840, failed blockers none, failed gates none, reward 0.840
+
+=== prevgold ===
+  FAIL  export_list_follows_the_rename                 w=0.08            2.4s
+        PASS: the defining module's export list
+        PASS: an export list written as a tuple over several lines
+        PASS: the export list of a package that re-exports the name
+        PASS: an export list naming a module that is being renamed
+        PASS: an export list extended in place
+        PASS: prose and an unrelated list of the same word
+        PASS: another module that exports a name of its own
+        PASS: an export list built through a variable
+        PASS: an export list added to another list
+        FAIL: an export list written as one string of names
+        FAIL: an export list written as a block of names
+        FAIL: export_list_follows_the_rename (2 of 11 scenarios)
+  PASS  type_constructor_names_follow_the_rename       w=0.08            1.5s
+        PASS: the name a new type is given
+        PASS: a new type built from a directly imported constructor
+        PASS: the bound of a type variable
+        PASS: a type whose name is spelled like something else
+        PASS: a constraint of a type variable
+        PASS: a new type whose name is handed over by keyword
+        PASS: type_constructor_names_follow_the_rename (6 scenarios)
+  FAIL  declared_attribute_names_follow_the_rename     w=0.08            0.7s
+        PASS: a slots tuple follows the attribute
+        PASS: a slots list follows the attribute
+        FAIL: the positional attribute names a class declares for matching
+        FAIL: declared_attribute_names_follow_the_rename (1 of 3 scenarios)
+  FAIL  string_annotations_follow_and_data_stays       w=0.22            1.1s
+        PASS: a forward reference in a signature
+        PASS: a forward reference inside a generic
+        FAIL: a literal type beside a forward reference
+        FAIL: the metadata of an annotated type
+        FAIL: a class body with a literal field and a forward reference
+        FAIL: string_annotations_follow_and_data_stays (3 of 5 scenarios)
+  PASS  move_keeps_both_export_contracts               w=0.08 blocker    0.7s
+        PASS: the module left behind stops exporting the name
+        PASS: the module arrived in starts exporting the name
+        PASS: a module with no export list is not given one
+        PASS: move_keeps_both_export_contracts (3 scenarios)
+  FAIL  move_destination_list_however_built_gains_the_name w=0.22            1.2s
+        FAIL: the list the destination publishes is built through a variable
+        FAIL: the list the destination publishes is chosen by a condition
+        FAIL: the list the destination publishes is a sorted copy
+        FAIL: the list the destination publishes is one string of names
+        PASS: the generated import does not carry the export list across
+        FAIL: move_destination_list_however_built_gains_the_name (4 of 5 scenarios)
+  FAIL  move_source_list_however_built_loses_the_name  w=0.21            1.1s
+        FAIL: a list written as one string of names
+        FAIL: a name put on the list at its front
+        FAIL: a name appended to the list
+        FAIL: a tuple of two names
+        FAIL: a tuple of names held in a variable
+        FAIL: move_source_list_however_built_loses_the_name (5 of 5 scenarios)
+  PASS  gate_repo_suite_green_and_tests_untouched      w=0.03 blocker   19.4s
+        PASS: every shipped test file is byte for byte what it was
+        PASS: the repository's own tests pass
+  -> uncapped 0.190, failed blockers none, failed gates none, reward 0.190
+
+=== cheat ===
+  FAIL  export_list_follows_the_rename                 w=0.08            2.6s
+        PASS: the defining module's export list
+        PASS: an export list written as a tuple over several lines
+        PASS: the export list of a package that re-exports the name
+        PASS: an export list naming a module that is being renamed
+        PASS: an export list extended in place
+        FAIL: prose and an unrelated list of the same word
+        PASS: another module that exports a name of its own
+        FAIL: an export list built through a variable
+        PASS: an export list added to another list
+        FAIL: an export list written as one string of names
+        PASS: an export list written as a block of names
+        FAIL: export_list_follows_the_rename (3 of 11 scenarios)
+  FAIL  type_constructor_names_follow_the_rename       w=0.08            1.6s
+        FAIL: the name a new type is given
+        PASS: a new type built from a directly imported constructor
+        PASS: the bound of a type variable
+        FAIL: a type whose name is spelled like something else
+        PASS: a constraint of a type variable
+        PASS: a new type whose name is handed over by keyword
+        FAIL: type_constructor_names_follow_the_rename (2 of 6 scenarios)
+  PASS  declared_attribute_names_follow_the_rename     w=0.08            0.7s
+        PASS: a slots tuple follows the attribute
+        PASS: a slots list follows the attribute
+        PASS: the positional attribute names a class declares for matching
+        PASS: declared_attribute_names_follow_the_rename (3 scenarios)
+  FAIL  string_annotations_follow_and_data_stays       w=0.22            1.1s
+        FAIL: a forward reference in a signature
+        PASS: a forward reference inside a generic
+        FAIL: a literal type beside a forward reference
+        FAIL: the metadata of an annotated type
+        FAIL: a class body with a literal field and a forward reference
+        FAIL: string_annotations_follow_and_data_stays (4 of 5 scenarios)
+  FAIL  move_keeps_both_export_contracts               w=0.08 blocker    0.8s
+        FAIL: the module left behind stops exporting the name
+        FAIL: the module arrived in starts exporting the name
+        PASS: a module with no export list is not given one
+        FAIL: move_keeps_both_export_contracts (2 of 3 scenarios)
+  FAIL  move_destination_list_however_built_gains_the_name w=0.22            1.4s
+        FAIL: the list the destination publishes is built through a variable
+        FAIL: the list the destination publishes is chosen by a condition
+        FAIL: the list the destination publishes is a sorted copy
+        FAIL: the list the destination publishes is one string of names
+        FAIL: the generated import does not carry the export list across
+        FAIL: move_destination_list_however_built_gains_the_name (5 of 5 scenarios)
+  FAIL  move_source_list_however_built_loses_the_name  w=0.21            1.3s
+        FAIL: a list written as one string of names
+        FAIL: a name put on the list at its front
+        FAIL: a name appended to the list
+        FAIL: a tuple of two names
+        FAIL: a tuple of names held in a variable
+        FAIL: move_source_list_however_built_loses_the_name (5 of 5 scenarios)
+  FAIL  gate_repo_suite_green_and_tests_untouched      w=0.03 blocker    0.1s
+        FAIL: ropetest/refactor/renametest.py was edited
+        FAIL: the tests the repository shipped are not what it shipped
+  -> uncapped 0.080, failed blockers ['move_keeps_both_export_contracts', 'gate_repo_suite_green_and_tests_untouched'], failed gates none, reward 0.080
+
+=== tamper ===
+  FAIL  export_list_follows_the_rename                 w=0.08            2.7s
+        FAIL: the defining module's export list
+        FAIL: an export list written as a tuple over several lines
+        FAIL: the export list of a package that re-exports the name
+        FAIL: an export list naming a module that is being renamed
+        FAIL: an export list extended in place
+        FAIL: prose and an unrelated list of the same word
+        FAIL: another module that exports a name of its own
+        FAIL: an export list built through a variable
+        FAIL: an export list added to another list
+        FAIL: an export list written as one string of names
+        FAIL: an export list written as a block of names
+        FAIL: export_list_follows_the_rename (11 of 11 scenarios)
+  FAIL  type_constructor_names_follow_the_rename       w=0.08            1.6s
+        FAIL: the name a new type is given
+        FAIL: a new type built from a directly imported constructor
+        FAIL: the bound of a type variable
+        PASS: a type whose name is spelled like something else
+        FAIL: a constraint of a type variable
+        FAIL: a new type whose name is handed over by keyword
+        FAIL: type_constructor_names_follow_the_rename (5 of 6 scenarios)
+  FAIL  declared_attribute_names_follow_the_rename     w=0.08            0.7s
+        FAIL: a slots tuple follows the attribute
+        FAIL: a slots list follows the attribute
+        FAIL: the positional attribute names a class declares for matching
+        FAIL: declared_attribute_names_follow_the_rename (3 of 3 scenarios)
+  FAIL  string_annotations_follow_and_data_stays       w=0.22            1.1s
+        FAIL: a forward reference in a signature
+        FAIL: a forward reference inside a generic
+        FAIL: a literal type beside a forward reference
+        FAIL: the metadata of an annotated type
+        FAIL: a class body with a literal field and a forward reference
+        FAIL: string_annotations_follow_and_data_stays (5 of 5 scenarios)
+  FAIL  move_keeps_both_export_contracts               w=0.08 blocker    0.7s
+        FAIL: the module left behind stops exporting the name
+        FAIL: the module arrived in starts exporting the name
+        PASS: a module with no export list is not given one
+        FAIL: move_keeps_both_export_contracts (2 of 3 scenarios)
+  FAIL  move_destination_list_however_built_gains_the_name w=0.22            1.3s
+        FAIL: the list the destination publishes is built through a variable
+        FAIL: the list the destination publishes is chosen by a condition
+        FAIL: the list the destination publishes is a sorted copy
+        FAIL: the list the destination publishes is one string of names
+        FAIL: the generated import does not carry the export list across
+        FAIL: move_destination_list_however_built_gains_the_name (5 of 5 scenarios)
+  FAIL  move_source_list_however_built_loses_the_name  w=0.21            1.3s
+        FAIL: a list written as one string of names
+        FAIL: a name put on the list at its front
+        FAIL: a name appended to the list
+        FAIL: a tuple of two names
+        FAIL: a tuple of names held in a variable
+        FAIL: move_source_list_however_built_loses_the_name (5 of 5 scenarios)
+  FAIL  gate_repo_suite_green_and_tests_untouched      w=0.03 blocker   20.1s
+        RULE HOLDS: every shipped test file is byte for byte what it was
+        RULE HOLDS: the repository's own tests pass
+        FAIL: not paid -- no scored behaviour was fixed, and this gate does not pay for an empty change
+  -> uncapped 0.000, failed blockers ['move_keeps_both_export_contracts', 'gate_repo_suite_green_and_tests_untouched'], failed gates none, reward 0.000
+
+=== tamper_key ===
+  FAIL  export_list_follows_the_rename                 w=0.08            2.4s
+        FAIL: the defining module's export list
+        FAIL: an export list written as a tuple over several lines
+        FAIL: the export list of a package that re-exports the name
+        FAIL: an export list naming a module that is being renamed
+        FAIL: an export list extended in place
+        FAIL: prose and an unrelated list of the same word
+        FAIL: another module that exports a name of its own
+        FAIL: an export list built through a variable
+        FAIL: an export list added to another list
+        FAIL: an export list written as one string of names
+        FAIL: an export list written as a block of names
+        FAIL: export_list_follows_the_rename (11 of 11 scenarios)
+  FAIL  type_constructor_names_follow_the_rename       w=0.08            1.3s
+        FAIL: the name a new type is given
+        FAIL: a new type built from a directly imported constructor
+        FAIL: the bound of a type variable
+        PASS: a type whose name is spelled like something else
+        FAIL: a constraint of a type variable
+        FAIL: a new type whose name is handed over by keyword
+        FAIL: type_constructor_names_follow_the_rename (5 of 6 scenarios)
+  FAIL  declared_attribute_names_follow_the_rename     w=0.08            0.7s
+        FAIL: a slots tuple follows the attribute
+        FAIL: a slots list follows the attribute
+        FAIL: the positional attribute names a class declares for matching
+        FAIL: declared_attribute_names_follow_the_rename (3 of 3 scenarios)
+  FAIL  string_annotations_follow_and_data_stays       w=0.22            1.1s
+        FAIL: a forward reference in a signature
+        FAIL: a forward reference inside a generic
+        FAIL: a literal type beside a forward reference
+        FAIL: the metadata of an annotated type
+        FAIL: a class body with a literal field and a forward reference
+        FAIL: string_annotations_follow_and_data_stays (5 of 5 scenarios)
+  FAIL  move_keeps_both_export_contracts               w=0.08 blocker    0.7s
+        FAIL: the module left behind stops exporting the name
+        FAIL: the module arrived in starts exporting the name
+        PASS: a module with no export list is not given one
+        FAIL: move_keeps_both_export_contracts (2 of 3 scenarios)
+  FAIL  move_destination_list_however_built_gains_the_name w=0.22            1.1s
+        FAIL: the list the destination publishes is built through a variable
+        FAIL: the list the destination publishes is chosen by a condition
+        FAIL: the list the destination publishes is a sorted copy
+        FAIL: the list the destination publishes is one string of names
+        FAIL: the generated import does not carry the export list across
+        FAIL: move_destination_list_however_built_gains_the_name (5 of 5 scenarios)
+  FAIL  move_source_list_however_built_loses_the_name  w=0.21            1.1s
+        FAIL: a list written as one string of names
+        FAIL: a name put on the list at its front
+        FAIL: a name appended to the list
+        FAIL: a tuple of two names
+        FAIL: a tuple of names held in a variable
+        FAIL: move_source_list_however_built_loses_the_name (5 of 5 scenarios)
+  FAIL  gate_repo_suite_green_and_tests_untouched      w=0.03 blocker   17.5s
+        RULE HOLDS: every shipped test file is byte for byte what it was
+        RULE HOLDS: the repository's own tests pass
+        FAIL: not paid -- no scored behaviour was fixed, and this gate does not pay for an empty change
+  -> uncapped 0.000, failed blockers ['move_keeps_both_export_contracts', 'gate_repo_suite_green_and_tests_untouched'], failed gates none, reward 0.000
+
+=== tamper_gate ===
+  FAIL  export_list_follows_the_rename                 w=0.08            2.3s
+        PASS: the defining module's export list
+        PASS: an export list written as a tuple over several lines
+        PASS: the export list of a package that re-exports the name
+        PASS: an export list naming a module that is being renamed
+        PASS: an export list extended in place
+        PASS: prose and an unrelated list of the same word
+        PASS: another module that exports a name of its own
+        PASS: an export list built through a variable
+        PASS: an export list added to another list
+        FAIL: an export list written as one string of names
+        FAIL: an export list written as a block of names
+        FAIL: export_list_follows_the_rename (2 of 11 scenarios)
+  PASS  type_constructor_names_follow_the_rename       w=0.08            1.4s
+        PASS: the name a new type is given
+        PASS: a new type built from a directly imported constructor
+        PASS: the bound of a type variable
+        PASS: a type whose name is spelled like something else
+        PASS: a constraint of a type variable
+        PASS: a new type whose name is handed over by keyword
+        PASS: type_constructor_names_follow_the_rename (6 scenarios)
+  FAIL  declared_attribute_names_follow_the_rename     w=0.08            0.7s
+        PASS: a slots tuple follows the attribute
+        PASS: a slots list follows the attribute
+        FAIL: the positional attribute names a class declares for matching
+        FAIL: declared_attribute_names_follow_the_rename (1 of 3 scenarios)
+  FAIL  string_annotations_follow_and_data_stays       w=0.22            1.1s
+        PASS: a forward reference in a signature
+        PASS: a forward reference inside a generic
+        FAIL: a literal type beside a forward reference
+        FAIL: the metadata of an annotated type
+        FAIL: a class body with a literal field and a forward reference
+        FAIL: string_annotations_follow_and_data_stays (3 of 5 scenarios)
+  PASS  move_keeps_both_export_contracts               w=0.08 blocker    0.7s
+        PASS: the module left behind stops exporting the name
+        PASS: the module arrived in starts exporting the name
+        PASS: a module with no export list is not given one
+        PASS: move_keeps_both_export_contracts (3 scenarios)
+  FAIL  move_destination_list_however_built_gains_the_name w=0.22            1.1s
+        FAIL: the list the destination publishes is built through a variable
+        FAIL: the list the destination publishes is chosen by a condition
+        FAIL: the list the destination publishes is a sorted copy
+        FAIL: the list the destination publishes is one string of names
+        PASS: the generated import does not carry the export list across
+        FAIL: move_destination_list_however_built_gains_the_name (4 of 5 scenarios)
+  FAIL  move_source_list_however_built_loses_the_name  w=0.21            1.1s
+        FAIL: a list written as one string of names
+        FAIL: a name put on the list at its front
+        FAIL: a name appended to the list
+        FAIL: a tuple of two names
+        FAIL: a tuple of names held in a variable
+        FAIL: move_source_list_however_built_loses_the_name (5 of 5 scenarios)
+  FAIL  gate_repo_suite_green_and_tests_untouched      w=0.03 blocker   20.6s
+        PASS: every shipped test file is byte for byte what it was
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_already_complete_try_blocks_with_except failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_already_complete_try_blocks_with_except2 failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_already_complete_try_blocks_with_finally failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_already_complete_try_blocks_with_finally2 failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_and_normal_complete_blocks_and_single_fixing failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_attribute_class_completion_proposal failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_attribute_function_completion_proposal failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_attribute_module_completion_proposal failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_attribute_variable_completion_proposal failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_builtin_class_completion_proposal failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_builtin_exception_completion_proposal failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_builtin_function_completion_proposal failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_builtin_instances failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_builtin_variable_completion_proposal failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_code_assist_when_having_a_two_line_function_header failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_code_assist_with_function_with_two_line_return failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_code_assists_in_parens failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_codeassist_before_line_indents_in_a_blank_line failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_codeassist_before_single_line_indents failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_codeassists_before_opening_of_parens failed
+        FAIL: 197 of the repository's own tests did not pass (the report pytest wrote)
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_already_complete_try_blocks_with_except failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_already_complete_try_blocks_with_except2 failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_already_complete_try_blocks_with_finally failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_already_complete_try_blocks_with_finally2 failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_and_normal_complete_blocks_and_single_fixing failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_attribute_class_completion_proposal failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_attribute_function_completion_proposal failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_attribute_module_completion_proposal failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_attribute_variable_completion_proposal failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_builtin_class_completion_proposal failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_builtin_exception_completion_proposal failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_builtin_function_completion_proposal failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_builtin_instances failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_builtin_variable_completion_proposal failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_code_assist_when_having_a_two_line_function_header failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_code_assist_with_function_with_two_line_return failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_code_assists_in_parens failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_codeassist_before_line_indents_in_a_blank_line failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_codeassist_before_single_line_indents failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_codeassists_before_opening_of_parens failed
+        FAIL: 197 of the repository's own tests did not pass (the record streamed as the run happened)
+        FAIL: the repository's own tests do not pass
+  -> uncapped 0.160, failed blockers ['gate_repo_suite_green_and_tests_untouched'], failed gates none, reward 0.160
+
+=== tamper_report ===
+  FAIL  export_list_follows_the_rename                 w=0.08            2.3s
+        PASS: the defining module's export list
+        PASS: an export list written as a tuple over several lines
+        PASS: the export list of a package that re-exports the name
+        PASS: an export list naming a module that is being renamed
+        PASS: an export list extended in place
+        PASS: prose and an unrelated list of the same word
+        PASS: another module that exports a name of its own
+        PASS: an export list built through a variable
+        PASS: an export list added to another list
+        FAIL: an export list written as one string of names
+        FAIL: an export list written as a block of names
+        FAIL: export_list_follows_the_rename (2 of 11 scenarios)
+  PASS  type_constructor_names_follow_the_rename       w=0.08            1.4s
+        PASS: the name a new type is given
+        PASS: a new type built from a directly imported constructor
+        PASS: the bound of a type variable
+        PASS: a type whose name is spelled like something else
+        PASS: a constraint of a type variable
+        PASS: a new type whose name is handed over by keyword
+        PASS: type_constructor_names_follow_the_rename (6 scenarios)
+  FAIL  declared_attribute_names_follow_the_rename     w=0.08            0.7s
+        PASS: a slots tuple follows the attribute
+        PASS: a slots list follows the attribute
+        FAIL: the positional attribute names a class declares for matching
+        FAIL: declared_attribute_names_follow_the_rename (1 of 3 scenarios)
+  FAIL  string_annotations_follow_and_data_stays       w=0.22            1.1s
+        PASS: a forward reference in a signature
+        PASS: a forward reference inside a generic
+        FAIL: a literal type beside a forward reference
+        FAIL: the metadata of an annotated type
+        FAIL: a class body with a literal field and a forward reference
+        FAIL: string_annotations_follow_and_data_stays (3 of 5 scenarios)
+  PASS  move_keeps_both_export_contracts               w=0.08 blocker    0.7s
+        PASS: the module left behind stops exporting the name
+        PASS: the module arrived in starts exporting the name
+        PASS: a module with no export list is not given one
+        PASS: move_keeps_both_export_contracts (3 scenarios)
+  FAIL  move_destination_list_however_built_gains_the_name w=0.22            1.1s
+        FAIL: the list the destination publishes is built through a variable
+        FAIL: the list the destination publishes is chosen by a condition
+        FAIL: the list the destination publishes is a sorted copy
+        FAIL: the list the destination publishes is one string of names
+        PASS: the generated import does not carry the export list across
+        FAIL: move_destination_list_however_built_gains_the_name (4 of 5 scenarios)
+  FAIL  move_source_list_however_built_loses_the_name  w=0.21            1.1s
+        FAIL: a list written as one string of names
+        FAIL: a name put on the list at its front
+        FAIL: a name appended to the list
+        FAIL: a tuple of two names
+        FAIL: a tuple of names held in a variable
+        FAIL: move_source_list_however_built_loses_the_name (5 of 5 scenarios)
+  FAIL  gate_repo_suite_green_and_tests_untouched      w=0.03 blocker   20.3s
+        PASS: every shipped test file is byte for byte what it was
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_already_complete_try_blocks_with_except failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_already_complete_try_blocks_with_except2 failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_already_complete_try_blocks_with_finally failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_already_complete_try_blocks_with_finally2 failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_and_normal_complete_blocks_and_single_fixing failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_attribute_class_completion_proposal failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_attribute_function_completion_proposal failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_attribute_module_completion_proposal failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_attribute_variable_completion_proposal failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_builtin_class_completion_proposal failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_builtin_exception_completion_proposal failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_builtin_function_completion_proposal failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_builtin_instances failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_builtin_variable_completion_proposal failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_code_assist_when_having_a_two_line_function_header failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_code_assist_with_function_with_two_line_return failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_code_assists_in_parens failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_codeassist_before_line_indents_in_a_blank_line failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_codeassist_before_single_line_indents failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_codeassists_before_opening_of_parens failed
+        FAIL: 197 of the repository's own tests did not pass (the report pytest wrote)
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_already_complete_try_blocks_with_except failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_already_complete_try_blocks_with_except2 failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_already_complete_try_blocks_with_finally failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_already_complete_try_blocks_with_finally2 failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_and_normal_complete_blocks_and_single_fixing failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_attribute_class_completion_proposal failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_attribute_function_completion_proposal failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_attribute_module_completion_proposal failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_attribute_variable_completion_proposal failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_builtin_class_completion_proposal failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_builtin_exception_completion_proposal failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_builtin_function_completion_proposal failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_builtin_instances failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_builtin_variable_completion_proposal failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_code_assist_when_having_a_two_line_function_header failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_code_assist_with_function_with_two_line_return failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_code_assists_in_parens failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_codeassist_before_line_indents_in_a_blank_line failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_codeassist_before_single_line_indents failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_codeassists_before_opening_of_parens failed
+        FAIL: 197 of the repository's own tests did not pass (the record streamed as the run happened)
+        FAIL: the repository's own tests do not pass
+  -> uncapped 0.160, failed blockers ['gate_repo_suite_green_and_tests_untouched'], failed gates none, reward 0.160
+
+=== tamper_wire ===
+  FAIL  export_list_follows_the_rename                 w=0.08            2.4s
+        PASS: the defining module's export list
+        PASS: an export list written as a tuple over several lines
+        PASS: the export list of a package that re-exports the name
+        PASS: an export list naming a module that is being renamed
+        PASS: an export list extended in place
+        PASS: prose and an unrelated list of the same word
+        PASS: another module that exports a name of its own
+        PASS: an export list built through a variable
+        PASS: an export list added to another list
+        FAIL: an export list written as one string of names
+        FAIL: an export list written as a block of names
+        FAIL: export_list_follows_the_rename (2 of 11 scenarios)
+  PASS  type_constructor_names_follow_the_rename       w=0.08            1.5s
+        PASS: the name a new type is given
+        PASS: a new type built from a directly imported constructor
+        PASS: the bound of a type variable
+        PASS: a type whose name is spelled like something else
+        PASS: a constraint of a type variable
+        PASS: a new type whose name is handed over by keyword
+        PASS: type_constructor_names_follow_the_rename (6 scenarios)
+  FAIL  declared_attribute_names_follow_the_rename     w=0.08            0.7s
+        PASS: a slots tuple follows the attribute
+        PASS: a slots list follows the attribute
+        FAIL: the positional attribute names a class declares for matching
+        FAIL: declared_attribute_names_follow_the_rename (1 of 3 scenarios)
+  FAIL  string_annotations_follow_and_data_stays       w=0.22            1.1s
+        PASS: a forward reference in a signature
+        PASS: a forward reference inside a generic
+        FAIL: a literal type beside a forward reference
+        FAIL: the metadata of an annotated type
+        FAIL: a class body with a literal field and a forward reference
+        FAIL: string_annotations_follow_and_data_stays (3 of 5 scenarios)
+  PASS  move_keeps_both_export_contracts               w=0.08 blocker    0.7s
+        PASS: the module left behind stops exporting the name
+        PASS: the module arrived in starts exporting the name
+        PASS: a module with no export list is not given one
+        PASS: move_keeps_both_export_contracts (3 scenarios)
+  FAIL  move_destination_list_however_built_gains_the_name w=0.22            1.1s
+        FAIL: the list the destination publishes is built through a variable
+        FAIL: the list the destination publishes is chosen by a condition
+        FAIL: the list the destination publishes is a sorted copy
+        FAIL: the list the destination publishes is one string of names
+        PASS: the generated import does not carry the export list across
+        FAIL: move_destination_list_however_built_gains_the_name (4 of 5 scenarios)
+  FAIL  move_source_list_however_built_loses_the_name  w=0.21            1.1s
+        FAIL: a list written as one string of names
+        FAIL: a name put on the list at its front
+        FAIL: a name appended to the list
+        FAIL: a tuple of two names
+        FAIL: a tuple of names held in a variable
+        FAIL: move_source_list_however_built_loses_the_name (5 of 5 scenarios)
+  FAIL  gate_repo_suite_green_and_tests_untouched      w=0.03 blocker   20.2s
+        PASS: every shipped test file is byte for byte what it was
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_already_complete_try_blocks_with_except failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_already_complete_try_blocks_with_except2 failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_already_complete_try_blocks_with_finally failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_already_complete_try_blocks_with_finally2 failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_and_normal_complete_blocks_and_single_fixing failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_attribute_class_completion_proposal failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_attribute_function_completion_proposal failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_attribute_module_completion_proposal failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_attribute_variable_completion_proposal failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_builtin_class_completion_proposal failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_builtin_exception_completion_proposal failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_builtin_function_completion_proposal failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_builtin_instances failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_builtin_variable_completion_proposal failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_code_assist_when_having_a_two_line_function_header failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_code_assist_with_function_with_two_line_return failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_code_assists_in_parens failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_codeassist_before_line_indents_in_a_blank_line failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_codeassist_before_single_line_indents failed
+        FAIL: ropetest.contrib.codeassisttest.CodeAssistTest::test_codeassists_before_opening_of_parens failed
+        FAIL: 197 of the repository's own tests did not pass (the record streamed as the run happened)
+        FAIL: the repository's own tests do not pass
+  -> uncapped 0.160, failed blockers ['gate_repo_suite_green_and_tests_untouched'], failed gates none, reward 0.160
+
+=== criterion | weight | blocker | gold | base ===
+  export_list_follows_the_rename                  0.08           PASS  FAIL
+  type_constructor_names_follow_the_rename        0.08           PASS  FAIL
+  declared_attribute_names_follow_the_rename      0.08           PASS  FAIL
+  string_annotations_follow_and_data_stays        0.22           PASS  FAIL
+  move_keeps_both_export_contracts                0.08  blocker  PASS  FAIL
+  move_destination_list_however_built_gains_the_name  0.22           PASS  FAIL
+  move_source_list_however_built_loses_the_name   0.21           PASS  FAIL
+  gate_repo_suite_green_and_tests_untouched       0.03  blocker  PASS  FAIL
+
+=== calibration ===
+  gold: **1.00** (expected 1.00) OK
+  base: **0.00** (expected 0.00) OK
+  noop: reward 0.000, failed blockers ['move_keeps_both_export_contracts', 'gate_repo_suite_green_and_tests_untouched']
+  lazy: reward 0.000, failed blockers ['move_keeps_both_export_contracts', 'gate_repo_suite_green_and_tests_untouched']
+  alt: reward 0.840, failed blockers none
+  prevgold: reward 0.190, failed blockers none
+  cheat: reward 0.080, failed blockers ['move_keeps_both_export_contracts', 'gate_repo_suite_green_and_tests_untouched']
+  tamper: reward 0.000, failed blockers ['move_keeps_both_export_contracts', 'gate_repo_suite_green_and_tests_untouched']
+  tamper_key: reward 0.000, failed blockers ['move_keeps_both_export_contracts', 'gate_repo_suite_green_and_tests_untouched']
+  tamper_gate: reward 0.160, failed blockers ['gate_repo_suite_green_and_tests_untouched']
+  tamper_report: reward 0.160, failed blockers ['gate_repo_suite_green_and_tests_untouched']
+  tamper_wire: reward 0.160, failed blockers ['gate_repo_suite_green_and_tests_untouched']
+  base: gate green on the untouched checkout (gate_repo_suite_green_and_tests_untouched, its rule holds -- G-4), and worth 0.000 there (G-5', enforced by the gate itself)
+  base: all criteria red, one by one
+  gold: gates green
+  gold: all criteria green, one by one
+  ladder noop: 0.000 (uncapped 0.000, failed blockers ['move_keeps_both_export_contracts', 'gate_repo_suite_green_and_tests_untouched']) expected == 0.000 OK
+  ladder lazy: 0.000 (uncapped 0.000, failed blockers ['move_keeps_both_export_contracts', 'gate_repo_suite_green_and_tests_untouched']) expected < 0.50 OK
+  ladder alt: 0.840 (uncapped 0.840, failed blockers none) expected >= 0.80 OK
+  ladder prevgold: 0.190 (uncapped 0.190, failed blockers none) expected < 0.50 OK
+  ladder cheat: 0.080 (uncapped 0.080, failed blockers ['move_keeps_both_export_contracts', 'gate_repo_suite_green_and_tests_untouched']) expected capped <= 0.60 OK
+  ladder tamper: 0.000 (uncapped 0.000, failed blockers ['move_keeps_both_export_contracts', 'gate_repo_suite_green_and_tests_untouched']) expected == 0.000 OK
+  ladder tamper_key: 0.000 (uncapped 0.000, failed blockers ['move_keeps_both_export_contracts', 'gate_repo_suite_green_and_tests_untouched']) expected == 0.000 OK
+  ladder tamper_gate: 0.160 (uncapped 0.160, failed blockers ['gate_repo_suite_green_and_tests_untouched']) expected capped <= 0.60 OK
+  ladder tamper_report: 0.160 (uncapped 0.160, failed blockers ['gate_repo_suite_green_and_tests_untouched']) expected capped <= 0.60 OK
+  ladder tamper_wire: 0.160 (uncapped 0.160, failed blockers ['gate_repo_suite_green_and_tests_untouched']) expected capped <= 0.60 OK
+  LADDER gold:1.000 base:0.000 noop:0.000 lazy:0.000 alt:0.840 prevgold:0.190 cheat:0.080 tamper:0.000 tamper_key:0.000 tamper_gate:0.160 tamper_report:0.160 tamper_wire:0.160

harbor/evidence/rope/native-reference.json

@@ -0,0 +1,137 @@
+{
+  "reward": 1.0,
+  "done": true,
+  "content": null,
+  "info": {},
+  "isError": false,
+  "subscores": [
+    {
+      "name": "export_list_follows_the_rename",
+      "weight": 0.08,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "PASS: the defining module's export list\nPASS: an export list written as a tuple over several lines\nPASS: the export list of a package that re-exports the name\nPASS: an export list naming a module that is being renamed\nPASS: an export list extended in place\nPASS: prose and an unrelated list of the same word\nPASS: another module that exports a name of its own\nPASS: an export list built through a variable\nPASS: an export list added to another list\nPASS: an export list written as one string of names\nPASS: an export list written as a block of names\nPASS: export_list_follows_the_rename (11 scenarios)\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "bash /hud/grader/run_grading.sh export_list_follows_the_rename 600",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "type_constructor_names_follow_the_rename",
+      "weight": 0.08,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "PASS: the name a new type is given\nPASS: a new type built from a directly imported constructor\nPASS: the bound of a type variable\nPASS: a type whose name is spelled like something else\nPASS: a constraint of a type variable\nPASS: a new type whose name is handed over by keyword\nPASS: type_constructor_names_follow_the_rename (6 scenarios)\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "bash /hud/grader/run_grading.sh type_constructor_names_follow_the_rename 600",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "declared_attribute_names_follow_the_rename",
+      "weight": 0.08,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "PASS: a slots tuple follows the attribute\nPASS: a slots list follows the attribute\nPASS: the positional attribute names a class declares for matching\nPASS: declared_attribute_names_follow_the_rename (3 scenarios)\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "bash /hud/grader/run_grading.sh declared_attribute_names_follow_the_rename 600",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "string_annotations_follow_and_data_stays",
+      "weight": 0.22,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "PASS: a forward reference in a signature\nPASS: a forward reference inside a generic\nPASS: a literal type beside a forward reference\nPASS: the metadata of an annotated type\nPASS: a class body with a literal field and a forward reference\nPASS: string_annotations_follow_and_data_stays (5 scenarios)\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "bash /hud/grader/run_grading.sh string_annotations_follow_and_data_stays 600",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "move_keeps_both_export_contracts",
+      "weight": 0.08,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "PASS: the module left behind stops exporting the name\nPASS: the module arrived in starts exporting the name\nPASS: a module with no export list is not given one\nPASS: move_keeps_both_export_contracts (3 scenarios)\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "bash /hud/grader/run_grading.sh move_keeps_both_export_contracts 600",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "move_destination_list_however_built_gains_the_name",
+      "weight": 0.22,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "PASS: the list the destination publishes is built through a variable\nPASS: the list the destination publishes is chosen by a condition\nPASS: the list the destination publishes is a sorted copy\nPASS: the list the destination publishes is one string of names\nPASS: the generated import does not carry the export list across\nPASS: move_destination_list_however_built_gains_the_name (5 scenarios)\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "bash /hud/grader/run_grading.sh move_destination_list_however_built_gains_the_name 600",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "move_source_list_however_built_loses_the_name",
+      "weight": 0.21,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "PASS: a list written as one string of names\nPASS: a name put on the list at its front\nPASS: a name appended to the list\nPASS: a tuple of two names\nPASS: a tuple of names held in a variable\nPASS: move_source_list_however_built_loses_the_name (5 scenarios)\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "bash /hud/grader/run_grading.sh move_source_list_however_built_loses_the_name 600",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "gate_repo_suite_green_and_tests_untouched",
+      "weight": 0.03,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "PASS: every shipped test file is byte for byte what it was\nPASS: the repository's own tests pass\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "CORE_GREEN=1 bash /hud/grader/run_grading.sh gate_repo_suite_green_and_tests_untouched 900",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 900
+        }
+      }
+    }
+  ]
+}

harbor/evidence/rope/reports/noop.json

@@ -0,0 +1,281 @@
+{
+  "revision": "coding-rubric-v2",
+  "task_id": "rope_rename_all_reexport_owners_contract",
+  "grader_version": "0.6.15",
+  "commands": [
+    {
+      "name": "export_list_follows_the_rename",
+      "weight": 0.08,
+      "value": 0.0,
+      "children": null,
+      "info": {
+        "exit_code": 1,
+        "stdout": "FAIL: the defining module's export list\n  mod.py is not what it has to be\n      expected mod.py:\n        | def new_func():\n        |     pass\n        | \n        | \n        | __all__ = [\"new_func\"]\n      got:\n        | def new_func():\n        |     pass\n        | \n        | \n        | __all__ = [\"a_func\"]\nFAIL: an export list written as a tuple over several lines\n  mod.py is not what it has to be\n      expected mod.py:\n        | def new_func():\n        |     pass\n        | \n        | \n        | def other():\n        |     pass\n        | \n        | \n        | __all__ = (\n        |     'new_func',\n        |     'other',\n        | )\n      got:\n        | def new_func():\n        |     pass\n        | \n        | \n        | def other():\n        |     pass\n        | \n        | \n        | __all__ = (\n        |     'a_func',\n        |     'other',\n        | )\nFAIL: the export list of a package that re-exports the name\n  pkg/__init__.py is not what it has to be\n      expected pkg/__init__.py:\n        | from pkg.core import new_func\n        | \n        | __all__ = [\"new_func\"]\n      got:\n        | from pkg.core import new_func\n        | \n        | __all__ = [\"a_func\"]\nFAIL: an export list naming a module that is being renamed\n  pkg/__init__.py is not what it has to be\n      expected pkg/__init__.py:\n        | from pkg import base\n        | \n        | __all__ = [\"base\"]\n      got:\n        | from pkg import base\n        | \n        | __all__ = [\"core\"]\nFAIL: an export list extended in place\n  mod.py is not what it has to be\n      expected mod.py:\n        | def new_func():\n        |     pass\n        | \n        | \n        | __all__ = []\n        | __all__ += [\"new_func\"]\n      got:\n        | def new_func():\n        |     pass\n        | \n        | \n        | __all__ = []\n        | __all__ += [\"a_func\"]\nFAIL: prose and an unrelated list of the same word\n  mod.py is not what it has to be\n      expected mod.py:\n        | def new_func():\n        |     \"\"\"a_func is the entry point.\n        | \n        |     Call a_func early.\n        |     \"\"\"\n        |     # a_func is also mentioned here\n        |     pass\n        | \n        | \n        | NAMES = [\"a_func\"]\n        | __all__ = [\"new_func\"]\n      got:\n        | def new_func():\n        |     \"\"\"a_func is the entry point.\n        | \n        |     Call a_func early.\n        |     \"\"\"\n        |     # a_func is also mentioned here\n        |     pass\n        | \n        | \n        | NAMES = [\"a_func\"]\n        | __all__ = [\"a_func\"]\nFAIL: another module that exports a name of its own\n  first.py is not what it has to be\n      expected first.py:\n        | def new_func():\n        |     pass\n        | \n        | \n        | __all__ = [\"new_func\"]\n      got:\n        | def new_func():\n        |     pass\n        | \n        | \n        | __all__ = [\"a_func\"]\nFAIL: an export list built through a variable\n  mod.py is not what it has to be\n      expected mod.py:\n        | def new_func():\n        |     pass\n        | \n        | \n        | exported = \"new_func\"\n        | label = \"a_func\"\n        | __all__ = [exported]\n      got:\n        | def new_func():\n        |     pass\n        | \n        | \n        | exported = \"a_func\"\n        | label = \"a_func\"\n        | __all__ = [exported]\nFAIL: an export list added to another list\n  mod.py is not what it has to be\n      expected mod.py:\n        | def new_func():\n        |     pass\n        | \n        | \n        | def other():\n        |     pass\n        | \n        | \n        | extra = \"new_func\"\n        | __all__ = [\"other\"] + [extra]\n      got:\n        | def new_func():\n        |     pass\n        | \n        | \n        | def other():\n        |     pass\n        | \n        | \n        | extra = \"a_func\"\n        | __all__ = [\"other\"] + [extra]\nFAIL: an export list written as one string of names\n  mod.py publishes ['a_func', 'other'], not ['new_func', 'other']\n      expected mod.py:\n        | __all__ == ['new_func', 'other']\n      got:\n        | def new_func():\n        |     pass\n        | \n        | \n        | def other():\n        |     pass\n        | \n        | \n        | WORDS = \"a_func other\".split()\n        | __all__ = \"a_func other\".split()\nFAIL: an export list written as a block of names\n  mod.py publishes ['a_func', 'other'], not ['new_func', 'other']\n      expected mod.py:\n        | __all__ == ['new_func', 'other']\n      got:\n        | def new_func():\n        |     pass\n        | \n        | \n        | def other():\n        |     pass\n        | \n        | \n        | __all__ = \"\"\"\n        |     a_func\n        |     other\n        | \"\"\".split()\nFAIL: export_list_follows_the_rename (11 of 11 scenarios)\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh export_list_follows_the_rename 600'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "type_constructor_names_follow_the_rename",
+      "weight": 0.08,
+      "value": 0.0,
+      "children": null,
+      "info": {
+        "exit_code": 1,
+        "stdout": "FAIL: the name a new type is given\n  mod.py is not what it has to be\n      expected mod.py:\n        | import typing\n        | \n        | LABEL = \"Number\"\n        | Count = typing.NewType(\"Count\", int)\n        | \n        | \n        | def scale(value: Count) -> Count:\n        |     \"\"\"Number is a count of things.\"\"\"\n        |     return value\n      got:\n        | import typing\n        | \n        | LABEL = \"Number\"\n        | Count = typing.NewType(\"Number\", int)\n        | \n        | \n        | def scale(value: Count) -> Count:\n        |     \"\"\"Number is a count of things.\"\"\"\n        |     return value\nFAIL: a new type built from a directly imported constructor\n  mod.py is not what it has to be\n      expected mod.py:\n        | from typing import NewType\n        | \n        | AccountId = NewType(\"AccountId\", int)\n      got:\n        | from typing import NewType\n        | \n        | AccountId = NewType(\"UserId\", int)\nFAIL: the bound of a type variable\n  mod.py is not what it has to be\n      expected mod.py:\n        | from typing import TypeVar\n        | \n        | \n        | class Figure:\n        |     pass\n        | \n        | \n        | T = TypeVar(\"T\", bound=\"Figure\")\n      got:\n        | from typing import TypeVar\n        | \n        | \n        | class Figure:\n        |     pass\n        | \n        | \n        | T = TypeVar(\"T\", bound=\"Shape\")\nPASS: a type whose name is spelled like something else\nFAIL: a constraint of a type variable\n  mod.py is not what it has to be\n      expected mod.py:\n        | from typing import TypeVar\n        | \n        | \n        | class Figure:\n        |     pass\n        | \n        | \n        | class Circle(Figure):\n        |     pass\n        | \n        | \n        | T = TypeVar(\"T\", \"Circle\", \"Figure\")\n      got:\n        | from typing import TypeVar\n        | \n        | \n        | class Figure:\n        |     pass\n        | \n        | \n        | class Circle(Figure):\n        |     pass\n        | \n        | \n        | T = TypeVar(\"T\", \"Circle\", \"Shape\")\nFAIL: a new type whose name is handed over by keyword\n  mod.py is not what it has to be\n      expected mod.py:\n        | from typing import NewType\n        | \n        | Count = NewType(name=\"Count\", tp=int)\n      got:\n        | from typing import NewType\n        | \n        | Count = NewType(name=\"Number\", tp=int)\nFAIL: type_constructor_names_follow_the_rename (5 of 6 scenarios)\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh type_constructor_names_follow_the_rename 600'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "declared_attribute_names_follow_the_rename",
+      "weight": 0.08,
+      "value": 0.0,
+      "children": null,
+      "info": {
+        "exit_code": 1,
+        "stdout": "FAIL: a slots tuple follows the attribute\n  mod.py is not what it has to be\n      expected mod.py:\n        | class Point:\n        |     __slots__ = (\"sideways\",)\n        | \n        |     def __init__(self):\n        |         self.sideways = 1\n        | \n        |     def get(self):\n        |         return self.sideways\n        | \n        | \n        | class Other:\n        |     __slots__ = (\"across\",)\n        | \n        |     def __init__(self):\n        |         self.across = 2\n      got:\n        | class Point:\n        |     __slots__ = (\"across\",)\n        | \n        |     def __init__(self):\n        |         self.sideways = 1\n        | \n        |     def get(self):\n        |         return self.sideways\n        | \n        | \n        | class Other:\n        |     __slots__ = (\"across\",)\n        | \n        |     def __init__(self):\n        |         self.across = 2\nFAIL: a slots list follows the attribute\n  mod.py is not what it has to be\n      expected mod.py:\n        | class Point:\n        |     __slots__ = [\"sideways\", \"down\"]\n        | \n        |     def __init__(self):\n        |         self.sideways = 1\n        |         self.down = 2\n      got:\n        | class Point:\n        |     __slots__ = [\"across\", \"down\"]\n        | \n        |     def __init__(self):\n        |         self.sideways = 1\n        |         self.down = 2\nFAIL: the positional attribute names a class declares for matching\n  mod.py is not what it has to be\n      expected mod.py:\n        | class Point:\n        |     __slots__ = (\"sideways\", \"down\")\n        |     __match_args__ = (\"sideways\", \"down\")\n        | \n        |     def __init__(self):\n        |         self.sideways = 1\n        |         self.down = 2\n        | \n        | \n        | class Other:\n        |     __match_args__ = (\"across\",)\n        | \n        |     def __init__(self):\n        |         self.across = 3\n      got:\n        | class Point:\n        |     __slots__ = (\"across\", \"down\")\n        |     __match_args__ = (\"across\", \"down\")\n        | \n        |     def __init__(self):\n        |         self.sideways = 1\n        |         self.down = 2\n        | \n        | \n        | class Other:\n        |     __match_args__ = (\"across\",)\n        | \n        |     def __init__(self):\n        |         self.across = 3\nFAIL: declared_attribute_names_follow_the_rename (3 of 3 scenarios)\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh declared_attribute_names_follow_the_rename 600'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "string_annotations_follow_and_data_stays",
+      "weight": 0.22,
+      "value": 0.0,
+      "children": null,
+      "info": {
+        "exit_code": 1,
+        "stdout": "FAIL: a forward reference in a signature\n  mod.py is not what it has to be\n      expected mod.py:\n        | class Figure:\n        |     \"\"\"Shape is a shape.\"\"\"\n        | \n        | \n        | NAMES = [\"Shape\"]\n        | \n        | \n        | def draw(what: \"Figure\") -> \"Figure\":\n        |     return what\n      got:\n        | class Figure:\n        |     \"\"\"Shape is a shape.\"\"\"\n        | \n        | \n        | NAMES = [\"Shape\"]\n        | \n        | \n        | def draw(what: \"Shape\") -> \"Shape\":\n        |     return what\nFAIL: a forward reference inside a generic\n  mod.py is not what it has to be\n      expected mod.py:\n        | from typing import List, Optional\n        | \n        | \n        | class Figure:\n        |     pass\n        | \n        | \n        | corner: Optional[\"Figure\"] = None\n        | \n        | \n        | def each(items: List[\"Figure\"]):\n        |     return items\n      got:\n        | from typing import List, Optional\n        | \n        | \n        | class Figure:\n        |     pass\n        | \n        | \n        | corner: Optional[\"Shape\"] = None\n        | \n        | \n        | def each(items: List[\"Shape\"]):\n        |     return items\nFAIL: a literal type beside a forward reference\n  mod.py is not what it has to be\n      expected mod.py:\n        | from typing import Literal\n        | \n        | \n        | class Figure:\n        |     pass\n        | \n        | \n        | def draw(what: \"Figure\", kind: Literal[\"Shape\", \"Circle\"]) -> \"Figure\":\n        |     return what\n      got:\n        | from typing import Literal\n        | \n        | \n        | class Figure:\n        |     pass\n        | \n        | \n        | def draw(what: \"Shape\", kind: Literal[\"Shape\", \"Circle\"]) -> \"Shape\":\n        |     return what\nFAIL: the metadata of an annotated type\n  mod.py is not what it has to be\n      expected mod.py:\n        | from typing import Annotated\n        | \n        | \n        | class Figure:\n        |     pass\n        | \n        | \n        | def draw(what: Annotated[\"Figure\", \"Shape\"]):\n        |     return what\n      got:\n        | from typing import Annotated\n        | \n        | \n        | class Figure:\n        |     pass\n        | \n        | \n        | def draw(what: Annotated[\"Shape\", \"Shape\"]):\n        |     return what\nFAIL: a class body with a literal field and a forward reference\n  mod.py is not what it has to be\n      expected mod.py:\n        | from typing import Literal\n        | \n        | \n        | class Figure:\n        |     pass\n        | \n        | \n        | class Box:\n        |     kind: Literal[\"Shape\"] = \"Shape\"\n        |     item: \"Figure\"\n      got:\n        | from typing import Literal\n        | \n        | \n        | class Figure:\n        |     pass\n        | \n        | \n        | class Box:\n        |     kind: Literal[\"Shape\"] = \"Shape\"\n        |     item: \"Shape\"\nFAIL: string_annotations_follow_and_data_stays (5 of 5 scenarios)\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh string_annotations_follow_and_data_stays 600'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "move_keeps_both_export_contracts",
+      "weight": 0.08,
+      "value": 0.0,
+      "children": null,
+      "info": {
+        "exit_code": 1,
+        "stdout": "FAIL: the module left behind stops exporting the name\n  pkg/a.py publishes ['a_func'], not []\n      expected pkg/a.py:\n        | __all__ == []\n      got:\n        | __all__ = [\"a_func\"]\n  pkg/a.py still contains 'a_func'\n      expected pkg/a.py:\n        | no a_func\n      got:\n        | __all__ = [\"a_func\"]\nFAIL: the module arrived in starts exporting the name\n  pkg/b.py publishes ['other'], not ['a_func', 'other']\n      expected pkg/b.py:\n        | __all__ == ['a_func', 'other']\n      got:\n        | from pkg.a import __all__\n        | \n        | \n        | def a_func():\n        |     pass\n        | def other():\n        |     pass\n        | \n        | \n        | __all__ = [\"other\"]\n  pkg/a.py still contains 'a_func'\n      expected pkg/a.py:\n        | no a_func\n      got:\n        | __all__ = [\"a_func\"]\nPASS: a module with no export list is not given one\nFAIL: move_keeps_both_export_contracts (2 of 3 scenarios)\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh move_keeps_both_export_contracts 600'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "move_destination_list_however_built_gains_the_name",
+      "weight": 0.22,
+      "value": 0.0,
+      "children": null,
+      "info": {
+        "exit_code": 1,
+        "stdout": "FAIL: the list the destination publishes is built through a variable\n  pkg/a.py publishes ['a_func'], not []\n      expected pkg/a.py:\n        | __all__ == []\n      got:\n        | __all__ = [\"a_func\"]\n  pkg/b.py publishes ['other'], not ['a_func', 'other']\n      expected pkg/b.py:\n        | __all__ == ['a_func', 'other']\n      got:\n        | from pkg.a import __all__\n        | \n        | \n        | def a_func():\n        |     pass\n        | def other():\n        |     pass\n        | \n        | \n        | _names = [\"other\"]\n        | __all__ = _names\nFAIL: the list the destination publishes is chosen by a condition\n  pkg/a.py publishes ['a_func'], not []\n      expected pkg/a.py:\n        | __all__ == []\n      got:\n        | __all__ = [\"a_func\"]\n  pkg/b.py publishes ['other'], not ['a_func', 'other']\n      expected pkg/b.py:\n        | __all__ == ['a_func', 'other']\n      got:\n        | from pkg.a import __all__\n        | \n        | \n        | def a_func():\n        |     pass\n        | def other():\n        |     pass\n        | \n        | \n        | FLAG = True\n        | __all__ = [\"other\"] if FLAG else []\nFAIL: the list the destination publishes is a sorted copy\n  pkg/a.py publishes ['a_func'], not []\n      expected pkg/a.py:\n        | __all__ == []\n      got:\n        | __all__ = [\"a_func\"]\n  pkg/b.py publishes ['other'], not ['a_func', 'other']\n      expected pkg/b.py:\n        | __all__ == ['a_func', 'other']\n      got:\n        | from pkg.a import __all__\n        | \n        | \n        | def a_func():\n        |     pass\n        | def other():\n        |     pass\n        | \n        | \n        | __all__ = sorted([\"other\"])\nFAIL: the list the destination publishes is one string of names\n  pkg/a.py publishes ['a_func'], not []\n      expected pkg/a.py:\n        | __all__ == []\n      got:\n        | __all__ = [\"a_func\"]\n  pkg/b.py publishes ['other'], not ['a_func', 'other']\n      expected pkg/b.py:\n        | __all__ == ['a_func', 'other']\n      got:\n        | from pkg.a import __all__\n        | \n        | \n        | def a_func():\n        |     pass\n        | def other():\n        |     pass\n        | \n        | \n        | __all__ = \"other\".split()\nFAIL: the generated import does not carry the export list across\n  pkg/b.py still contains '__all__'\n      expected pkg/b.py:\n        | no __all__\n      got:\n        | from pkg.a import other, __all__\n        | \n        | \n        | def a_func():\n        |     pass\nFAIL: move_destination_list_however_built_gains_the_name (5 of 5 scenarios)\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh move_destination_list_however_built_gains_the_name 600'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "move_source_list_however_built_loses_the_name",
+      "weight": 0.21,
+      "value": 0.0,
+      "children": null,
+      "info": {
+        "exit_code": 1,
+        "stdout": "FAIL: a list written as one string of names\n  pkg/a.py publishes ['a_func', 'other'], not ['other']\n      expected pkg/a.py:\n        | __all__ == ['other']\n      got:\n        | def other():\n        |     pass\n        | \n        | \n        | __all__ = \"a_func other\".split()\nFAIL: a name put on the list at its front\n  pkg/a.py publishes ['a_func', 'other'], not ['other']\n      expected pkg/a.py:\n        | __all__ == ['other']\n      got:\n        | def other():\n        |     pass\n        | \n        | \n        | __all__ = [\"other\"]\n        | __all__.insert(0, \"a_func\")\nFAIL: a name appended to the list\n  pkg/a.py publishes ['a_func', 'other'], not ['other']\n      expected pkg/a.py:\n        | __all__ == ['other']\n      got:\n        | def other():\n        |     pass\n        | \n        | \n        | __all__ = [\"other\"]\n        | __all__.append(\"a_func\")\nFAIL: a tuple of two names\n  pkg/a.py publishes ['a_func', 'other'], not ['other']\n      expected pkg/a.py:\n        | __all__ == ['other']\n      got:\n        | def other():\n        |     pass\n        | \n        | \n        | __all__ = (\"a_func\", \"other\")\nFAIL: a tuple of names held in a variable\n  pkg/a.py publishes ['a_func', 'other'], not ['other']\n      expected pkg/a.py:\n        | __all__ == ['other']\n      got:\n        | def other():\n        |     pass\n        | \n        | \n        | _names = (\"a_func\", \"other\")\n        | __all__ = _names\nFAIL: move_source_list_however_built_loses_the_name (5 of 5 scenarios)\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh move_source_list_however_built_loses_the_name 600'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "gate_repo_suite_green_and_tests_untouched",
+      "weight": 0.03,
+      "value": 0.0,
+      "children": null,
+      "info": {
+        "exit_code": 1,
+        "stdout": "RULE HOLDS: every shipped test file is byte for byte what it was\nRULE HOLDS: the repository's own tests pass\nFAIL: not paid -- no scored behaviour was fixed, and this gate does not pay for an empty change\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=0 bash /hud/grader/run_grading.sh gate_repo_suite_green_and_tests_untouched 900'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 900
+        }
+      }
+    }
+  ],
+  "status": "graded",
+  "patch_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
+  "result": {
+    "reward": 0.0,
+    "done": true,
+    "content": "Blocker criteria failed (gate_repo_suite_green_and_tests_untouched, move_keeps_both_export_contracts); cap 0.60, reward 0.000 (uncapped 0.000).",
+    "info": {
+      "uncapped_reward": 0.0,
+      "failed_blockers": [
+        "gate_repo_suite_green_and_tests_untouched",
+        "move_keeps_both_export_contracts"
+      ],
+      "blocker_cap": 0.6
+    },
+    "isError": false,
+    "subscores": [
+      {
+        "name": "export_list_follows_the_rename",
+        "weight": 0.08,
+        "value": 0.0,
+        "children": null,
+        "info": {
+          "exit_code": 1,
+          "stdout": "FAIL: the defining module's export list\n  mod.py is not what it has to be\n      expected mod.py:\n        | def new_func():\n        |     pass\n        | \n        | \n        | __all__ = [\"new_func\"]\n      got:\n        | def new_func():\n        |     pass\n        | \n        | \n        | __all__ = [\"a_func\"]\nFAIL: an export list written as a tuple over several lines\n  mod.py is not what it has to be\n      expected mod.py:\n        | def new_func():\n        |     pass\n        | \n        | \n        | def other():\n        |     pass\n        | \n        | \n        | __all__ = (\n        |     'new_func',\n        |     'other',\n        | )\n      got:\n        | def new_func():\n        |     pass\n        | \n        | \n        | def other():\n        |     pass\n        | \n        | \n        | __all__ = (\n        |     'a_func',\n        |     'other',\n        | )\nFAIL: the export list of a package that re-exports the name\n  pkg/__init__.py is not what it has to be\n      expected pkg/__init__.py:\n        | from pkg.core import new_func\n        | \n        | __all__ = [\"new_func\"]\n      got:\n        | from pkg.core import new_func\n        | \n        | __all__ = [\"a_func\"]\nFAIL: an export list naming a module that is being renamed\n  pkg/__init__.py is not what it has to be\n      expected pkg/__init__.py:\n        | from pkg import base\n        | \n        | __all__ = [\"base\"]\n      got:\n        | from pkg import base\n        | \n        | __all__ = [\"core\"]\nFAIL: an export list extended in place\n  mod.py is not what it has to be\n      expected mod.py:\n        | def new_func():\n        |     pass\n        | \n        | \n        | __all__ = []\n        | __all__ += [\"new_func\"]\n      got:\n        | def new_func():\n        |     pass\n        | \n        | \n        | __all__ = []\n        | __all__ += [\"a_func\"]\nFAIL: prose and an unrelated list of the same word\n  mod.py is not what it has to be\n      expected mod.py:\n        | def new_func():\n        |     \"\"\"a_func is the entry point.\n        | \n        |     Call a_func early.\n        |     \"\"\"\n        |     # a_func is also mentioned here\n        |     pass\n        | \n        | \n        | NAMES = [\"a_func\"]\n        | __all__ = [\"new_func\"]\n      got:\n        | def new_func():\n        |     \"\"\"a_func is the entry point.\n        | \n        |     Call a_func early.\n        |     \"\"\"\n        |     # a_func is also mentioned here\n        |     pass\n        | \n        | \n        | NAMES = [\"a_func\"]\n        | __all__ = [\"a_func\"]\nFAIL: another module that exports a name of its own\n  first.py is not what it has to be\n      expected first.py:\n        | def new_func():\n        |     pass\n        | \n        | \n        | __all__ = [\"new_func\"]\n      got:\n        | def new_func():\n        |     pass\n        | \n        | \n        | __all__ = [\"a_func\"]\nFAIL: an export list built through a variable\n  mod.py is not what it has to be\n      expected mod.py:\n        | def new_func():\n        |     pass\n        | \n        | \n        | exported = \"new_func\"\n        | label = \"a_func\"\n        | __all__ = [exported]\n      got:\n        | def new_func():\n        |     pass\n        | \n        | \n        | exported = \"a_func\"\n        | label = \"a_func\"\n        | __all__ = [exported]\nFAIL: an export list added to another list\n  mod.py is not what it has to be\n      expected mod.py:\n        | def new_func():\n        |     pass\n        | \n        | \n        | def other():\n        |     pass\n        | \n        | \n        | extra = \"new_func\"\n        | __all__ = [\"other\"] + [extra]\n      got:\n        | def new_func():\n        |     pass\n        | \n        | \n        | def other():\n        |     pass\n        | \n        | \n        | extra = \"a_func\"\n        | __all__ = [\"other\"] + [extra]\nFAIL: an export list written as one string of names\n  mod.py publishes ['a_func', 'other'], not ['new_func', 'other']\n      expected mod.py:\n        | __all__ == ['new_func', 'other']\n      got:\n        | def new_func():\n        |     pass\n        | \n        | \n        | def other():\n        |     pass\n        | \n        | \n        | WORDS = \"a_func other\".split()\n        | __all__ = \"a_func other\".split()\nFAIL: an export list written as a block of names\n  mod.py publishes ['a_func', 'other'], not ['new_func', 'other']\n      expected mod.py:\n        | __all__ == ['new_func', 'other']\n      got:\n        | def new_func():\n        |     pass\n        | \n        | \n        | def other():\n        |     pass\n        | \n        | \n        | __all__ = \"\"\"\n        |     a_func\n        |     other\n        | \"\"\".split()\nFAIL: export_list_follows_the_rename (11 of 11 scenarios)\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh export_list_follows_the_rename 600'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 600
+          }
+        }
+      },
+      {
+        "name": "type_constructor_names_follow_the_rename",
+        "weight": 0.08,
+        "value": 0.0,
+        "children": null,
+        "info": {
+          "exit_code": 1,
+          "stdout": "FAIL: the name a new type is given\n  mod.py is not what it has to be\n      expected mod.py:\n        | import typing\n        | \n        | LABEL = \"Number\"\n        | Count = typing.NewType(\"Count\", int)\n        | \n        | \n        | def scale(value: Count) -> Count:\n        |     \"\"\"Number is a count of things.\"\"\"\n        |     return value\n      got:\n        | import typing\n        | \n        | LABEL = \"Number\"\n        | Count = typing.NewType(\"Number\", int)\n        | \n        | \n        | def scale(value: Count) -> Count:\n        |     \"\"\"Number is a count of things.\"\"\"\n        |     return value\nFAIL: a new type built from a directly imported constructor\n  mod.py is not what it has to be\n      expected mod.py:\n        | from typing import NewType\n        | \n        | AccountId = NewType(\"AccountId\", int)\n      got:\n        | from typing import NewType\n        | \n        | AccountId = NewType(\"UserId\", int)\nFAIL: the bound of a type variable\n  mod.py is not what it has to be\n      expected mod.py:\n        | from typing import TypeVar\n        | \n        | \n        | class Figure:\n        |     pass\n        | \n        | \n        | T = TypeVar(\"T\", bound=\"Figure\")\n      got:\n        | from typing import TypeVar\n        | \n        | \n        | class Figure:\n        |     pass\n        | \n        | \n        | T = TypeVar(\"T\", bound=\"Shape\")\nPASS: a type whose name is spelled like something else\nFAIL: a constraint of a type variable\n  mod.py is not what it has to be\n      expected mod.py:\n        | from typing import TypeVar\n        | \n        | \n        | class Figure:\n        |     pass\n        | \n        | \n        | class Circle(Figure):\n        |     pass\n        | \n        | \n        | T = TypeVar(\"T\", \"Circle\", \"Figure\")\n      got:\n        | from typing import TypeVar\n        | \n        | \n        | class Figure:\n        |     pass\n        | \n        | \n        | class Circle(Figure):\n        |     pass\n        | \n        | \n        | T = TypeVar(\"T\", \"Circle\", \"Shape\")\nFAIL: a new type whose name is handed over by keyword\n  mod.py is not what it has to be\n      expected mod.py:\n        | from typing import NewType\n        | \n        | Count = NewType(name=\"Count\", tp=int)\n      got:\n        | from typing import NewType\n        | \n        | Count = NewType(name=\"Number\", tp=int)\nFAIL: type_constructor_names_follow_the_rename (5 of 6 scenarios)\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh type_constructor_names_follow_the_rename 600'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 600
+          }
+        }
+      },
+      {
+        "name": "declared_attribute_names_follow_the_rename",
+        "weight": 0.08,
+        "value": 0.0,
+        "children": null,
+        "info": {
+          "exit_code": 1,
+          "stdout": "FAIL: a slots tuple follows the attribute\n  mod.py is not what it has to be\n      expected mod.py:\n        | class Point:\n        |     __slots__ = (\"sideways\",)\n        | \n        |     def __init__(self):\n        |         self.sideways = 1\n        | \n        |     def get(self):\n        |         return self.sideways\n        | \n        | \n        | class Other:\n        |     __slots__ = (\"across\",)\n        | \n        |     def __init__(self):\n        |         self.across = 2\n      got:\n        | class Point:\n        |     __slots__ = (\"across\",)\n        | \n        |     def __init__(self):\n        |         self.sideways = 1\n        | \n        |     def get(self):\n        |         return self.sideways\n        | \n        | \n        | class Other:\n        |     __slots__ = (\"across\",)\n        | \n        |     def __init__(self):\n        |         self.across = 2\nFAIL: a slots list follows the attribute\n  mod.py is not what it has to be\n      expected mod.py:\n        | class Point:\n        |     __slots__ = [\"sideways\", \"down\"]\n        | \n        |     def __init__(self):\n        |         self.sideways = 1\n        |         self.down = 2\n      got:\n        | class Point:\n        |     __slots__ = [\"across\", \"down\"]\n        | \n        |     def __init__(self):\n        |         self.sideways = 1\n        |         self.down = 2\nFAIL: the positional attribute names a class declares for matching\n  mod.py is not what it has to be\n      expected mod.py:\n        | class Point:\n        |     __slots__ = (\"sideways\", \"down\")\n        |     __match_args__ = (\"sideways\", \"down\")\n        | \n        |     def __init__(self):\n        |         self.sideways = 1\n        |         self.down = 2\n        | \n        | \n        | class Other:\n        |     __match_args__ = (\"across\",)\n        | \n        |     def __init__(self):\n        |         self.across = 3\n      got:\n        | class Point:\n        |     __slots__ = (\"across\", \"down\")\n        |     __match_args__ = (\"across\", \"down\")\n        | \n        |     def __init__(self):\n        |         self.sideways = 1\n        |         self.down = 2\n        | \n        | \n        | class Other:\n        |     __match_args__ = (\"across\",)\n        | \n        |     def __init__(self):\n        |         self.across = 3\nFAIL: declared_attribute_names_follow_the_rename (3 of 3 scenarios)\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh declared_attribute_names_follow_the_rename 600'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 600
+          }
+        }
+      },
+      {
+        "name": "string_annotations_follow_and_data_stays",
+        "weight": 0.22,
+        "value": 0.0,
+        "children": null,
+        "info": {
+          "exit_code": 1,
+          "stdout": "FAIL: a forward reference in a signature\n  mod.py is not what it has to be\n      expected mod.py:\n        | class Figure:\n        |     \"\"\"Shape is a shape.\"\"\"\n        | \n        | \n        | NAMES = [\"Shape\"]\n        | \n        | \n        | def draw(what: \"Figure\") -> \"Figure\":\n        |     return what\n      got:\n        | class Figure:\n        |     \"\"\"Shape is a shape.\"\"\"\n        | \n        | \n        | NAMES = [\"Shape\"]\n        | \n        | \n        | def draw(what: \"Shape\") -> \"Shape\":\n        |     return what\nFAIL: a forward reference inside a generic\n  mod.py is not what it has to be\n      expected mod.py:\n        | from typing import List, Optional\n        | \n        | \n        | class Figure:\n        |     pass\n        | \n        | \n        | corner: Optional[\"Figure\"] = None\n        | \n        | \n        | def each(items: List[\"Figure\"]):\n        |     return items\n      got:\n        | from typing import List, Optional\n        | \n        | \n        | class Figure:\n        |     pass\n        | \n        | \n        | corner: Optional[\"Shape\"] = None\n        | \n        | \n        | def each(items: List[\"Shape\"]):\n        |     return items\nFAIL: a literal type beside a forward reference\n  mod.py is not what it has to be\n      expected mod.py:\n        | from typing import Literal\n        | \n        | \n        | class Figure:\n        |     pass\n        | \n        | \n        | def draw(what: \"Figure\", kind: Literal[\"Shape\", \"Circle\"]) -> \"Figure\":\n        |     return what\n      got:\n        | from typing import Literal\n        | \n        | \n        | class Figure:\n        |     pass\n        | \n        | \n        | def draw(what: \"Shape\", kind: Literal[\"Shape\", \"Circle\"]) -> \"Shape\":\n        |     return what\nFAIL: the metadata of an annotated type\n  mod.py is not what it has to be\n      expected mod.py:\n        | from typing import Annotated\n        | \n        | \n        | class Figure:\n        |     pass\n        | \n        | \n        | def draw(what: Annotated[\"Figure\", \"Shape\"]):\n        |     return what\n      got:\n        | from typing import Annotated\n        | \n        | \n        | class Figure:\n        |     pass\n        | \n        | \n        | def draw(what: Annotated[\"Shape\", \"Shape\"]):\n        |     return what\nFAIL: a class body with a literal field and a forward reference\n  mod.py is not what it has to be\n      expected mod.py:\n        | from typing import Literal\n        | \n        | \n        | class Figure:\n        |     pass\n        | \n        | \n        | class Box:\n        |     kind: Literal[\"Shape\"] = \"Shape\"\n        |     item: \"Figure\"\n      got:\n        | from typing import Literal\n        | \n        | \n        | class Figure:\n        |     pass\n        | \n        | \n        | class Box:\n        |     kind: Literal[\"Shape\"] = \"Shape\"\n        |     item: \"Shape\"\nFAIL: string_annotations_follow_and_data_stays (5 of 5 scenarios)\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh string_annotations_follow_and_data_stays 600'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 600
+          }
+        }
+      },
+      {
+        "name": "move_keeps_both_export_contracts",
+        "weight": 0.08,
+        "value": 0.0,
+        "children": null,
+        "info": {
+          "exit_code": 1,
+          "stdout": "FAIL: the module left behind stops exporting the name\n  pkg/a.py publishes ['a_func'], not []\n      expected pkg/a.py:\n        | __all__ == []\n      got:\n        | __all__ = [\"a_func\"]\n  pkg/a.py still contains 'a_func'\n      expected pkg/a.py:\n        | no a_func\n      got:\n        | __all__ = [\"a_func\"]\nFAIL: the module arrived in starts exporting the name\n  pkg/b.py publishes ['other'], not ['a_func', 'other']\n      expected pkg/b.py:\n        | __all__ == ['a_func', 'other']\n      got:\n        | from pkg.a import __all__\n        | \n        | \n        | def a_func():\n        |     pass\n        | def other():\n        |     pass\n        | \n        | \n        | __all__ = [\"other\"]\n  pkg/a.py still contains 'a_func'\n      expected pkg/a.py:\n        | no a_func\n      got:\n        | __all__ = [\"a_func\"]\nPASS: a module with no export list is not given one\nFAIL: move_keeps_both_export_contracts (2 of 3 scenarios)\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh move_keeps_both_export_contracts 600'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 600
+          }
+        }
+      },
+      {
+        "name": "move_destination_list_however_built_gains_the_name",
+        "weight": 0.22,
+        "value": 0.0,
+        "children": null,
+        "info": {
+          "exit_code": 1,
+          "stdout": "FAIL: the list the destination publishes is built through a variable\n  pkg/a.py publishes ['a_func'], not []\n      expected pkg/a.py:\n        | __all__ == []\n      got:\n        | __all__ = [\"a_func\"]\n  pkg/b.py publishes ['other'], not ['a_func', 'other']\n      expected pkg/b.py:\n        | __all__ == ['a_func', 'other']\n      got:\n        | from pkg.a import __all__\n        | \n        | \n        | def a_func():\n        |     pass\n        | def other():\n        |     pass\n        | \n        | \n        | _names = [\"other\"]\n        | __all__ = _names\nFAIL: the list the destination publishes is chosen by a condition\n  pkg/a.py publishes ['a_func'], not []\n      expected pkg/a.py:\n        | __all__ == []\n      got:\n        | __all__ = [\"a_func\"]\n  pkg/b.py publishes ['other'], not ['a_func', 'other']\n      expected pkg/b.py:\n        | __all__ == ['a_func', 'other']\n      got:\n        | from pkg.a import __all__\n        | \n        | \n        | def a_func():\n        |     pass\n        | def other():\n        |     pass\n        | \n        | \n        | FLAG = True\n        | __all__ = [\"other\"] if FLAG else []\nFAIL: the list the destination publishes is a sorted copy\n  pkg/a.py publishes ['a_func'], not []\n      expected pkg/a.py:\n        | __all__ == []\n      got:\n        | __all__ = [\"a_func\"]\n  pkg/b.py publishes ['other'], not ['a_func', 'other']\n      expected pkg/b.py:\n        | __all__ == ['a_func', 'other']\n      got:\n        | from pkg.a import __all__\n        | \n        | \n        | def a_func():\n        |     pass\n        | def other():\n        |     pass\n        | \n        | \n        | __all__ = sorted([\"other\"])\nFAIL: the list the destination publishes is one string of names\n  pkg/a.py publishes ['a_func'], not []\n      expected pkg/a.py:\n        | __all__ == []\n      got:\n        | __all__ = [\"a_func\"]\n  pkg/b.py publishes ['other'], not ['a_func', 'other']\n      expected pkg/b.py:\n        | __all__ == ['a_func', 'other']\n      got:\n        | from pkg.a import __all__\n        | \n        | \n        | def a_func():\n        |     pass\n        | def other():\n        |     pass\n        | \n        | \n        | __all__ = \"other\".split()\nFAIL: the generated import does not carry the export list across\n  pkg/b.py still contains '__all__'\n      expected pkg/b.py:\n        | no __all__\n      got:\n        | from pkg.a import other, __all__\n        | \n        | \n        | def a_func():\n        |     pass\nFAIL: move_destination_list_however_built_gains_the_name (5 of 5 scenarios)\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh move_destination_list_however_built_gains_the_name 600'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 600
+          }
+        }
+      },
+      {
+        "name": "move_source_list_however_built_loses_the_name",
+        "weight": 0.21,
+        "value": 0.0,
+        "children": null,
+        "info": {
+          "exit_code": 1,
+          "stdout": "FAIL: a list written as one string of names\n  pkg/a.py publishes ['a_func', 'other'], not ['other']\n      expected pkg/a.py:\n        | __all__ == ['other']\n      got:\n        | def other():\n        |     pass\n        | \n        | \n        | __all__ = \"a_func other\".split()\nFAIL: a name put on the list at its front\n  pkg/a.py publishes ['a_func', 'other'], not ['other']\n      expected pkg/a.py:\n        | __all__ == ['other']\n      got:\n        | def other():\n        |     pass\n        | \n        | \n        | __all__ = [\"other\"]\n        | __all__.insert(0, \"a_func\")\nFAIL: a name appended to the list\n  pkg/a.py publishes ['a_func', 'other'], not ['other']\n      expected pkg/a.py:\n        | __all__ == ['other']\n      got:\n        | def other():\n        |     pass\n        | \n        | \n        | __all__ = [\"other\"]\n        | __all__.append(\"a_func\")\nFAIL: a tuple of two names\n  pkg/a.py publishes ['a_func', 'other'], not ['other']\n      expected pkg/a.py:\n        | __all__ == ['other']\n      got:\n        | def other():\n        |     pass\n        | \n        | \n        | __all__ = (\"a_func\", \"other\")\nFAIL: a tuple of names held in a variable\n  pkg/a.py publishes ['a_func', 'other'], not ['other']\n      expected pkg/a.py:\n        | __all__ == ['other']\n      got:\n        | def other():\n        |     pass\n        | \n        | \n        | _names = (\"a_func\", \"other\")\n        | __all__ = _names\nFAIL: move_source_list_however_built_loses_the_name (5 of 5 scenarios)\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh move_source_list_however_built_loses_the_name 600'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 600
+          }
+        }
+      },
+      {
+        "name": "gate_repo_suite_green_and_tests_untouched",
+        "weight": 0.03,
+        "value": 0.0,
+        "children": null,
+        "info": {
+          "exit_code": 1,
+          "stdout": "RULE HOLDS: every shipped test file is byte for byte what it was\nRULE HOLDS: the repository's own tests pass\nFAIL: not paid -- no scored behaviour was fixed, and this gate does not pay for an empty change\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=0 bash /hud/grader/run_grading.sh gate_repo_suite_green_and_tests_untouched 900'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 900
+          }
+        }
+      }
+    ]
+  }
+}

harbor/evidence/rope/reports/reference.json

@@ -0,0 +1,274 @@
+{
+  "revision": "coding-rubric-v2",
+  "task_id": "rope_rename_all_reexport_owners_contract",
+  "grader_version": "0.6.15",
+  "commands": [
+    {
+      "name": "export_list_follows_the_rename",
+      "weight": 0.08,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "PASS: the defining module's export list\nPASS: an export list written as a tuple over several lines\nPASS: the export list of a package that re-exports the name\nPASS: an export list naming a module that is being renamed\nPASS: an export list extended in place\nPASS: prose and an unrelated list of the same word\nPASS: another module that exports a name of its own\nPASS: an export list built through a variable\nPASS: an export list added to another list\nPASS: an export list written as one string of names\nPASS: an export list written as a block of names\nPASS: export_list_follows_the_rename (11 scenarios)\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh export_list_follows_the_rename 600'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "type_constructor_names_follow_the_rename",
+      "weight": 0.08,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "PASS: the name a new type is given\nPASS: a new type built from a directly imported constructor\nPASS: the bound of a type variable\nPASS: a type whose name is spelled like something else\nPASS: a constraint of a type variable\nPASS: a new type whose name is handed over by keyword\nPASS: type_constructor_names_follow_the_rename (6 scenarios)\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh type_constructor_names_follow_the_rename 600'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "declared_attribute_names_follow_the_rename",
+      "weight": 0.08,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "PASS: a slots tuple follows the attribute\nPASS: a slots list follows the attribute\nPASS: the positional attribute names a class declares for matching\nPASS: declared_attribute_names_follow_the_rename (3 scenarios)\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh declared_attribute_names_follow_the_rename 600'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "string_annotations_follow_and_data_stays",
+      "weight": 0.22,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "PASS: a forward reference in a signature\nPASS: a forward reference inside a generic\nPASS: a literal type beside a forward reference\nPASS: the metadata of an annotated type\nPASS: a class body with a literal field and a forward reference\nPASS: string_annotations_follow_and_data_stays (5 scenarios)\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh string_annotations_follow_and_data_stays 600'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "move_keeps_both_export_contracts",
+      "weight": 0.08,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "PASS: the module left behind stops exporting the name\nPASS: the module arrived in starts exporting the name\nPASS: a module with no export list is not given one\nPASS: move_keeps_both_export_contracts (3 scenarios)\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh move_keeps_both_export_contracts 600'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "move_destination_list_however_built_gains_the_name",
+      "weight": 0.22,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "PASS: the list the destination publishes is built through a variable\nPASS: the list the destination publishes is chosen by a condition\nPASS: the list the destination publishes is a sorted copy\nPASS: the list the destination publishes is one string of names\nPASS: the generated import does not carry the export list across\nPASS: move_destination_list_however_built_gains_the_name (5 scenarios)\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh move_destination_list_however_built_gains_the_name 600'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "move_source_list_however_built_loses_the_name",
+      "weight": 0.21,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "PASS: a list written as one string of names\nPASS: a name put on the list at its front\nPASS: a name appended to the list\nPASS: a tuple of two names\nPASS: a tuple of names held in a variable\nPASS: move_source_list_however_built_loses_the_name (5 scenarios)\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh move_source_list_however_built_loses_the_name 600'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "gate_repo_suite_green_and_tests_untouched",
+      "weight": 0.03,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "PASS: every shipped test file is byte for byte what it was\nPASS: the repository's own tests pass\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=1 bash /hud/grader/run_grading.sh gate_repo_suite_green_and_tests_untouched 900'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 900
+        }
+      }
+    }
+  ],
+  "status": "graded",
+  "patch_sha256": "0d2bb7844e4fbbf9603570156b1857e77c12abca0eaf486a93f5e4118f01b9dc",
+  "result": {
+    "reward": 1.0,
+    "done": true,
+    "content": null,
+    "info": {},
+    "isError": false,
+    "subscores": [
+      {
+        "name": "export_list_follows_the_rename",
+        "weight": 0.08,
+        "value": 1.0,
+        "children": null,
+        "info": {
+          "exit_code": 0,
+          "stdout": "PASS: the defining module's export list\nPASS: an export list written as a tuple over several lines\nPASS: the export list of a package that re-exports the name\nPASS: an export list naming a module that is being renamed\nPASS: an export list extended in place\nPASS: prose and an unrelated list of the same word\nPASS: another module that exports a name of its own\nPASS: an export list built through a variable\nPASS: an export list added to another list\nPASS: an export list written as one string of names\nPASS: an export list written as a block of names\nPASS: export_list_follows_the_rename (11 scenarios)\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh export_list_follows_the_rename 600'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 600
+          }
+        }
+      },
+      {
+        "name": "type_constructor_names_follow_the_rename",
+        "weight": 0.08,
+        "value": 1.0,
+        "children": null,
+        "info": {
+          "exit_code": 0,
+          "stdout": "PASS: the name a new type is given\nPASS: a new type built from a directly imported constructor\nPASS: the bound of a type variable\nPASS: a type whose name is spelled like something else\nPASS: a constraint of a type variable\nPASS: a new type whose name is handed over by keyword\nPASS: type_constructor_names_follow_the_rename (6 scenarios)\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh type_constructor_names_follow_the_rename 600'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 600
+          }
+        }
+      },
+      {
+        "name": "declared_attribute_names_follow_the_rename",
+        "weight": 0.08,
+        "value": 1.0,
+        "children": null,
+        "info": {
+          "exit_code": 0,
+          "stdout": "PASS: a slots tuple follows the attribute\nPASS: a slots list follows the attribute\nPASS: the positional attribute names a class declares for matching\nPASS: declared_attribute_names_follow_the_rename (3 scenarios)\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh declared_attribute_names_follow_the_rename 600'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 600
+          }
+        }
+      },
+      {
+        "name": "string_annotations_follow_and_data_stays",
+        "weight": 0.22,
+        "value": 1.0,
+        "children": null,
+        "info": {
+          "exit_code": 0,
+          "stdout": "PASS: a forward reference in a signature\nPASS: a forward reference inside a generic\nPASS: a literal type beside a forward reference\nPASS: the metadata of an annotated type\nPASS: a class body with a literal field and a forward reference\nPASS: string_annotations_follow_and_data_stays (5 scenarios)\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh string_annotations_follow_and_data_stays 600'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 600
+          }
+        }
+      },
+      {
+        "name": "move_keeps_both_export_contracts",
+        "weight": 0.08,
+        "value": 1.0,
+        "children": null,
+        "info": {
+          "exit_code": 0,
+          "stdout": "PASS: the module left behind stops exporting the name\nPASS: the module arrived in starts exporting the name\nPASS: a module with no export list is not given one\nPASS: move_keeps_both_export_contracts (3 scenarios)\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh move_keeps_both_export_contracts 600'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 600
+          }
+        }
+      },
+      {
+        "name": "move_destination_list_however_built_gains_the_name",
+        "weight": 0.22,
+        "value": 1.0,
+        "children": null,
+        "info": {
+          "exit_code": 0,
+          "stdout": "PASS: the list the destination publishes is built through a variable\nPASS: the list the destination publishes is chosen by a condition\nPASS: the list the destination publishes is a sorted copy\nPASS: the list the destination publishes is one string of names\nPASS: the generated import does not carry the export list across\nPASS: move_destination_list_however_built_gains_the_name (5 scenarios)\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh move_destination_list_however_built_gains_the_name 600'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 600
+          }
+        }
+      },
+      {
+        "name": "move_source_list_however_built_loses_the_name",
+        "weight": 0.21,
+        "value": 1.0,
+        "children": null,
+        "info": {
+          "exit_code": 0,
+          "stdout": "PASS: a list written as one string of names\nPASS: a name put on the list at its front\nPASS: a name appended to the list\nPASS: a tuple of two names\nPASS: a tuple of names held in a variable\nPASS: move_source_list_however_built_loses_the_name (5 scenarios)\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'bash /hud/grader/run_grading.sh move_source_list_however_built_loses_the_name 600'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 600
+          }
+        }
+      },
+      {
+        "name": "gate_repo_suite_green_and_tests_untouched",
+        "weight": 0.03,
+        "value": 1.0,
+        "children": null,
+        "info": {
+          "exit_code": 0,
+          "stdout": "PASS: every shipped test file is byte for byte what it was\nPASS: the repository's own tests pass\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc 'CORE_GREEN=1 bash /hud/grader/run_grading.sh gate_repo_suite_green_and_tests_untouched 900'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 900
+          }
+        }
+      }
+    ]
+  }
+}

harbor/evidence/rope/summary.json

@@ -0,0 +1,149 @@
+{
+  "name": "rope",
+  "measure": {
+    "ok": true,
+    "hud": "0.6.15",
+    "path": "/opt/agentvenv/bin:/usr/local/bin:/usr/bin:/bin:/usr/local/games:/usr/games",
+    "handed": [
+      "PIP_DISABLE_PIP_VERSION_CHECK",
+      "PYTHONDONTWRITEBYTECODE"
+    ],
+    "repo_dir": "/workspace/repo",
+    "python_usr_local": true,
+    "seconds": 4
+  },
+  "parity": {
+    "ok": true,
+    "reward": 1.0,
+    "criteria": 8,
+    "unmet": [],
+    "info": null,
+    "seconds": 30
+  },
+  "task_id": "rope_rename_all_reexport_owners_contract",
+  "image": "ec-swe-rope:source-v3",
+  "build": {
+    "ok": true,
+    "image_id": "sha256:e33bc697f20f1281afeb4f9862caa9dc697b1c045eb68e0d5425565a1d7196e5",
+    "ladder": {
+      "gold": 1.0,
+      "base": 0.0,
+      "noop": 0.0,
+      "lazy": 0.0,
+      "alt": 0.84,
+      "prevgold": 0.19,
+      "cheat": 0.08,
+      "tamper": 0.0,
+      "tamper_key": 0.0,
+      "tamper_gate": 0.16,
+      "tamper_report": 0.16,
+      "tamper_wire": 0.16
+    },
+    "source_digest": "d7a94af16d5e99d1",
+    "source": "edited",
+    "seconds": 362
+  },
+  "prepare": {
+    "ok": true,
+    "bundle": "rope-r5",
+    "converter": "b8bba68767da888f",
+    "pinned_files": 25,
+    "criteria": 8,
+    "seconds": 1
+  },
+  "qualified": false,
+  "controls": {
+    "ok": false,
+    "checks": {
+      "noop_is_zero": true,
+      "reference_is_one": true,
+      "reference_equals_native": true,
+      "agent_is_isolated": true
+    },
+    "results": {
+      "noop": {
+        "run": "rope-r5-control-noop-1791900129",
+        "reward": 0.0,
+        "stop": "completed",
+        "error": null,
+        "seconds": 130,
+        "probe": false,
+        "status": "graded",
+        "verifier_error": null,
+        "exit_codes": [
+          1
+        ],
+        "attempts": 1
+      },
+      "reference": {
+        "run": "rope-r5-control-reference-1791900260",
+        "reward": 1.0,
+        "stop": "completed",
+        "error": null,
+        "seconds": 98,
+        "probe": true,
+        "status": "graded",
+        "verifier_error": null,
+        "exit_codes": [
+          0
+        ],
+        "subscores": {
+          "export_list_follows_the_rename": [
+            1.0,
+            0.08
+          ],
+          "type_constructor_names_follow_the_rename": [
+            1.0,
+            0.08
+          ],
+          "declared_attribute_names_follow_the_rename": [
+            1.0,
+            0.08
+          ],
+          "string_annotations_follow_and_data_stays": [
+            1.0,
+            0.22
+          ],
+          "move_keeps_both_export_contracts": [
+            1.0,
+            0.08
+          ],
+          "move_destination_list_however_built_gains_the_name": [
+            1.0,
+            0.22
+          ],
+          "move_source_list_however_built_loses_the_name": [
+            1.0,
+            0.21
+          ],
+          "gate_repo_suite_green_and_tests_untouched": [
+            1.0,
+            0.03
+          ]
+        },
+        "attempts": 1
+      }
+    },
+    "bundle": "rope-r5",
+    "complete": false,
+    "seconds": 228
+  },
+  "leak": {
+    "scan_finished": true,
+    "hits_outside_checkout": [],
+    "other_files_holding_only_upstream_lines": 0,
+    "library_files_sharing_a_line": 0,
+    "most_lines_shared_by_one_library_file": 0,
+    "private_paths_absent": true,
+    "public_lines_shared_with_patch_and_graders": [
+      3,
+      0
+    ],
+    "private_lines_only_in_agent_checkout": [
+      0,
+      0
+    ],
+    "ok": true,
+    "seconds": 18
+  }
+}

harbor/evidence/zod/agent-image-scan.txt

@@ -0,0 +1,36 @@
+needles: 1582
+2 matching line(s): /workspace/repo/packages/zod/src/v4/core/tests/compile.test.ts
+1 matching line(s): /workspace/repo/packages/zod/src/v4/core/json-schema-processors.ts
+2 matching line(s): /workspace/repo/packages/zod/src/v4/classic/tests/record.test.ts
+234 matching line(s): /workspace/repo/packages/zod/src/v4/classic/tests/from-json-schema.test.ts
+9 matching line(s): /workspace/repo/packages/zod/src/v4/classic/tests/to-json-schema-methods.test.ts
+1 matching line(s): /workspace/repo/packages/zod/src/v4/classic/tests/catch.test.ts
+571 matching line(s): /workspace/repo/packages/zod/src/v4/classic/tests/to-json-schema.test.ts
+2 matching line(s): /workspace/repo/packages/zod/src/v4/classic/tests/datetime.test.ts
+3 matching line(s): /workspace/repo/packages/docs/content/json-schema.mdx
+82 matching line(s): /workspace/repo/node_modules/.pnpm/zod@4.0.8/node_modules/zod/src/v4/classic/tests/to-json-schema.test.ts
+3 matching line(s): /workspace/repo/node_modules/.pnpm/@seriousme+openapi-schema-validator@2.9.0/node_modules/@seriousme/openapi-schema-validator/README.md
+2 matching line(s): /workspace/repo/node_modules/.pnpm/@seriousme+openapi-schema-validator@2.9.0/node_modules/@seriousme/openapi-schema-validator/UPGRADING.md
+1 matching line(s): /workspace/repo/node_modules/.pnpm/zod@4.4.3/node_modules/zod/src/v4/classic/tests/record.test.ts
+74 matching line(s): /workspace/repo/node_modules/.pnpm/zod@4.4.3/node_modules/zod/src/v4/classic/tests/from-json-schema.test.ts
+9 matching line(s): /workspace/repo/node_modules/.pnpm/zod@4.4.3/node_modules/zod/src/v4/classic/tests/to-json-schema-methods.test.ts
+176 matching line(s): /workspace/repo/node_modules/.pnpm/zod@4.4.3/node_modules/zod/src/v4/classic/tests/to-json-schema.test.ts
+3 matching line(s): /tmp/agent-home/.local/share/pnpm/store/v11/files/70/cb2bcdec2238b686581c1875801ebbe4655e31e481abb02808b641794cf61202fa15dc6d132007920e5514a5b30eca9631f1f497dc5bd161bb30e306db1242
+82 matching line(s): /tmp/agent-home/.local/share/pnpm/store/v11/files/8b/8b12266aa4f1770c38b9315660998b4ec629d6dc619baf4a6d546195109f73e2d04bd95ee6969b2f9b1b2265f28162d26e962ae076d3c19ebefa0116f337a7
+74 matching line(s): /tmp/agent-home/.local/share/pnpm/store/v11/files/32/6254f362a42af3d2190672411f230955afe95289cec7e3faa916c4321e1dcc9e1fc392d30878ccba466469557186aed95195422285205dfe5c9058384aebd8
+176 matching line(s): /tmp/agent-home/.local/share/pnpm/store/v11/files/0c/f56b6dbef29f709b57525b1d28ee2c4b4d0fba92e84218ff26813643513b3de6f18aaa311f93bd2b37e4a0b803d7fe2edec9ca8b8450bbd4af0c1d5ade13b7
+9 matching line(s): /tmp/agent-home/.local/share/pnpm/store/v11/files/ba/92c366de0fa10d4e1eb680244d2729bd9db73730528ac46e5eaf0edc5b586a958196da3e5fb34d860984ee358eb69f9f997fa8d57926cb81804b3ceec53f76
+1 matching line(s): /tmp/agent-home/.local/share/pnpm/store/v11/files/6b/27923e8875d5efbcf5fcc377fc75e63098b6e0d7de4a4d73391a9605a0fba28d55179223b21fa094f2cabda5a055399c8f3638268f22e8dc3f6db2685cd730
+2 matching line(s): /tmp/agent-home/.local/share/pnpm/store/v11/files/af/cef11e126bb20cd74708e91498bcd10cbe0d17c183f370ebd213a51f8cf2a016d6b14d8f86015c3b1d7286b311f95ebbe8c584ba64473485e54cd4fe2fb46f
+scan finished
+absent: /hud
+absent: /app
+absent: /tests
+present: /opt/task-tools/baseline.git
+reference-patch lines in the agent's tracked files: 1, of which the pristine checkout lacks: 0
+grader lines in the agent's tracked files: 751, of which the pristine checkout lacks: 0
+reference-patch lines in total: 43, grader lines in total: 1531
+private lines in 7 other files outside the checkout: 221, of which upstream lacks: 0
+--- private lines only the agent's checkout holds (expect none)
+--- the agent's checkout: uncommitted changes (expect none), then its history
+818c759 baseline

harbor/evidence/zod/native-reference.json

@@ -0,0 +1,204 @@
+{
+  "reward": 1.0,
+  "done": true,
+  "content": null,
+  "info": {
+    "blocker_policy": "weight",
+    "failed_blockers": []
+  },
+  "isError": false,
+  "subscores": [
+    {
+      "name": "g1_catch_value_degrades",
+      "weight": 0.04,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "VITEST passes=6 failures=0 pending=0 todo=0 tests=6 expected=6\nCRITERION g1_catch_value_degrades: PASS\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "/bin/bash /hud/grader/run_grading.sh g1_catch_value_degrades",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "g6_unencodable_default_degrades",
+      "weight": 0.04,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "VITEST passes=6 failures=0 pending=0 todo=0 tests=6 expected=6\nCRITERION g6_unencodable_default_degrades: PASS\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "/bin/bash /hud/grader/run_grading.sh g6_unencodable_default_degrades",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "g5_default_value_conforms",
+      "weight": 0.11,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "VITEST passes=6 failures=0 pending=0 todo=0 tests=6 expected=6\nCRITERION g5_default_value_conforms: PASS\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "/bin/bash /hud/grader/run_grading.sh g5_default_value_conforms",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "g2_empty_positional_list",
+      "weight": 0.09,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "VITEST passes=4 failures=0 pending=0 todo=0 tests=4 expected=4\nCRITERION g2_empty_positional_list: PASS\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "/bin/bash /hud/grader/run_grading.sh g2_empty_positional_list",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "g2_empty_tuple_rest_element",
+      "weight": 0.05,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "VITEST passes=2 failures=0 pending=0 todo=0 tests=2 expected=2\nCRITERION g2_empty_tuple_rest_element: PASS\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "/bin/bash /hud/grader/run_grading.sh g2_empty_tuple_rest_element",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "g3_pattern_keyed_record_openapi",
+      "weight": 0.11,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "VITEST passes=6 failures=0 pending=0 todo=0 tests=6 expected=6\nCRITERION g3_pattern_keyed_record_openapi: PASS\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "/bin/bash /hud/grader/run_grading.sh g3_pattern_keyed_record_openapi",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "g4_metadata_examples_openapi",
+      "weight": 0.09,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "VITEST passes=6 failures=0 pending=0 todo=0 tests=6 expected=6\nCRITERION g4_metadata_examples_openapi: PASS\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "/bin/bash /hud/grader/run_grading.sh g4_metadata_examples_openapi",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "g7_null_literal_openapi",
+      "weight": 0.09,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "VITEST passes=6 failures=0 pending=0 todo=0 tests=6 expected=6\nCRITERION g7_null_literal_openapi: PASS\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "/bin/bash /hud/grader/run_grading.sh g7_null_literal_openapi",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "g8_empty_union_accepts_nothing",
+      "weight": 0.09,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "VITEST passes=6 failures=0 pending=0 todo=0 tests=6 expected=6\nCRITERION g8_empty_union_accepts_nothing: PASS\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "/bin/bash /hud/grader/run_grading.sh g8_empty_union_accepts_nothing",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "g9_enum_members_unique",
+      "weight": 0.09,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "VITEST passes=6 failures=0 pending=0 todo=0 tests=6 expected=6\nCRITERION g9_enum_members_unique: PASS\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "/bin/bash /hud/grader/run_grading.sh g9_enum_members_unique",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "d1_established_behavior",
+      "weight": 0.1,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "VITEST passes=22 failures=0 pending=0 todo=0 tests=22 expected=22\nCRITERION d1_established_behavior: PASS\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "/bin/bash /hud/grader/run_grading.sh d1_established_behavior",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 900
+        }
+      }
+    },
+    {
+      "name": "preservation_converter_suites",
+      "weight": 0.10000000000000009,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "VITEST passes=734 failures=0 pending=0 todo=0 tests=734 expected=734\nCRITERION preservation_converter_suites: PASS\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "/bin/bash /hud/grader/run_grading.sh preservation_converter_suites",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 900
+        }
+      }
+    }
+  ]
+}

harbor/evidence/zod/reports/noop.json

@@ -0,0 +1,108 @@
+{
+  "revision": "coding-rubric-v2",
+  "task_id": "zod_json_schema_target_legality",
+  "grader_version": "0.6.17",
+  "commands": [],
+  "status": "empty_patch",
+  "patch_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
+  "result": {
+    "reward": 0.0,
+    "done": true,
+    "content": "Empty diff: no change to grade.",
+    "info": {
+      "failed_blockers": [
+        "g1_catch_value_degrades",
+        "d1_established_behavior",
+        "preservation_converter_suites"
+      ],
+      "blocker_policy": "weight"
+    },
+    "isError": false,
+    "subscores": [
+      {
+        "name": "g1_catch_value_degrades",
+        "weight": 0.04,
+        "value": 0.0,
+        "children": null,
+        "info": null
+      },
+      {
+        "name": "g6_unencodable_default_degrades",
+        "weight": 0.04,
+        "value": 0.0,
+        "children": null,
+        "info": null
+      },
+      {
+        "name": "g5_default_value_conforms",
+        "weight": 0.11,
+        "value": 0.0,
+        "children": null,
+        "info": null
+      },
+      {
+        "name": "g2_empty_positional_list",
+        "weight": 0.09,
+        "value": 0.0,
+        "children": null,
+        "info": null
+      },
+      {
+        "name": "g2_empty_tuple_rest_element",
+        "weight": 0.05,
+        "value": 0.0,
+        "children": null,
+        "info": null
+      },
+      {
+        "name": "g3_pattern_keyed_record_openapi",
+        "weight": 0.11,
+        "value": 0.0,
+        "children": null,
+        "info": null
+      },
+      {
+        "name": "g4_metadata_examples_openapi",
+        "weight": 0.09,
+        "value": 0.0,
+        "children": null,
+        "info": null
+      },
+      {
+        "name": "g7_null_literal_openapi",
+        "weight": 0.09,
+        "value": 0.0,
+        "children": null,
+        "info": null
+      },
+      {
+        "name": "g8_empty_union_accepts_nothing",
+        "weight": 0.09,
+        "value": 0.0,
+        "children": null,
+        "info": null
+      },
+      {
+        "name": "g9_enum_members_unique",
+        "weight": 0.09,
+        "value": 0.0,
+        "children": null,
+        "info": null
+      },
+      {
+        "name": "d1_established_behavior",
+        "weight": 0.1,
+        "value": 0.0,
+        "children": null,
+        "info": null
+      },
+      {
+        "name": "preservation_converter_suites",
+        "weight": 0.1,
+        "value": 0.0,
+        "children": null,
+        "info": null
+      }
+    ]
+  }
+}

harbor/evidence/zod/reports/reference.json

@@ -0,0 +1,405 @@
+{
+  "revision": "coding-rubric-v2",
+  "task_id": "zod_json_schema_target_legality",
+  "grader_version": "0.6.17",
+  "commands": [
+    {
+      "name": "g1_catch_value_degrades",
+      "weight": 0.04,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "VITEST passes=6 failures=0 pending=0 todo=0 tests=6 expected=6\nCRITERION g1_catch_value_degrades: PASS\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc '/bin/bash /hud/grader/run_grading.sh g1_catch_value_degrades'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "g6_unencodable_default_degrades",
+      "weight": 0.04,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "VITEST passes=6 failures=0 pending=0 todo=0 tests=6 expected=6\nCRITERION g6_unencodable_default_degrades: PASS\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc '/bin/bash /hud/grader/run_grading.sh g6_unencodable_default_degrades'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "g5_default_value_conforms",
+      "weight": 0.11,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "VITEST passes=6 failures=0 pending=0 todo=0 tests=6 expected=6\nCRITERION g5_default_value_conforms: PASS\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc '/bin/bash /hud/grader/run_grading.sh g5_default_value_conforms'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "g2_empty_positional_list",
+      "weight": 0.09,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "VITEST passes=4 failures=0 pending=0 todo=0 tests=4 expected=4\nCRITERION g2_empty_positional_list: PASS\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc '/bin/bash /hud/grader/run_grading.sh g2_empty_positional_list'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "g2_empty_tuple_rest_element",
+      "weight": 0.05,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "VITEST passes=2 failures=0 pending=0 todo=0 tests=2 expected=2\nCRITERION g2_empty_tuple_rest_element: PASS\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc '/bin/bash /hud/grader/run_grading.sh g2_empty_tuple_rest_element'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "g3_pattern_keyed_record_openapi",
+      "weight": 0.11,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "VITEST passes=6 failures=0 pending=0 todo=0 tests=6 expected=6\nCRITERION g3_pattern_keyed_record_openapi: PASS\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc '/bin/bash /hud/grader/run_grading.sh g3_pattern_keyed_record_openapi'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "g4_metadata_examples_openapi",
+      "weight": 0.09,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "VITEST passes=6 failures=0 pending=0 todo=0 tests=6 expected=6\nCRITERION g4_metadata_examples_openapi: PASS\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc '/bin/bash /hud/grader/run_grading.sh g4_metadata_examples_openapi'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "g7_null_literal_openapi",
+      "weight": 0.09,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "VITEST passes=6 failures=0 pending=0 todo=0 tests=6 expected=6\nCRITERION g7_null_literal_openapi: PASS\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc '/bin/bash /hud/grader/run_grading.sh g7_null_literal_openapi'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "g8_empty_union_accepts_nothing",
+      "weight": 0.09,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "VITEST passes=6 failures=0 pending=0 todo=0 tests=6 expected=6\nCRITERION g8_empty_union_accepts_nothing: PASS\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc '/bin/bash /hud/grader/run_grading.sh g8_empty_union_accepts_nothing'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "g9_enum_members_unique",
+      "weight": 0.09,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "VITEST passes=6 failures=0 pending=0 todo=0 tests=6 expected=6\nCRITERION g9_enum_members_unique: PASS\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc '/bin/bash /hud/grader/run_grading.sh g9_enum_members_unique'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 600
+        }
+      }
+    },
+    {
+      "name": "d1_established_behavior",
+      "weight": 0.1,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "VITEST passes=22 failures=0 pending=0 todo=0 tests=22 expected=22\nCRITERION d1_established_behavior: PASS\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc '/bin/bash /hud/grader/run_grading.sh d1_established_behavior'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 900
+        }
+      }
+    },
+    {
+      "name": "preservation_converter_suites",
+      "weight": 0.10000000000000009,
+      "value": 1.0,
+      "children": null,
+      "info": {
+        "exit_code": 0,
+        "stdout": "VITEST passes=734 failures=0 pending=0 todo=0 tests=734 expected=734\nCRITERION preservation_converter_suites: PASS\n",
+        "stderr": "",
+        "_parameters": {
+          "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc '/bin/bash /hud/grader/run_grading.sh preservation_converter_suites'",
+          "cwd": "/workspace/repo",
+          "timeout_seconds": 900
+        }
+      }
+    }
+  ],
+  "status": "graded",
+  "patch_sha256": "4cbdcd45459e6ca72da3fe41ef8cd70b715d5aa61f7af94c4e1183b2e299a182",
+  "result": {
+    "reward": 1.0,
+    "done": true,
+    "content": null,
+    "info": {
+      "blocker_policy": "weight",
+      "failed_blockers": []
+    },
+    "isError": false,
+    "subscores": [
+      {
+        "name": "g1_catch_value_degrades",
+        "weight": 0.04,
+        "value": 1.0,
+        "children": null,
+        "info": {
+          "exit_code": 0,
+          "stdout": "VITEST passes=6 failures=0 pending=0 todo=0 tests=6 expected=6\nCRITERION g1_catch_value_degrades: PASS\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc '/bin/bash /hud/grader/run_grading.sh g1_catch_value_degrades'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 600
+          }
+        }
+      },
+      {
+        "name": "g6_unencodable_default_degrades",
+        "weight": 0.04,
+        "value": 1.0,
+        "children": null,
+        "info": {
+          "exit_code": 0,
+          "stdout": "VITEST passes=6 failures=0 pending=0 todo=0 tests=6 expected=6\nCRITERION g6_unencodable_default_degrades: PASS\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc '/bin/bash /hud/grader/run_grading.sh g6_unencodable_default_degrades'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 600
+          }
+        }
+      },
+      {
+        "name": "g5_default_value_conforms",
+        "weight": 0.11,
+        "value": 1.0,
+        "children": null,
+        "info": {
+          "exit_code": 0,
+          "stdout": "VITEST passes=6 failures=0 pending=0 todo=0 tests=6 expected=6\nCRITERION g5_default_value_conforms: PASS\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc '/bin/bash /hud/grader/run_grading.sh g5_default_value_conforms'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 600
+          }
+        }
+      },
+      {
+        "name": "g2_empty_positional_list",
+        "weight": 0.09,
+        "value": 1.0,
+        "children": null,
+        "info": {
+          "exit_code": 0,
+          "stdout": "VITEST passes=4 failures=0 pending=0 todo=0 tests=4 expected=4\nCRITERION g2_empty_positional_list: PASS\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc '/bin/bash /hud/grader/run_grading.sh g2_empty_positional_list'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 600
+          }
+        }
+      },
+      {
+        "name": "g2_empty_tuple_rest_element",
+        "weight": 0.05,
+        "value": 1.0,
+        "children": null,
+        "info": {
+          "exit_code": 0,
+          "stdout": "VITEST passes=2 failures=0 pending=0 todo=0 tests=2 expected=2\nCRITERION g2_empty_tuple_rest_element: PASS\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc '/bin/bash /hud/grader/run_grading.sh g2_empty_tuple_rest_element'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 600
+          }
+        }
+      },
+      {
+        "name": "g3_pattern_keyed_record_openapi",
+        "weight": 0.11,
+        "value": 1.0,
+        "children": null,
+        "info": {
+          "exit_code": 0,
+          "stdout": "VITEST passes=6 failures=0 pending=0 todo=0 tests=6 expected=6\nCRITERION g3_pattern_keyed_record_openapi: PASS\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc '/bin/bash /hud/grader/run_grading.sh g3_pattern_keyed_record_openapi'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 600
+          }
+        }
+      },
+      {
+        "name": "g4_metadata_examples_openapi",
+        "weight": 0.09,
+        "value": 1.0,
+        "children": null,
+        "info": {
+          "exit_code": 0,
+          "stdout": "VITEST passes=6 failures=0 pending=0 todo=0 tests=6 expected=6\nCRITERION g4_metadata_examples_openapi: PASS\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc '/bin/bash /hud/grader/run_grading.sh g4_metadata_examples_openapi'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 600
+          }
+        }
+      },
+      {
+        "name": "g7_null_literal_openapi",
+        "weight": 0.09,
+        "value": 1.0,
+        "children": null,
+        "info": {
+          "exit_code": 0,
+          "stdout": "VITEST passes=6 failures=0 pending=0 todo=0 tests=6 expected=6\nCRITERION g7_null_literal_openapi: PASS\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc '/bin/bash /hud/grader/run_grading.sh g7_null_literal_openapi'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 600
+          }
+        }
+      },
+      {
+        "name": "g8_empty_union_accepts_nothing",
+        "weight": 0.09,
+        "value": 1.0,
+        "children": null,
+        "info": {
+          "exit_code": 0,
+          "stdout": "VITEST passes=6 failures=0 pending=0 todo=0 tests=6 expected=6\nCRITERION g8_empty_union_accepts_nothing: PASS\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc '/bin/bash /hud/grader/run_grading.sh g8_empty_union_accepts_nothing'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 600
+          }
+        }
+      },
+      {
+        "name": "g9_enum_members_unique",
+        "weight": 0.09,
+        "value": 1.0,
+        "children": null,
+        "info": {
+          "exit_code": 0,
+          "stdout": "VITEST passes=6 failures=0 pending=0 todo=0 tests=6 expected=6\nCRITERION g9_enum_members_unique: PASS\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc '/bin/bash /hud/grader/run_grading.sh g9_enum_members_unique'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 600
+          }
+        }
+      },
+      {
+        "name": "d1_established_behavior",
+        "weight": 0.1,
+        "value": 1.0,
+        "children": null,
+        "info": {
+          "exit_code": 0,
+          "stdout": "VITEST passes=22 failures=0 pending=0 todo=0 tests=22 expected=22\nCRITERION d1_established_behavior: PASS\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc '/bin/bash /hud/grader/run_grading.sh d1_established_behavior'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 900
+          }
+        }
+      },
+      {
+        "name": "preservation_converter_suites",
+        "weight": 0.10000000000000009,
+        "value": 1.0,
+        "children": null,
+        "info": {
+          "exit_code": 0,
+          "stdout": "VITEST passes=734 failures=0 pending=0 todo=0 tests=734 expected=734\nCRITERION preservation_converter_suites: PASS\n",
+          "stderr": "",
+          "_parameters": {
+            "command": "setpriv --reuid 1000 --regid 1000 --clear-groups --no-new-privs env HOME=/tmp/agent-home bash -lc '/bin/bash /hud/grader/run_grading.sh preservation_converter_suites'",
+            "cwd": "/workspace/repo",
+            "timeout_seconds": 900
+          }
+        }
+      }
+    ]
+  }
+}

harbor/evidence/zod/summary.json

@@ -0,0 +1,150 @@
+{
+  "name": "zod",
+  "task_id": "zod_json_schema_target_legality",
+  "image": "ec-swe-zod:source-v7",
+  "build": {
+    "ok": true,
+    "image_id": "sha256:6c7eb55e9901704ab8e6b81e924c3c9ba35bfef05b814c6487399dc494123852",
+    "ladder": {},
+    "seconds": 0
+  },
+  "qualified": false,
+  "measure": {
+    "ok": true,
+    "hud": "0.6.17",
+    "path": "/usr/local/bin:/usr/bin:/bin:/usr/local/games:/usr/games",
+    "handed": [],
+    "repo_dir": "/workspace/repo",
+    "python_usr_local": false,
+    "seconds": 7,
+    "reused": true
+  },
+  "parity": {
+    "ok": true,
+    "reward": 1.0,
+    "criteria": 12,
+    "unmet": [],
+    "info": {
+      "blocker_policy": "weight",
+      "failed_blockers": []
+    },
+    "seconds": 578,
+    "reused": true
+  },
+  "prepare": {
+    "ok": true,
+    "bundle": "zod-r4",
+    "converter": "b8bba68767da888f",
+    "pinned_files": 26,
+    "criteria": 12,
+    "seconds": 3
+  },
+  "controls": {
+    "ok": false,
+    "checks": {
+      "noop_is_zero": true,
+      "reference_is_one": true,
+      "reference_equals_native": true,
+      "agent_is_isolated": true
+    },
+    "results": {
+      "noop": {
+        "run": "zod-r4-control-noop-1791892270",
+        "reward": 0.0,
+        "stop": "completed",
+        "error": null,
+        "seconds": 1299,
+        "probe": false,
+        "status": "empty_patch",
+        "verifier_error": null,
+        "exit_codes": [],
+        "attempts": 1
+      },
+      "reference": {
+        "run": "zod-r4-control-reference-1791892270",
+        "reward": 1.0,
+        "stop": "completed",
+        "error": null,
+        "seconds": 1383,
+        "probe": true,
+        "status": "graded",
+        "verifier_error": null,
+        "exit_codes": [
+          0
+        ],
+        "subscores": {
+          "g1_catch_value_degrades": [
+            1.0,
+            0.04
+          ],
+          "g6_unencodable_default_degrades": [
+            1.0,
+            0.04
+          ],
+          "g5_default_value_conforms": [
+            1.0,
+            0.11
+          ],
+          "g2_empty_positional_list": [
+            1.0,
+            0.09
+          ],
+          "g2_empty_tuple_rest_element": [
+            1.0,
+            0.05
+          ],
+          "g3_pattern_keyed_record_openapi": [
+            1.0,
+            0.11
+          ],
+          "g4_metadata_examples_openapi": [
+            1.0,
+            0.09
+          ],
+          "g7_null_literal_openapi": [
+            1.0,
+            0.09
+          ],
+          "g8_empty_union_accepts_nothing": [
+            1.0,
+            0.09
+          ],
+          "g9_enum_members_unique": [
+            1.0,
+            0.09
+          ],
+          "d1_established_behavior": [
+            1.0,
+            0.1
+          ],
+          "preservation_converter_suites": [
+            1.0,
+            0.10000000000000009
+          ]
+        },
+        "attempts": 1
+      }
+    },
+    "bundle": "zod-r4",
+    "complete": false,
+    "seconds": 1383
+  },
+  "leak": {
+    "scan_finished": true,
+    "hits_outside_checkout": [],
+    "other_files_holding_only_upstream_lines": 7,
+    "library_files_sharing_a_line": 0,
+    "most_lines_shared_by_one_library_file": 0,
+    "private_paths_absent": true,
+    "public_lines_shared_with_patch_and_graders": [
+      1,
+      751
+    ],
+    "private_lines_only_in_agent_checkout": [
+      0,
+      0
+    ],
+    "ok": true,
+    "seconds": 88
+  }
+}

harbor/manifest.json

@@ -0,0 +1,527 @@
+{
+  "schema_version": 1,
+  "created": "2026-10-13",
+  "contains": "reference solutions and hidden graders; keep away from evaluated models",
+  "package_revision": "coding-rubric-v2",
+  "runtime": {
+    "python": "3.12",
+    "harbor": "0.23.0",
+    "platform": "linux/amd64"
+  },
+  "qualified_with": "stock Harbor 0.23.0 (oracle and nop agents) on Docker",
+  "tasks": [
+    {
+      "task_id": "coredns_rewrite_svcb_opt",
+      "name": "coredns",
+      "dataset_instance_id": "coredns__coredns-coredns__4097",
+      "repository": "coredns/coredns",
+      "language": "Go",
+      "source_folder": "coredns-v4",
+      "dockerfile": "Dockerfile.hud",
+      "source_image": "ec-swe-coredns:source-v4",
+      "source_image_id_at_qualification": "sha256:5f122f58a039aefe65ccf2fd057eceb288c09278edde20f2ff0f2598f52c22e9",
+      "grader_version": "0.6.18",
+      "criteria": 7,
+      "reward_rule": "authored",
+      "blocker_cap": null,
+      "author_calibration": {},
+      "provenance": {
+        "repository": "coredns/coredns",
+        "environment_id": "26a880f2-79fd-403a-8189-570cff511312",
+        "build": 4,
+        "build_id": "0f706606-696c-411a-b95a-727766f336ed",
+        "hosted_task": "coredns__4097",
+        "task_version": 1,
+        "source_archive_sha256": "f3e54c704918dc7b5043354e24681e708f15164f4069f857d4abf9de0faf796d",
+        "hosted_revision_argument": "not recorded; the scenario does not read it"
+      },
+      "checks_on_shipped_files": "all",
+      "further_checks": "broken fix 0; missing grader file: no reward; planted score 0; author's patches: none supplied"
+    },
+    {
+      "task_id": "epicyon_deleted_post_announce_cache_contract",
+      "name": "epicyon",
+      "dataset_instance_id": "bashrc2__epicyon-epicyon_deleted_post_announce_cache_contract-1c066303",
+      "repository": "bashrc2/epicyon",
+      "language": "Python",
+      "source_folder": "epicyon-v7",
+      "dockerfile": "Dockerfile.hud",
+      "source_image": "ec-swe-epicyon:source-v7",
+      "source_image_id_at_qualification": "sha256:74752fd4e8ebf481d6f0c01fd9b7e542c539f850bf9eed8503f41aaeabd7d99d",
+      "grader_version": "0.6.15",
+      "criteria": 11,
+      "reward_rule": "weighted_cap",
+      "blocker_cap": 0.6,
+      "author_calibration": {
+        "gold": 1.0,
+        "base": 0.0,
+        "noop": 0.0,
+        "lazy": 0.085,
+        "alt": 1.0,
+        "cheat": 0.6
+      },
+      "provenance": {
+        "repository": "bashrc2/epicyon",
+        "environment_id": "c29bfa96-5ecd-4ebd-9741-1f74147966c2",
+        "build": 7,
+        "build_id": "dd339e97-119f-4162-be02-d283a0784ff5",
+        "hosted_task": "epicyon_deleted_post_announce_cache_contract-1c066303",
+        "task_version": 1,
+        "source_archive_sha256": "f8c5e69a0a61ee283feb452cc2ed4e0645fe4ceff039b772af6119581f7151fb",
+        "hosted_revision_argument": "not recorded; the scenario does not read it"
+      },
+      "source_edited": {
+        "changed": [
+          "env.py",
+          "pyproject.toml"
+        ],
+        "removed": [],
+        "what": "the environment name was changed to a neutral one (ec-swe-epicyon-deleted-post-announce-cache)"
+      },
+      "checks_on_shipped_files": "reference, empty change, image scan",
+      "further_checks": "not yet; on the preceding build, with stock Harbor: broken fix 0; missing grader file: no reward; planted score 0; author's patches: 4 of 4"
+    },
+    {
+      "task_id": "rope_rename_all_reexport_owners_contract",
+      "name": "rope",
+      "dataset_instance_id": "python-rope__rope-rope_rename_all_reexport_owners_contract-1c066303",
+      "repository": "python-rope/rope",
+      "language": "Python",
+      "source_folder": "rope-v3",
+      "dockerfile": "Dockerfile.hud",
+      "source_image": "ec-swe-rope:source-v3",
+      "source_image_id_at_qualification": "sha256:e33bc697f20f1281afeb4f9862caa9dc697b1c045eb68e0d5425565a1d7196e5",
+      "grader_version": "0.6.15",
+      "criteria": 8,
+      "reward_rule": "weighted_cap",
+      "blocker_cap": 0.6,
+      "author_calibration": {
+        "gold": 1.0,
+        "base": 0.0,
+        "noop": 0.0,
+        "lazy": 0.0,
+        "alt": 0.84,
+        "prevgold": 0.19,
+        "cheat": 0.08,
+        "tamper": 0.0,
+        "tamper_key": 0.0,
+        "tamper_gate": 0.16,
+        "tamper_report": 0.16,
+        "tamper_wire": 0.16
+      },
+      "provenance": {
+        "repository": "python-rope/rope",
+        "environment_id": "98a7b668-7394-4dc8-a222-d1759208e2c0",
+        "build": 3,
+        "build_id": "2eef5e21-3eec-45df-aaee-043567b1bdf1",
+        "hosted_task": "rope_rename_all_reexport_owners_contract-1c066303",
+        "task_version": 1,
+        "source_archive_sha256": "09367f6384cda46d917e9ff0e0a3968e6d85ffbbc6c767c1825f6b69d97dffc3",
+        "hosted_revision_argument": "not recorded; the scenario does not read it"
+      },
+      "source_edited": {
+        "changed": [
+          "env.py",
+          "pyproject.toml"
+        ],
+        "removed": [],
+        "what": "the environment name was changed to a neutral one (ec-swe-rope-rename-all-reexport-owners)"
+      },
+      "checks_on_shipped_files": "reference, empty change, image scan",
+      "further_checks": "not yet; on the preceding build, with stock Harbor: broken fix 0; missing grader file: no reward; planted score 0; author's patches: 10 of 10"
+    },
+    {
+      "task_id": "zod_json_schema_target_legality",
+      "name": "zod",
+      "dataset_instance_id": "colinhacks__zod-zod-json-schema-target-legality",
+      "repository": "colinhacks/zod",
+      "language": "TypeScript",
+      "source_folder": "zod-v7",
+      "dockerfile": "Dockerfile.hud",
+      "source_image": "ec-swe-zod:source-v7",
+      "source_image_id_at_qualification": "sha256:6c7eb55e9901704ab8e6b81e924c3c9ba35bfef05b814c6487399dc494123852",
+      "grader_version": "0.6.17",
+      "criteria": 12,
+      "reward_rule": "authored",
+      "blocker_cap": null,
+      "author_calibration": {},
+      "provenance": {
+        "repository": "colinhacks/zod",
+        "environment_id": "130424c5-8a1d-4cb7-b5b0-ffa04b564d03",
+        "build": 7,
+        "build_id": "af9abdd0-1f19-4d21-80da-d60924ccc128",
+        "hosted_task": "zod-json-schema-target-legality",
+        "task_version": 1,
+        "source_archive_sha256": "feb1e6acfc057e95a3a89532cb7f322f2c08f6bfea65556d66a04ceca877a5bc",
+        "hosted_revision_argument": "not recorded; the scenario does not read it"
+      },
+      "checks_on_shipped_files": "reference, empty change, image scan",
+      "further_checks": "not yet; on an earlier build, with a different launcher: broken fix 0; missing grader file: no reward. Planted score: not run; author's patches: none supplied"
+    }
+  ],
+  "unpackaged": [
+    {
+      "source_edited": {
+        "changed": [
+          "env.py",
+          "tasks.py"
+        ],
+        "removed": [],
+        "what": "the environment name was changed to a neutral one (ec-swe-click)"
+      },
+      "task_id": "click",
+      "name": "click",
+      "dataset_instance_id": "pallets__click-click-help-defaults",
+      "repository": "pallets/click",
+      "source_folder": "click-v9",
+      "dockerfile": "Dockerfile.hud",
+      "source_image": "ec-swe-click:source-v9",
+      "package": null,
+      "why_no_package": "one criterion is judged by a model"
+    },
+    {
+      "source_edited": {
+        "changed": [
+          "env.py",
+          "tasks.py"
+        ],
+        "removed": [],
+        "what": "the environment name was changed to a neutral one (ec-swe-sequelize)"
+      },
+      "task_id": "sequelize",
+      "name": "sequelize",
+      "dataset_instance_id": "sequelize__sequelize-sequelize-bulkcreate-ignore-duplicates",
+      "repository": "sequelize/sequelize",
+      "source_folder": "sequelize-v27",
+      "dockerfile": "Dockerfile.hud",
+      "source_image": "ec-swe-sequelize:source-v27",
+      "package": null,
+      "why_no_package": "one criterion is judged by a model"
+    },
+    {
+      "source_edited": {
+        "changed": [
+          "env.py"
+        ],
+        "removed": [],
+        "what": "the environment name was changed to a neutral one (ec-swe-hooks2675-v2)"
+      },
+      "task_id": "hooks",
+      "name": "hooks",
+      "dataset_instance_id": "alibaba__hooks-ahooks_request_cache_races",
+      "repository": "alibaba/hooks",
+      "source_folder": "hooks-v4",
+      "dockerfile": "Dockerfile.hud",
+      "source_image": "ec-swe-hooks:source-v4",
+      "package": null,
+      "why_no_package": "one criterion is judged by a model"
+    },
+    {
+      "source_edited": {
+        "changed": [
+          "preflight.sh",
+          "tasks.py"
+        ],
+        "removed": [
+          ".hud/config.json",
+          "task/grader/maintainer_review.judge.json.as-submitted-2026-09-18",
+          "tasks.py.bak-2026-09-20"
+        ],
+        "what": "comments that recorded internal review history were shortened in two files, and three files the image build does not read were removed; tasks.py is the same program"
+      },
+      "task_id": "druid",
+      "name": "druid",
+      "dataset_instance_id": "alibaba__druid-druid_lenient_script",
+      "repository": "alibaba/druid",
+      "source_folder": "druid-v9",
+      "dockerfile": "Dockerfile.hud",
+      "source_image": "ec-swe-druid:source-v9",
+      "package": null,
+      "why_no_package": "one criterion is judged by a model"
+    },
+    {
+      "source_edited": {
+        "changed": [
+          "env.py",
+          "pyproject.toml"
+        ],
+        "removed": [],
+        "what": "the environment name was changed to a neutral one (ec-swe-sickchill-episode-rename-sidecars)"
+      },
+      "task_id": "sickchill",
+      "name": "sickchill",
+      "dataset_instance_id": "SickChill__SickChill-sickchill_episode_metadata_files_contract-2e2a49e4",
+      "repository": "SickChill/SickChill",
+      "source_folder": "sickchill-v7",
+      "dockerfile": "Dockerfile.hud",
+      "source_image": "ec-swe-sickchill:source-v7",
+      "package": null,
+      "why_no_package": "its gate runs repository tests that fetch from pypi.org, so it cannot pass offline"
+    },
+    {
+      "task_id": "mikro-orm",
+      "name": "mikro-orm",
+      "dataset_instance_id": "mikro-orm__mikro-orm-mikro-orm-pessimistic-locking",
+      "repository": "mikro-orm/mikro-orm",
+      "source_folder": "mikro-orm-v7",
+      "dockerfile": "Dockerfile.hud",
+      "source_image": "ec-swe-mikro-orm:source-v7",
+      "package": null,
+      "why_no_package": "its grader starts database servers under their own accounts, which needs root; the packaged verifier runs graders unprivileged"
+    },
+    {
+      "source_edited": {
+        "changed": [
+          "env.py",
+          "pyproject.toml"
+        ],
+        "removed": [],
+        "what": "the environment name was changed to a neutral one (ec-swe-babylonjs-scene-remove-owner-refs)"
+      },
+      "task_id": "babylonjs",
+      "name": "babylonjs",
+      "dataset_instance_id": "BabylonJS__Babylon.js-babylonjs_scene_remove_owner_refs_contract-382616f8",
+      "repository": "BabylonJS/Babylon.js",
+      "source_folder": "babylonjs-v8",
+      "dockerfile": "Dockerfile.hud",
+      "source_image": "ec-swe-babylonjs:source-v8",
+      "package": null,
+      "why_no_package": "too heavy for the qualification host: its setup's ownership change of a very large dependency tree exceeded its ten-minute limit twice, and on a third try grading the reference fix had not finished after two hours"
+    },
+    {
+      "task_id": "aioftp",
+      "name": "aioftp",
+      "dataset_instance_id": "aio-libs__aioftp-aioftp-stream-abort-desynchronization",
+      "repository": "aio-libs/aioftp",
+      "source_folder": "aioftp-v11",
+      "dockerfile": "Dockerfile.hud",
+      "source_image": "ec-swe-aioftp:source-v11",
+      "package": null,
+      "why_no_package": "different task template; not converted yet"
+    },
+    {
+      "task_id": "apscheduler",
+      "name": "apscheduler",
+      "dataset_instance_id": "agronholm__apscheduler-apscheduler-expired-job-lease",
+      "repository": "agronholm/apscheduler",
+      "source_folder": "apscheduler-v17",
+      "dockerfile": "Dockerfile.hud",
+      "source_image": "ec-swe-apscheduler:source-v17",
+      "package": null,
+      "why_no_package": "different task template; not converted yet"
+    }
+  ],
+  "being_rechecked": [
+    {
+      "task_id": "badger",
+      "name": "badger",
+      "dataset_instance_id": "dgraph-io__badger-journal_recovery_after_unclean_stop",
+      "repository": "dgraph-io/badger",
+      "source_folder": "badger-v2",
+      "dockerfile": "Dockerfile.hud",
+      "source_edited": {
+        "changed": [
+          "env.py",
+          "pyproject.toml",
+          "tasks.py"
+        ],
+        "removed": [],
+        "what": "the environment name was changed to a neutral one (ec-swe-badger-journal)"
+      },
+      "package": null,
+      "why_no_package": "its build files were given a neutral environment name; the package is being rebuilt and re-checked on them"
+    },
+    {
+      "task_id": "bazel-remote",
+      "name": "bazel-remote",
+      "dataset_instance_id": "buchgr__bazel-remote-bazel_remote_disk_eviction_accounting_contract-afe5ea56",
+      "repository": "buchgr/bazel-remote",
+      "source_folder": "bazel-remote-v2",
+      "dockerfile": "Dockerfile.hud",
+      "source_edited": {
+        "changed": [
+          "env.py",
+          "pyproject.toml"
+        ],
+        "removed": [],
+        "what": "the environment name was changed to a neutral one (ec-swe-bazel-remote-disk-eviction-accounting)"
+      },
+      "package": null,
+      "why_no_package": "its build files were given a neutral environment name; the package is being rebuilt and re-checked on them"
+    },
+    {
+      "task_id": "celery",
+      "name": "celery",
+      "dataset_instance_id": "celery__celery-celery_retry_expires_countdown_contract",
+      "repository": "celery/celery",
+      "source_folder": "celery-v12",
+      "dockerfile": "Dockerfile.hud",
+      "source_edited": {
+        "changed": [
+          "env.py",
+          "pyproject.toml"
+        ],
+        "removed": [],
+        "what": "the environment name was changed to a neutral one (ec-swe-celery-retry-expires-countdown)"
+      },
+      "package": null,
+      "why_no_package": "its build files were given a neutral environment name; the package is being rebuilt and re-checked on them"
+    },
+    {
+      "task_id": "geode",
+      "name": "geode",
+      "dataset_instance_id": "apache__geode-geode_oql_index_null_semantics_contract-1c066303",
+      "repository": "apache/geode",
+      "source_folder": "geode-v3",
+      "dockerfile": "Dockerfile.hud",
+      "source_edited": {
+        "changed": [
+          "env.py",
+          "pyproject.toml"
+        ],
+        "removed": [],
+        "what": "the environment name was changed to a neutral one (ec-swe-geode-oql-index-null-semantics)"
+      },
+      "package": null,
+      "why_no_package": "its build files were given a neutral environment name; the package is being rebuilt and re-checked on them"
+    },
+    {
+      "task_id": "kratos",
+      "name": "kratos",
+      "dataset_instance_id": "ory__kratos-kratos_cleanup_stale_flows_courier_contract",
+      "repository": "ory/kratos",
+      "source_folder": "kratos-v10",
+      "dockerfile": "Dockerfile.hud",
+      "source_edited": {
+        "changed": [
+          "env.py",
+          "pyproject.toml"
+        ],
+        "removed": [],
+        "what": "the environment name was changed to a neutral one (ec-swe-kratos-cleanup-stale-flows-courier)"
+      },
+      "package": null,
+      "why_no_package": "its build files were given a neutral environment name; the package is being rebuilt and re-checked on them"
+    },
+    {
+      "task_id": "microblogpub",
+      "name": "microblogpub",
+      "dataset_instance_id": "tsileo__microblog.pub-microblogpub_deleted_object_resurrect_lookup_contract-382616f8",
+      "repository": "tsileo/microblog.pub",
+      "source_folder": "microblogpub-v4",
+      "dockerfile": "Dockerfile.hud",
+      "source_edited": {
+        "changed": [
+          "env.py",
+          "pyproject.toml"
+        ],
+        "removed": [],
+        "what": "the environment name was changed to a neutral one (ec-swe-microblogpub-deleted-object-resurrect-lookup)"
+      },
+      "package": null,
+      "why_no_package": "its build files were given a neutral environment name; the package is being rebuilt and re-checked on them"
+    },
+    {
+      "task_id": "open-xml-sdk",
+      "name": "open-xml-sdk",
+      "dataset_instance_id": "dotnet__Open-XML-SDK-open_xml_sdk_deletepart_refs_contract-1c066303",
+      "repository": "dotnet/Open-XML-SDK",
+      "source_folder": "open-xml-sdk-v5",
+      "dockerfile": "Dockerfile.hud",
+      "source_edited": {
+        "changed": [
+          "env.py",
+          "pyproject.toml"
+        ],
+        "removed": [],
+        "what": "the environment name was changed to a neutral one (ec-swe-open-xml-sdk-deletepart-refs)"
+      },
+      "package": null,
+      "why_no_package": "its build files were given a neutral environment name; the package is being rebuilt and re-checked on them"
+    },
+    {
+      "task_id": "openfire",
+      "name": "openfire",
+      "dataset_instance_id": "igniterealtime__Openfire-openfire_userdeleting_last_owner_contract-0163de9c",
+      "repository": "igniterealtime/Openfire",
+      "source_folder": "openfire-v9",
+      "dockerfile": "Dockerfile.hud",
+      "source_edited": {
+        "changed": [
+          "env.py",
+          "pyproject.toml"
+        ],
+        "removed": [],
+        "what": "the environment name was changed to a neutral one (ec-swe-openfire-userdeleting-last-owner)"
+      },
+      "package": null,
+      "why_no_package": "its build files were given a neutral environment name; the package is being rebuilt and re-checked on them"
+    },
+    {
+      "task_id": "terser",
+      "name": "terser",
+      "dataset_instance_id": "terser__terser-retained_top_level_names_keep_their_values",
+      "repository": "terser/terser",
+      "source_folder": "terser-v3",
+      "dockerfile": "Dockerfile.hud",
+      "source_edited": {
+        "changed": [
+          "env.py",
+          "tasks.py"
+        ],
+        "removed": [],
+        "what": "the environment name was changed to a neutral one (ec-swe-terser)"
+      },
+      "package": null,
+      "why_no_package": "its build files were given a neutral environment name; the package is being rebuilt and re-checked on them"
+    },
+    {
+      "task_id": "umya",
+      "name": "umya",
+      "dataset_instance_id": "MathNya__umya-spreadsheet-umya_structural_edit_reference_owners_contract",
+      "repository": "MathNya/umya-spreadsheet",
+      "source_folder": "umya-v4",
+      "dockerfile": "Dockerfile.hud",
+      "source_edited": {
+        "changed": [
+          "env.py",
+          "pyproject.toml"
+        ],
+        "removed": [],
+        "what": "the environment name was changed to a neutral one (ec-swe-umya-structural-edit-reference-owners)"
+      },
+      "package": null,
+      "why_no_package": "its build files were given a neutral environment name; the package is being rebuilt and re-checked on them"
+    },
+    {
+      "task_id": "zettlr",
+      "name": "zettlr",
+      "dataset_instance_id": "Zettlr__Zettlr-zettlr_clickable_keywords",
+      "repository": "Zettlr/Zettlr",
+      "source_folder": "zettlr-v32",
+      "dockerfile": "Dockerfile.hud",
+      "source_edited": {
+        "changed": [
+          "env.py"
+        ],
+        "removed": [],
+        "what": "the environment name was changed to a neutral one (ec-swe-zettir-v1)"
+      },
+      "package": null,
+      "why_no_package": "its build files were given a neutral environment name; the package is being rebuilt and re-checked on them"
+    }
+  ],
+  "not_included": {
+    "badger": "its build files were given a neutral environment name; the package is being rebuilt and re-checked on them",
+    "bazel-remote": "its build files were given a neutral environment name; the package is being rebuilt and re-checked on them",
+    "celery": "its build files were given a neutral environment name; the package is being rebuilt and re-checked on them",
+    "geode": "its build files were given a neutral environment name; the package is being rebuilt and re-checked on them",
+    "kratos": "its build files were given a neutral environment name; the package is being rebuilt and re-checked on them",
+    "microblogpub": "its build files were given a neutral environment name; the package is being rebuilt and re-checked on them",
+    "open-xml-sdk": "its build files were given a neutral environment name; the package is being rebuilt and re-checked on them",
+    "openfire": "its build files were given a neutral environment name; the package is being rebuilt and re-checked on them",
+    "terser": "its build files were given a neutral environment name; the package is being rebuilt and re-checked on them",
+    "umya": "its build files were given a neutral environment name; the package is being rebuilt and re-checked on them",
+    "zettlr": "its build files were given a neutral environment name; the package is being rebuilt and re-checked on them"
+  }
+}

harbor/run_checks.py

@@ -0,0 +1,153 @@
+"""Run the scripted checks on these tasks with Harbor's own agents.
+
+    python run_checks.py --task <task id> --output checks-001
+    python run_checks.py --task <task id> --case noop --output smoke-001
+    python run_checks.py --output all-checks-001          # every task, every case
+
+Needs the `harbor` command (pip install harbor==0.23.0) and Docker, on Linux,
+macOS or WSL, and the task's source image (build_images.py). No model is
+called. Each case is one `harbor run` on a copy of the task:
+
+  noop       Harbor's nop agent changes nothing              -> reward 0
+  reference  Harbor's oracle agent applies the reference fix -> reward 1
+             (it first checks, from inside the agent's container, that it is
+             an unprivileged user, sees no grading files and gets no traffic out)
+  broken     the reference fix plus a syntax error           -> reward 0, graded
+  infra      a grader file is removed from the verifier      -> no reward at all
+  tamper     the "agent" keeps writing itself a full score   -> reward 0
+
+Results go to <output>/summary.json; the script exits 0 only if every case
+gave what is expected. Every run needs a new output directory name.
+"""
+
+from __future__ import annotations
+
+import argparse
+import json
+import re
+import shlex
+import shutil
+import subprocess
+import sys
+import time
+from pathlib import Path
+
+CASES = ["noop", "reference", "broken", "infra", "tamper"]
+EXPECTED = {"noop": 0.0, "reference": 1.0, "broken": 0.0, "infra": None, "tamper": 0.0}
+PROBE = r"""/opt/task-tools/python -c 'import os,socket; from pathlib import Path
+assert os.getuid()==1000, "the agent is not uid 1000"
+private=("/hud", "/app/tasks.py", "/tests", "/var/run/docker.sock")
+assert not any(Path(p).exists() for p in private), "a private path is visible"
+data=b""
+try:
+    s=socket.create_connection(("1.1.1.1",443),timeout=4); s.settimeout(4)
+    s.sendall(b"GET / HTTP/1.0\r\n\r\n"); data=s.recv(16)
+except OSError:
+    pass
+assert not data, "the network is reachable"
+print("agent isolation probe passed")'
+"""
+PLANT = """#!/bin/bash
+mkdir -p /logs/verifier 2>/dev/null
+( while true; do
+    echo 1 > /logs/verifier/reward.txt 2>/dev/null
+    echo '{"reward": 1}' > /logs/verifier/reward.json 2>/dev/null
+    sleep 0.2
+  done ) > /dev/null 2>&1 &
+sleep 2
+exit 0
+"""
+SOURCE_SUFFIXES = (".py", ".go", ".java", ".rs", ".cs", ".ts", ".tsx", ".js", ".mjs", ".cjs")
+TEST_PATH = re.compile(
+    r"/(tests?|__tests__|specs?|testing)/"
+    r"|/test[^/]*$|[._-](test|spec)\.[a-z]+$|_test\.go$|Tests?\.(java|cs)$|/[^/]*tests?\.py$",
+    re.IGNORECASE,
+)
+
+
+def variant_for(case: str, package: Path, scratch: Path) -> tuple[Path, str]:
+    """Copy the task and change what this case submits; return the copy and the agent."""
+    variant = scratch / case / package.name
+    shutil.copytree(package, variant)
+    solve = variant / "solution/solve.sh"
+    if case == "noop":
+        return variant, "nop"
+    if case == "reference":
+        marker = "cd /workspace/repo\n"
+        text = solve.read_text()
+        solve.write_text(text.replace(marker, marker + PROBE, 1), newline="\n")
+    elif case == "broken":
+        patch = (variant / "solution/reference.patch").read_text()
+        paths = re.findall(r"^\+\+\+ b/(\S+)", patch, flags=re.MULTILINE)
+        target = next(p for p in paths if p.endswith(SOURCE_SUFFIXES) and not TEST_PATH.search(f"/{p}"))
+        with solve.open("a", newline="\n") as handle:
+            handle.write(f"printf '\\nBROKEN SYNTAX !\\n' >> {shlex.quote(target)}\n")
+    elif case == "infra":
+        config = json.loads((variant / "tests/config.json").read_text())
+        pinned = sorted(f for f in config["verifier_files"] if f.startswith("/hud/grader/"))
+        script = variant / "tests/test.sh"
+        lines = script.read_text().splitlines()
+        lines.insert(len(lines) - 1, f"rm {shlex.quote(pinned[0])}")
+        script.write_text("\n".join(lines) + "\n", newline="\n")
+    elif case == "tamper":
+        solve.write_text(PLANT, newline="\n")
+    return variant, "oracle"
+
+
+def main() -> None:
+    parser = argparse.ArgumentParser(description=__doc__, formatter_class=argparse.RawDescriptionHelpFormatter)
+    parser.add_argument("--output", required=True)
+    parser.add_argument("--task", action="append", default=[])
+    parser.add_argument("--case", action="append", choices=CASES)
+    args = parser.parse_args()
+    root = Path(__file__).resolve().parent
+    if shutil.which("harbor") is None:
+        raise SystemExit("The harbor command is not on PATH: pip install harbor==0.23.0")
+    output = root / args.output
+    if Path(args.output).name != args.output:
+        raise SystemExit("--output must be a new directory name within the bundle")
+    output.mkdir(exist_ok=False)
+    known = json.loads((root / "tasks/manifest.json").read_text())["tasks"]
+    unknown = [name for name in args.task if name not in known]
+    if unknown:
+        raise SystemExit("Unknown task: " + ", ".join(unknown))
+    summary = []
+    for task in args.task or known:
+        for case in args.case or CASES:
+            variant, agent = variant_for(case, root / "tasks" / task, output / "variants" / task)
+            job = f"{task[:40]}-{case}"
+            started = time.time()
+            log = output / f"{job}.log"
+            with log.open("wb") as handle:
+                subprocess.run(
+                    ["harbor", "run", "-p", str(variant), "-a", agent, "--jobs-dir", str(output / "jobs"),
+                     "--job-name", job, "-n", "1"],
+                    stdout=handle, stderr=subprocess.STDOUT,
+                )
+            results = sorted((output / "jobs" / job).glob("*/result.json"))
+            reward, error, probe = None, "Harbor wrote no trial result", None
+            if results:
+                data = json.loads(results[0].read_text())
+                failure = data.get("exception_info")
+                error = f"{failure.get('exception_type')}: {failure.get('exception_message')}"[:300] if failure else None
+                if failure is None:
+                    reward = ((data.get("verifier_result") or {}).get("rewards") or {}).get("reward")
+                agent_dir = results[0].parent / "agent"
+                said = "".join(f.read_text(errors="replace") for f in agent_dir.rglob("*") if f.is_file()) if agent_dir.is_dir() else ""
+                probe = "agent isolation probe passed" in said
+            ok = reward == EXPECTED[case] and (error is None) == (case != "infra")
+            if case == "reference":
+                ok = ok and bool(probe)
+            row = {"task": task, "case": case, "reward": reward, "expected": EXPECTED[case], "ok": ok,
+                   "error": error, "seconds": round(time.time() - started)}
+            if case == "reference":
+                row["agent_isolated"] = probe
+            summary.append(row)
+            (output / "summary.json").write_text(json.dumps(summary, indent=2) + "\n")
+            print(json.dumps(row), flush=True)
+    shutil.rmtree(output / "variants", ignore_errors=True)
+    sys.exit(0 if summary and all(row["ok"] for row in summary) else 1)
+
+
+if __name__ == "__main__":
+    main()

harbor/sources/aioftp-v11/.dockerignore

@@ -0,0 +1,14 @@
+.venv/
+__pycache__
+*.pyc
+.pytest_cache/
+.ruff_cache/
+.hud/
+.git/
+.github/
+tests/
+!grading/baseline/tests/
+!grading/baseline/tests/**
+grading/adversarial/
+artifact-manifest.json
+scaffold-manifest.json

harbor/sources/aioftp-v11/.gitignore

@@ -0,0 +1,9 @@
+.venv/
+__pycache__/
+*.pyc
+.pytest_cache/
+.ruff_cache/
+.hud/
+dist/
+build/
+*.egg-info/

harbor/sources/aioftp-v11/Dockerfile.hud

@@ -0,0 +1,126 @@
+# syntax=docker/dockerfile:1
+# Stage `runtime-contract` is the credential-free, source-free runtime the graders, the
+# repository-native commands, and the agent prompt's tools run on. `validate-helper.py
+# runtime-preflight` builds only this target (no BuildKit secret) and probes every declared
+# executable and service here; the final `runtime` stage derives from it unchanged, so what the
+# probe proves is what the served image carries. Keep credentials and the private clone out of it.
+FROM ubuntu:24.04@sha256:224a1869083a311ef3f13648a154ba79832fbef6364d31493642ca03082da254 AS runtime-contract
+
+# Resolve uv once by immutable OCI digest. Independent local and HUD builds therefore copy the
+# same binary instead of asking PyPI for whatever version happens to be current.
+COPY --from=ghcr.io/astral-sh/uv:0.12.3@sha256:2d890623d310b57771ce840f0da5eed5fc6d657da05ffaa45d82797b53fa3abc /uv /uvx /bin/
+
+# Every apt package resolves against one dated Ubuntu snapshot. The image has no CA bundle yet,
+# so the first index fetch skips TLS peer checks only to install ca-certificates; apt still
+# verifies every index against the archive's signing keys, and every later fetch uses TLS.
+RUN echo 'APT::Snapshot "20260915T000000Z";' > /etc/apt/apt.conf.d/50snapshot \
+  && apt-get -o Acquire::https::Verify-Peer=false update -y \
+  && apt-get -o Acquire::https::Verify-Peer=false install -y --no-install-recommends ca-certificates \
+  && apt-get update -y \
+  && apt-get install -y --no-install-recommends \
+  bash \
+  build-essential \
+  curl \
+  git \
+  gnupg \
+  libpq5 \
+  openssl \
+  python-is-python3 \
+  python3 \
+  python3-dev \
+  python3-pip \
+  python3-venv \
+  python3-wheel \
+  rsync \
+  util-linux \
+  sqlite3 \
+  vim \
+  wget \
+  && rm -rf /var/lib/apt/lists/* \
+  && update-ca-certificates
+
+# Repository runtime: CPython 3.11.15 (uv-managed, hash-verified download) with the repository's
+# `dev` extra from a fully hashed lock. The editable path and the dist-info record stand in for
+# `pip install -e .[dev]` of the workspace cloned below, so `pytest` from the repository root, the
+# native test command, and the hidden probes all import `aioftp` from the tree being worked on.
+# `aioftp_ipv6_loopback` is a pytest11 plugin that skips only the suite's `::1` nodes, and only
+# where IPv6 loopback cannot be bound (HUD's sandbox has none); with `::1` available it is inert.
+# `pytest-timeout` fails one hung test after PYTEST_TIMEOUT seconds with its traceback (default
+# `signal` method; the rest of the run continues) instead of the whole shell call expiring with no
+# output. Only the agent shell inherits the variable: the grader runs pytest with a clean
+# environment and the pristine settings, where the plugin has no timeout and is inert.
+ENV UV_PYTHON_INSTALL_DIR=/opt/uv-python \
+    PYTEST_TIMEOUT=30
+COPY ./runtime/aioftp-dev-requirements.txt /opt/aioftp-runtime/requirements.txt
+COPY ./runtime/aioftp_ipv6_loopback.py /opt/aioftp-runtime/aioftp_ipv6_loopback.py
+RUN uv python install --no-bin 3.11.15 \
+  && uv venv --python 3.11.15 /opt/aioftp-venv \
+  && uv pip install --python /opt/aioftp-venv/bin/python --no-cache --require-hashes --no-deps -r /opt/aioftp-runtime/requirements.txt \
+  && site="$(/opt/aioftp-venv/bin/python -c 'import sysconfig; print(sysconfig.get_paths()["purelib"])')" \
+  && mkdir -p "$site/aioftp-0.28.3.dist-info" \
+  && printf 'Metadata-Version: 2.1\nName: aioftp\nVersion: 0.28.3\n' > "$site/aioftp-0.28.3.dist-info/METADATA" \
+  && printf 'uv\n' > "$site/aioftp-0.28.3.dist-info/INSTALLER" \
+  && : > "$site/aioftp-0.28.3.dist-info/RECORD" \
+  && echo /home/ubuntu/aioftp/src > "$site/__editable__.aioftp-0.28.3.pth" \
+  && cp /opt/aioftp-runtime/aioftp_ipv6_loopback.py "$site/aioftp_ipv6_loopback.py" \
+  && mkdir -p "$site/aioftp_ipv6_loopback-1.0.dist-info" \
+  && printf 'Metadata-Version: 2.1\nName: aioftp-ipv6-loopback\nVersion: 1.0\n' > "$site/aioftp_ipv6_loopback-1.0.dist-info/METADATA" \
+  && printf '[pytest11]\naioftp_ipv6_loopback = aioftp_ipv6_loopback\n' > "$site/aioftp_ipv6_loopback-1.0.dist-info/entry_points.txt" \
+  && : > "$site/aioftp_ipv6_loopback-1.0.dist-info/RECORD" \
+  && chmod -R a+rX /opt/uv-python /opt/aioftp-venv /opt/aioftp-runtime
+
+# Grader and HUD server virtualenv, from the package lock (Python 3.12 from the snapshot).
+COPY ./pyproject.toml ./uv.lock /hud/
+WORKDIR /hud
+ENV RUST_LOG=warn \
+    PYTHONPATH=/hud/.venv/lib/python3.12/site-packages:/hud \
+    PATH=/hud/.venv/bin:/opt/aioftp-venv/bin:$PATH \
+    AIOFTP_REPO_PYTHON=/opt/aioftp-venv/bin/python
+RUN uv venv --python /usr/bin/python3 /hud/.venv \
+  && uv sync --frozen --no-install-project --no-cache --python /hud/.venv/bin/python
+
+FROM runtime-contract AS setup
+
+ARG REPO_URL="https://github.com/aio-libs/aioftp"
+ARG BASE_COMMIT="4c18bdec8eb8f2e12a226fc513d07b92b5014167"
+ARG FOLDER_NAME="aioftp"
+
+ENV FOLDER_NAME=${FOLDER_NAME} \
+    REPO_URL=${REPO_URL} \
+    BASE_COMMIT=${BASE_COMMIT} \
+    HOME=/home/ubuntu \
+    PROJECT_DIR=/home/ubuntu/${FOLDER_NAME} \
+    PATCHES_DIR=/hud/patches
+
+# The base repository is public (the upstream itself, or the published base), so the pinned
+# commit is fetched by SHA with no credential and no other ref: the image never holds a branch
+# that could reach the gold, and no BuildKit secret is mounted anywhere in this Dockerfile.
+USER ubuntu
+RUN git init -q /home/ubuntu/${FOLDER_NAME} && \
+    cd /home/ubuntu/${FOLDER_NAME} && \
+    git fetch -q --depth 1 --no-tags ${REPO_URL} ${BASE_COMMIT} && \
+    git checkout -q --detach ${BASE_COMMIT} && \
+    rm -f .git/FETCH_HEAD
+
+USER root
+RUN chown -R ubuntu:ubuntu /home/ubuntu/${FOLDER_NAME} \
+    && mkdir -p /hud/patches \
+    && chown -R root:root /home/ubuntu/${FOLDER_NAME}/.git \
+    && chmod -R 700 /home/ubuntu/${FOLDER_NAME}/.git
+
+FROM setup AS runtime
+
+USER root
+WORKDIR /hud
+COPY ./env.py /hud/env.py
+COPY ./grading /hud/grading
+COPY ./tasks.py /hud/tasks.py
+COPY ./task /hud/task
+COPY ./gold.diff /hud/gold.diff
+RUN chmod -R 700 /hud \
+    && chown -R root:root /hud
+RUN chmod 700 /root
+
+ENV PROBLEM_ID=""
+EXPOSE 8765
+CMD ["hud", "serve", "env:env", "--host", "0.0.0.0", "--port", "8765"]

harbor/sources/aioftp-v11/env.py

@@ -0,0 +1,700 @@
+"""Coding environment (HUD v6) for aioftp_stream_abort_desynchronization.
+
+The environment publishes one ssh capability (a UID-walled Workspace); the agent's
+harness brings its own bash/file tools. A task is a @env.template async
+generator: yield a prompt, then yield a reward.
+
+Grading entry point contract: this module must never import the grader chain
+(``grading.*``) at module import time. The chain is resolved lazily inside
+``run_grading`` so that a missing or corrupted grader module degrades to a
+fail-closed JSON result instead of aborting the import and emitting no output
+at all. Prompt 05 populates the grader behind ``GRADER_ENTRY_POINT``; it must
+not add a module-level ``grading`` import here.
+"""
+
+import asyncio
+import base64
+import datetime
+import gzip
+import hashlib
+import importlib
+import inspect
+import json
+import logging
+import os
+import shutil
+import stat
+import subprocess
+import sys
+import tempfile
+import time
+import traceback
+from pathlib import Path
+
+from hud.environment import Environment, Workspace
+from hud.graders import EvaluationResult
+
+logger = logging.getLogger(__name__)
+
+TASK_ID = "aioftp_stream_abort_desynchronization"
+TASK_SLUG = "aioftp-stream-abort-desynchronization"
+HUD_ENV_NAME = "aioftp-stream-abort-desynchronization"
+FOLDER_NAME = os.environ.get("FOLDER_NAME", "aioftp")
+REPO_URL = os.environ.get("REPO_URL", "https://github.com/aio-libs/aioftp")
+BASE_COMMIT_SHA = "4c18bdec8eb8f2e12a226fc513d07b92b5014167"
+BASE_TREE_SHA = "cb1a5e0d57a012eea1e9bddbb9e1946d167c2f50"
+
+_explicit_dir = os.environ.get("PROJECT_DIR")
+if _explicit_dir:
+    PROJECT_DIR = _explicit_dir
+    PATCHES_DIR = os.environ.get("PATCHES_DIR", "/hud/patches")
+    _LOCAL_SUBSTRATE = False
+else:
+    _LOCAL_ROOT = Path(tempfile.gettempdir()) / f"hud-{FOLDER_NAME}"
+    PROJECT_DIR = str(_LOCAL_ROOT / f"{FOLDER_NAME}-{os.getpid()}")
+    PATCHES_DIR = str(_LOCAL_ROOT / f"patches-{os.getpid()}")
+    _LOCAL_SUBSTRATE = True
+
+env = Environment(name="aioftp-stream-abort-desynchronization")
+
+
+class UidWallWorkspace(Workspace):
+    """Workspace that drops the agent shell to UID 1000 (ubuntu) when running as root.
+
+    Maintains full signature compatibility with Workspace.shell_argv.
+    """
+
+    def shell_argv(self, command=None, *, cwd=None, env=None, tty: bool = False, **kwargs):
+        super_kwargs = {"cwd": cwd, "env": env, "tty": tty}
+        sig = inspect.signature(super().shell_argv)
+        if "kwargs" in sig.parameters:
+            super_kwargs.update(kwargs)
+        else:
+            for k, v in kwargs.items():
+                if k in sig.parameters:
+                    super_kwargs[k] = v
+        argv = super().shell_argv(command, **super_kwargs)
+        if sys.platform != "win32" and hasattr(os, "geteuid") and os.geteuid() == 0:
+            argv = ["setpriv", "--reuid", "1000", "--regid", "1000", "--clear-groups", "--", *argv]
+        return argv
+
+
+_ws = UidWallWorkspace(
+    PROJECT_DIR,
+    guest_path=PROJECT_DIR,
+    network=True,
+    env={"HOME": "/home/ubuntu"},
+)
+
+
+async def _git_step(args: list[str], cwd: str | None, label: str) -> None:
+    proc = await asyncio.create_subprocess_exec(
+        "git", *args,
+        cwd=cwd,
+        stdout=asyncio.subprocess.PIPE,
+        stderr=asyncio.subprocess.PIPE,
+    )
+    _, stderr = await proc.communicate()
+    if proc.returncode != 0:
+        raise RuntimeError(f"{label} failed: {stderr.decode(errors='replace')}")
+
+
+async def _ensure_repo() -> None:
+    """Local-substrate fallback: the same single-commit, credential-free fetch the image performs."""
+    if (Path(PROJECT_DIR) / ".git").exists():
+        return
+    Path(PROJECT_DIR).parent.mkdir(parents=True, exist_ok=True)
+    logger.info("Fetching %s@%s into %s", REPO_URL, BASE_COMMIT_SHA, PROJECT_DIR)
+    await _git_step(["init", "-q", PROJECT_DIR], None, "git init")
+    await _git_step(
+        ["-c", f"safe.directory={PROJECT_DIR}", "fetch", "-q", "--depth", "1", "--no-tags", REPO_URL, BASE_COMMIT_SHA],
+        PROJECT_DIR, f"git fetch {REPO_URL} {BASE_COMMIT_SHA}",
+    )
+    await _git_step(
+        ["-c", f"safe.directory={PROJECT_DIR}", "checkout", "-q", "--detach", BASE_COMMIT_SHA],
+        PROJECT_DIR, f"git checkout {BASE_COMMIT_SHA}",
+    )
+    (Path(PROJECT_DIR) / ".git" / "FETCH_HEAD").unlink(missing_ok=True)
+
+
+@env.initialize
+async def _up() -> None:
+    await _ensure_repo()
+    await _ws.start()
+    env.add_capability(_ws.capability("shell"))
+
+
+@env.shutdown
+async def _down() -> None:
+    await _ws.stop()
+    if _LOCAL_SUBSTRATE:
+        shutil.rmtree(PROJECT_DIR, ignore_errors=True)
+        shutil.rmtree(PATCHES_DIR, ignore_errors=True)
+
+
+def make_prompt(description: str) -> str:
+    return f"""You are working in a coding repository located at {PROJECT_DIR}.
+
+Use the tools provided to complete the following task:
+
+{description}
+"""
+
+
+# Grading entry point. Resolved lazily by ``run_grading``; never imported at module
+# import time. Prompt 05 implements ``grade_repository`` in ``grading/runner.py``.
+GRADER_ENTRY_POINT = ("grading.runner", "grade_repository")
+
+FAIL_CLOSED_LIST_KEYS = ("caps", "families", "subscores", "checks", "blockers")
+
+
+
+# --- Workspace snapshot ---------------------------------------------------------------------
+# Captured around the grader, never inside ``grading/**`` (so the grader scale id is untouched):
+# the agent's final workspace as a binary-safe git patch against the pinned base commit, attached
+# to the grade envelope's ``info["workspace_snapshot"]``. Prompt 08 decodes it beside the trace
+# export and Prompt 11 reconstructs the candidate from it instead of replaying the trajectory.
+#
+# Isolation: the env process runs as root against the root-owned ``.git`` through a scratch index
+# and a scratch object directory (the real object store is read-only through alternates), so the
+# agent's tree, the real index and the real objects are never written. Completeness: when the
+# agent edited an ignore file the capture re-adds with ``--force`` (complete beats compact), and
+# configured content filters are reported so diagnosis never trusts transformed bytes. Safety:
+# the capture runs after the grader, under one deadline, streams the diff to disk and reads it
+# only under the size cap; a capture failure degrades to ``unavailable`` and never touches the grade.
+
+WORKSPACE_SNAPSHOT_SCHEMA_VERSION = 2
+WORKSPACE_SNAPSHOT_MAX_BYTES_ENV = "HUD_WORKSPACE_SNAPSHOT_MAX_BYTES"  # raw patch bytes; 0 disables
+WORKSPACE_SNAPSHOT_DEADLINE_ENV = "HUD_WORKSPACE_SNAPSHOT_DEADLINE_SECONDS"  # whole capture
+WORKSPACE_SNAPSHOT_DEFAULT_MAX_BYTES = 4 * 1024 * 1024
+WORKSPACE_SNAPSHOT_DEFAULT_DEADLINE = 120.0
+WORKSPACE_SNAPSHOT_EXCLUDED_DIRS = ("node_modules", ".venv", ".hud", ".tmp", ".cache", "dist", "build", "cdk.out")
+# Prompt 11's forbidden path parts, refused here too so they never enter the grade event.
+WORKSPACE_SNAPSHOT_PROTECTED_NAMES = (".env", ".hud", "gold.diff", "grading", "mcp_server")
+WORKSPACE_SNAPSHOT_IGNORE_FILES = (".gitignore", ".gitattributes")
+WORKSPACE_SNAPSHOT_MAX_LISTED_FILES = 2000
+WORKSPACE_SNAPSHOT_GIT_TIMEOUT = 60
+WORKSPACE_SNAPSHOT_ENCODING = "gzip+base64"
+
+
+def _snapshot_limit(name: str, default, cast):
+    try:
+        return cast(os.environ.get(name, default))
+    except (TypeError, ValueError):
+        return cast(default)
+
+
+def _snapshot_payload(base_commit_sha: str, max_bytes: int | None, **fields) -> dict:
+    payload = {
+        "schema_version": WORKSPACE_SNAPSHOT_SCHEMA_VERSION,
+        "status": "unavailable",
+        "reason": None,
+        "base_commit_sha": base_commit_sha,
+        "head_commit_sha": None,
+        "patch_sha256": None,
+        "patch_size": None,
+        "changed_files": [],
+        "changed_file_count": None,
+        "changed_files_truncated": False,
+        "ignore_rules_modified": None,
+        "ignored_entry_count": None,
+        "capture_mode": None,
+        "filtered": None,
+        "content_filters": {"configured": [], "attributes_declared": False},
+        "captured_after_grading": True,
+        "excluded_dirs": list(WORKSPACE_SNAPSHOT_EXCLUDED_DIRS),
+        "protected_names": list(WORKSPACE_SNAPSHOT_PROTECTED_NAMES),
+        "encoding": WORKSPACE_SNAPSHOT_ENCODING,
+        "max_bytes": max_bytes,
+        "deadline_seconds": None,
+        "elapsed_seconds": None,
+        "captured_at_utc": datetime.datetime.now(datetime.UTC).strftime("%Y-%m-%dT%H:%M:%SZ"),
+    }
+    payload.update(fields)
+    return payload
+
+
+def _snapshot_pathspec_excludes() -> list[str]:
+    names = (*WORKSPACE_SNAPSHOT_EXCLUDED_DIRS, *WORKSPACE_SNAPSHOT_PROTECTED_NAMES)
+    specs: list[str] = []
+    for name in names:
+        specs.extend((f":(exclude){name}", f":(exclude,glob)**/{name}", f":(exclude,glob)**/{name}/**"))
+    return specs
+
+
+def capture_workspace_snapshot(
+    project_dir: str, base_commit_sha: str, max_bytes: int | None = None, deadline_seconds: float | None = None,
+) -> dict:
+    """The agent's final workspace as a ``git diff --binary`` against ``base_commit_sha``.
+
+    Status is ``captured`` (body in ``patch_b64``), ``empty`` (no change), ``too_large`` (hash and
+    listing only) or ``unavailable`` (reason only). Never raises.
+    """
+    if max_bytes is None:
+        max_bytes = _snapshot_limit(WORKSPACE_SNAPSHOT_MAX_BYTES_ENV, WORKSPACE_SNAPSHOT_DEFAULT_MAX_BYTES, int)
+    if deadline_seconds is None:
+        deadline_seconds = _snapshot_limit(WORKSPACE_SNAPSHOT_DEADLINE_ENV, WORKSPACE_SNAPSHOT_DEFAULT_DEADLINE, float)
+    started = time.monotonic()
+
+    def finish(payload: dict) -> dict:
+        payload["deadline_seconds"] = deadline_seconds
+        payload["elapsed_seconds"] = round(time.monotonic() - started, 3)
+        return payload
+
+    try:
+        project = Path(project_dir)
+        git_dir = project / ".git"
+        if not git_dir.exists():
+            return finish(_snapshot_payload(base_commit_sha, max_bytes, reason="not_a_git_repository"))
+        if max_bytes <= 0:
+            return finish(_snapshot_payload(base_commit_sha, max_bytes, reason="disabled"))
+        with tempfile.TemporaryDirectory(prefix="hud-workspace-snapshot-") as scratch_name:
+            scratch = Path(scratch_name)
+            objects = scratch / "objects"
+            objects.mkdir()
+            # Heavy runtime trees and protected names are excluded through a scratch excludes
+            # file, which layers under the repository's own ignore rules.
+            excludes_file = scratch / "excludes"
+            excludes_file.write_text(
+                "".join(f"{name}/\n" for name in WORKSPACE_SNAPSHOT_EXCLUDED_DIRS)
+                + "".join(f"{name}\n" for name in WORKSPACE_SNAPSHOT_PROTECTED_NAMES),
+                encoding="utf-8",
+            )
+            env = {
+                **os.environ,
+                "GIT_INDEX_FILE": str(scratch / "index"),
+                "GIT_DIR": str(git_dir),
+                "GIT_WORK_TREE": str(project),
+                "GIT_OBJECT_DIRECTORY": str(objects),
+                "GIT_ALTERNATE_OBJECT_DIRECTORIES": str(git_dir / "objects"),
+                "HOME": scratch_name,
+                "GIT_CONFIG_NOSYSTEM": "1",
+                "GIT_TERMINAL_PROMPT": "0",
+            }
+            base_argv = [
+                "git", "-c", f"safe.directory={project}", "-c", "core.quotePath=false",
+                "-c", f"core.excludesFile={excludes_file}",
+            ]
+
+            def git(*args: str, ok_codes: tuple = (0,), stdout=subprocess.PIPE) -> bytes:
+                remaining = deadline_seconds - (time.monotonic() - started)
+                if remaining <= 0:
+                    raise TimeoutError(f"snapshot deadline of {deadline_seconds:g}s exhausted before git {args[0]}")
+                completed = subprocess.run(
+                    [*base_argv, *args], cwd=str(project), env=env, stdout=stdout, stderr=subprocess.PIPE,
+                    check=False, timeout=min(WORKSPACE_SNAPSHOT_GIT_TIMEOUT, remaining),
+                )
+                if completed.returncode not in ok_codes:
+                    detail = completed.stderr.decode("utf-8", errors="replace").strip()[-200:]
+                    raise RuntimeError(f"git {args[0]} exited {completed.returncode}: {detail}")
+                return completed.stdout or b""
+
+            def name_status() -> list[dict]:
+                listing = git("diff", "--cached", "--name-status", "-z", "--no-renames", base_commit_sha, "--")
+                fields = [entry.decode("utf-8", errors="replace") for entry in listing.split(b"\0") if entry]
+                rows = [{"path": fields[i + 1], "status": fields[i][:1]} for i in range(0, len(fields) - 1, 2)]
+                rows.sort(key=lambda item: item["path"])
+                return rows
+
+            head = git("rev-parse", "HEAD").decode("utf-8", errors="replace").strip()
+            git("read-tree", base_commit_sha)
+            git("add", "-A", "--ignore-errors", "--", ".")
+            changed = name_status()
+            ignore_rules_modified = any(item["path"].rsplit("/", 1)[-1] in WORKSPACE_SNAPSHOT_IGNORE_FILES for item in changed)
+            capture_mode = "standard"
+            if ignore_rules_modified:
+                # Agent-edited ignore rules can hide files the grader saw: re-add everything the
+                # pathspec excludes do not name, ignore rules notwithstanding.
+                git("add", "-A", "--force", "--ignore-errors", "--", ".", *_snapshot_pathspec_excludes())
+                changed = name_status()
+                capture_mode = "forced"
+
+            config_out = git("config", "--get-regexp", r"^filter\..*\.(clean|smudge|process)$", ok_codes=(0, 1))
+            configured = sorted({line.split()[0] for line in config_out.decode("utf-8", errors="replace").splitlines() if line.strip()})
+            attribute_paths = [
+                entry.decode("utf-8", errors="replace")
+                for entry in git("ls-files", "-z", "--cached", "--", ".gitattributes", "*/.gitattributes").split(b"\0") if entry
+            ]
+            attributes_declared = False
+            for rel in [*attribute_paths, ".git/info/attributes"]:
+                candidate = project / rel
+                if candidate.is_file() and "filter=" in candidate.read_text(encoding="utf-8", errors="replace"):
+                    attributes_declared = True
+                    break
+            filtered = bool(configured) and attributes_declared
+
+            diff_path = scratch / "snapshot.diff"
+            with diff_path.open("wb") as handle:
+                git("diff", "--cached", "--binary", "--full-index", "--no-ext-diff", "--no-renames", base_commit_sha, "--", stdout=handle)
+            size = diff_path.stat().st_size
+            digest = hashlib.sha256()
+            with diff_path.open("rb") as handle:
+                while chunk := handle.read(1 << 20):
+                    digest.update(chunk)
+            ignored = git("ls-files", "-z", "--others", "--ignored", "--exclude-standard", "--directory")
+            patch = diff_path.read_bytes() if 0 < size <= max_bytes else b""
+
+        payload = _snapshot_payload(
+            base_commit_sha, max_bytes,
+            head_commit_sha=head,
+            patch_sha256=digest.hexdigest(),
+            patch_size=size,
+            changed_files=changed[:WORKSPACE_SNAPSHOT_MAX_LISTED_FILES],
+            changed_file_count=len(changed),
+            changed_files_truncated=len(changed) > WORKSPACE_SNAPSHOT_MAX_LISTED_FILES,
+            ignore_rules_modified=ignore_rules_modified,
+            ignored_entry_count=sum(1 for entry in ignored.split(b"\0") if entry),
+            capture_mode=capture_mode,
+            filtered=filtered,
+            content_filters={"configured": configured, "attributes_declared": attributes_declared},
+        )
+        if size == 0:
+            payload["status"] = "empty"
+        elif size > max_bytes:
+            payload.update(status="too_large", reason=f"patch of {size} bytes exceeds max_bytes {max_bytes}")
+        else:
+            payload.update(status="captured", patch_b64=base64.b64encode(gzip.compress(patch, 6)).decode("ascii"))
+        return finish(payload)
+    except Exception as exc:  # noqa: BLE001 - a snapshot failure must never change the grade
+        logger.warning("workspace snapshot unavailable: %s: %s", type(exc).__name__, exc)
+        return finish(_snapshot_payload(base_commit_sha, max_bytes, reason=f"{type(exc).__name__}: {str(exc)[:200]}"))
+
+
+def _attach_workspace_snapshot(result: EvaluationResult, snapshot: dict) -> EvaluationResult:
+    """Return ``result`` with ``info["workspace_snapshot"]`` set; the grader's own info is kept.
+
+    Runs outside the fail-closed grading guards, so it must never raise: on any failure the
+    grader's result is returned exactly as it was.
+    """
+    try:
+        existing = getattr(result, "info", None)
+        info = dict(existing) if isinstance(existing, dict) else {}
+        info["workspace_snapshot"] = snapshot
+        try:
+            return result.model_copy(update={"info": info})
+        except Exception:  # noqa: BLE001 - fall back to assignment on a non-pydantic result
+            result.info = info
+            return result
+    except Exception:  # noqa: BLE001 - the snapshot is diagnostic, never grade-bearing
+        logger.exception("could not attach the workspace snapshot to the grade result")
+        return result
+
+
+def fail_closed_payload(reason: str, detail: str = "", health: str = "fault:entry_point") -> dict:
+    """Structured, fail-closed grading result.
+
+    Emitted whenever the grader chain cannot be imported, resolved, or executed.
+    Every list-shaped key is present and explicitly empty so downstream readers
+    never have to distinguish "absent" from "empty". ``health`` is a non-healthy literal.
+    """
+    payload = {
+        "task_id": TASK_ID,
+        "reward": 0.0,
+        "is_error": True,
+        "grader_health": health,
+        "grader_health_reason": reason,
+    }
+    for key in FAIL_CLOSED_LIST_KEYS:
+        payload[key] = []
+    payload["blockers"] = [reason]
+    if detail:
+        payload["grader_health_detail"] = detail
+    return payload
+
+
+# --- Served grader scale ----------------------------------------------------------------------
+# The id of the scale a served grade is expressed on, computed exactly as
+# delivered_tasks/scripts/hud_task_evidence.compute_grader_scale_id does over the package: every
+# file under grading/ (adversarial fixtures and generated caches excluded) plus task/rubric.md.
+SERVED_ROOT = Path(__file__).resolve().parent
+SCALE_GENERATED_DIRECTORY_NAMES = {"__pycache__", ".pytest_cache", ".ruff_cache", ".venv"}
+
+
+def served_grader_scale_id(root: Path | None = None) -> str | None:
+    root = root or SERVED_ROOT
+    try:
+        digest = hashlib.sha256()
+        sources: list[tuple[str, Path]] = []
+        grading_dir = root / "grading"
+        if grading_dir.is_dir():
+            for path in sorted(grading_dir.rglob("*")):
+                relative = path.relative_to(root)
+                if not path.is_file() or any(
+                    "adversarial" in part or part in SCALE_GENERATED_DIRECTORY_NAMES for part in relative.parts
+                ):
+                    continue
+                sources.append((relative.as_posix(), path))
+        rubric = root / "task" / "rubric.md"
+        if rubric.is_file():
+            sources.append(("task/rubric.md", rubric))
+        for relative, path in sorted(sources):
+            digest.update(relative.encode("utf-8"))
+            digest.update(b"\0")
+            digest.update(path.read_bytes())
+            digest.update(b"\0")
+        return digest.hexdigest()
+    except OSError:
+        return None
+
+
+def _with_info(result: EvaluationResult, **fields) -> EvaluationResult:
+    """Return ``result`` with ``fields`` merged into its info; never raises."""
+    try:
+        existing = getattr(result, "info", None)
+        info = dict(existing) if isinstance(existing, dict) else {}
+        info.update(fields)
+        return result.model_copy(update={"info": info})
+    except Exception:  # noqa: BLE001 - diagnostic enrichment never changes the grade
+        logger.exception("could not enrich the grade result info")
+        return result
+
+
+def _resolve_grader():
+    """Import the grader chain lazily and return its callable entry point."""
+    module_name, attribute = GRADER_ENTRY_POINT
+    module = importlib.import_module(module_name)
+    grade = getattr(module, attribute)
+    if not callable(grade):
+        raise TypeError(f"{module_name}.{attribute} is not callable")
+    return grade
+
+
+def _evaluation_from_payload(payload: dict) -> EvaluationResult:
+    """Convert a fail-closed payload into an EvaluationResult carrying the complete diagnostics.
+
+    Faults report zero raw/final scores, an error flag, a non-healthy health literal and
+    explicit empty lists, so every reader can reject them as measurements.
+    """
+    info = {
+        "raw_score": 0.0,
+        "final_score": 0.0,
+        "caps": [],
+        "criteria": [],
+        "gates_fired": [],
+        "gate_statuses": {},
+        "grader_health": payload.get("grader_health") or "fault:entry_point",
+        "grader_health_reason": payload.get("grader_health_reason"),
+    }
+    return EvaluationResult(
+        reward=0.0,
+        done=True,
+        isError=True,
+        content=json.dumps(payload, indent=2, sort_keys=True, default=str),
+        info=info,
+        subscores=[],
+    )
+
+
+async def run_grading(**grader_kwargs) -> EvaluationResult:
+    """Run the grader chain, failing closed on any import or execution failure.
+
+    A missing or corrupted grader module must still produce a structured JSON
+    result with ``is_error`` true, ``reward`` 0.0, a non-healthy
+    ``grader_health``, and explicitly empty lists -- never an import-time crash
+    that leaves the container with no output.
+
+    After the grader has produced its result, the agent's final workspace is captured as a
+    git patch (``capture_workspace_snapshot``) and attached to that result's
+    ``info["workspace_snapshot"]``. The grader always runs first, so a slow or failing capture
+    can only degrade to ``unavailable``; it can neither delay nor change the grade.
+    """
+    outcome = await _run_grader_chain(**grader_kwargs)
+    project_dir = str(grader_kwargs.get("project_dir") or PROJECT_DIR)
+    try:
+        snapshot = await asyncio.to_thread(capture_workspace_snapshot, project_dir, BASE_COMMIT_SHA)
+    except Exception as exc:  # noqa: BLE001 - the snapshot is diagnostic, never grade-bearing
+        logger.exception("workspace snapshot capture raised; recording it unavailable")
+        snapshot = _snapshot_payload(BASE_COMMIT_SHA, None, reason=f"{type(exc).__name__}: {str(exc)[:200]}")
+    return _attach_workspace_snapshot(outcome, snapshot)
+
+
+async def _run_grader_chain(**grader_kwargs) -> EvaluationResult:
+    """The one shared grading function: run the grader, then attach the served scale id."""
+    outcome = await _grade_or_fail_closed(**grader_kwargs)
+    return _with_info(outcome, served_grader_scale_id=served_grader_scale_id())
+
+
+async def _grade_or_fail_closed(**grader_kwargs) -> EvaluationResult:
+    """Resolve and run the grader, failing closed on any import or execution failure."""
+    try:
+        grade = _resolve_grader()
+    except Exception as exc:  # noqa: BLE001 - fail closed on any resolution failure
+        reason = f"grader entry point {'.'.join(GRADER_ENTRY_POINT)} unavailable: {exc!r}"
+        logger.exception("Grader chain could not be resolved; failing closed")
+        return _evaluation_from_payload(fail_closed_payload(reason))
+
+    try:
+        result = grade(**grader_kwargs)
+        if inspect.isawaitable(result):
+            result = await result
+    except Exception as exc:  # noqa: BLE001 - fail closed on any grading failure
+        reason = f"grader execution failed: {exc!r}"
+        logger.exception("Grader chain raised; failing closed")
+        return _evaluation_from_payload(
+            fail_closed_payload(reason, detail=traceback.format_exc(limit=5))
+        )
+
+    if isinstance(result, EvaluationResult):
+        return result
+    if isinstance(result, dict):
+        return _evaluation_from_payload({**result, "is_error": True})
+    reason = f"grader returned unsupported result type {type(result).__name__}"
+    logger.error(reason)
+    return _evaluation_from_payload(fail_closed_payload(reason))
+
+
+TASK_PROMPT_PATH = Path(__file__).resolve().parent / "task" / "prompt.md"
+
+
+def last_resort_result(exc: BaseException) -> EvaluationResult:
+    """The template's final guard: a fail-closed grade naming only the exception class."""
+    payload = fail_closed_payload(
+        f"unhandled {type(exc).__name__} while grading", health="fault:unhandled"
+    )
+    return _with_info(_evaluation_from_payload(payload), served_grader_scale_id=served_grader_scale_id())
+
+
+async def _guarded(grading) -> EvaluationResult:
+    try:
+        return await grading
+    except Exception as exc:  # noqa: BLE001 - the last-resort path must always yield a grade
+        logger.exception("grading raised past its own guards; failing closed")
+        return last_resort_result(exc)
+
+
+@env.template(id="aioftp_stream_abort_desynchronization", description="Keep aioftp's FTP control connection in sync around transfer aborts.")
+async def aioftp_stream_abort_desynchronization(
+    task_id: str = "aioftp_stream_abort_desynchronization",
+    description: str | None = None,
+    **kwargs,
+):
+    """The agent task: serve the approved prompt, then grade the agent's workspace."""
+    if description is None:
+        description = TASK_PROMPT_PATH.read_text(encoding="utf-8")
+    _ = yield make_prompt(description)
+    yield await _guarded(run_grading(task_id=task_id, project_dir=PROJECT_DIR, **kwargs))
+
+
+# --- Reference variants ---------------------------------------------------------------------
+# The ec-swe-benchmarks PR check needs, in the linked taskset, a job whose name contains "gold"
+# with every trace at 1.0 and one whose name contains "base" with every trace at 0.0. A single-task
+# HUD job is named after its template id, so each variant is its own template (``<TASK_ID>_gold``,
+# ``<TASK_ID>_base``). Prompt 07 runs each once on every deployed build.
+#
+# Neither variant grades the agent's workspace. Only after the agent's final answer, a fresh tree is
+# cloned from the root-only repository metadata (``.git`` is root:root 0700 in the image, so the
+# agent cannot have changed it) into a root-only temporary directory, checked out at
+# ``BASE_COMMIT_SHA`` and verified against ``BASE_TREE_SHA``. For gold, ``/hud/gold.diff`` is applied
+# there, read in place and never copied. The tree is graded through the same grader entry point as
+# the agent task and then removed. A patch that does not apply, or any identity mismatch, fails
+# closed. No workspace snapshot is captured, so no reference content reaches the trace.
+
+GOLD_DIFF_PATH = "/hud/gold.diff"
+REFERENCE_KINDS = ("gold", "base")
+REFERENCE_GIT_TIMEOUT = 600
+REFERENCE_PROMPT = (
+    "This is a reference run of this environment. No change to the repository is needed, and no "
+    "files should be modified. Reply that you are done."
+)
+
+
+class ReferenceTreeError(RuntimeError):
+    """The reference tree could not be materialized exactly; the reference grade fails closed."""
+
+
+def _reference_git(*args: str, cwd: Path | None = None, safe: tuple = ()) -> str:
+    """Run one git command for the reference tree. Its output is never echoed: it may name gold paths."""
+    safe_flags = [flag for directory in safe for flag in ("-c", f"safe.directory={directory}")]
+    completed = subprocess.run(
+        ["git", *safe_flags, *args],
+        cwd=cwd,
+        capture_output=True,
+        text=True,
+        check=False,
+        timeout=REFERENCE_GIT_TIMEOUT,
+        env={**os.environ, "GIT_CONFIG_GLOBAL": os.devnull, "GIT_CONFIG_NOSYSTEM": "1", "GIT_TERMINAL_PROMPT": "0"},
+    )
+    if completed.returncode != 0:
+        raise ReferenceTreeError(f"git {args[0]} exited {completed.returncode}")
+    return completed.stdout
+
+
+def _require_root_only(path: Path, label: str) -> None:
+    """Refuse a path any other account can read, or one not owned by root when running as root."""
+    details = path.stat()
+    if stat.S_IMODE(details.st_mode) & 0o077:
+        raise ReferenceTreeError(f"{label} is accessible beyond its owner")
+    if hasattr(os, "geteuid") and os.geteuid() == 0 and details.st_uid != 0:
+        raise ReferenceTreeError(f"{label} is not owned by root")
+
+
+def materialize_reference_tree(reference: str) -> Path:
+    """Build the root-only reference tree for ``reference`` and return its temporary root.
+
+    The graded tree is ``<root>/tree``; the caller removes ``<root>``.
+    """
+    if reference not in REFERENCE_KINDS:
+        raise ReferenceTreeError(f"unknown reference kind {reference!r}")
+    git_dir = Path(PROJECT_DIR) / ".git"
+    if not git_dir.is_dir():
+        raise ReferenceTreeError("repository metadata is missing")
+    if not _LOCAL_SUBSTRATE:
+        _require_root_only(git_dir, "repository metadata")
+    root = Path(tempfile.mkdtemp(prefix="hud-reference-"))
+    try:
+        _require_root_only(root, "reference directory")
+        tree = root / "tree"
+        _reference_git(
+            "clone", "--no-hardlinks", "--no-checkout", "--quiet", str(git_dir), str(tree),
+            safe=(str(git_dir), str(PROJECT_DIR)),
+        )
+        _reference_git("checkout", "--quiet", "--detach", BASE_COMMIT_SHA, cwd=tree)
+        if _reference_git("rev-parse", "HEAD^{tree}", cwd=tree).strip() != BASE_TREE_SHA:
+            raise ReferenceTreeError("reference tree does not match the pinned base tree")
+        if _reference_git("status", "--porcelain", "--untracked-files=all", cwd=tree).strip():
+            raise ReferenceTreeError("reference tree is not a clean checkout of the pinned base")
+        if reference == "gold":
+            patch = Path(GOLD_DIFF_PATH)
+            if not patch.is_file():
+                raise ReferenceTreeError("the gold patch is unavailable")
+            _reference_git("apply", "--check", "--whitespace=nowarn", str(patch), cwd=tree)
+            _reference_git("apply", "--whitespace=nowarn", str(patch), cwd=tree)
+        return root
+    except BaseException:
+        shutil.rmtree(root, ignore_errors=True)
+        raise
+
+
+async def run_reference_grading(reference: str) -> EvaluationResult:
+    """Grade a fresh reference tree through the agent task's grader entry point, failing closed."""
+    try:
+        root = await asyncio.to_thread(materialize_reference_tree, reference)
+    except Exception as exc:  # noqa: BLE001 - any materialization failure fails closed
+        detail = str(exc) if isinstance(exc, ReferenceTreeError) else type(exc).__name__
+        logger.warning("reference %s tree unavailable: %s", reference, detail)
+        outcome = _with_info(
+            _evaluation_from_payload(
+                fail_closed_payload(f"reference_{reference}_unavailable: {detail}", health="fault:reference_tree")
+            ),
+            served_grader_scale_id=served_grader_scale_id(),
+        )
+    else:
+        try:
+            outcome = await _run_grader_chain(task_id=TASK_ID, project_dir=str(root / "tree"))
+        finally:
+            await asyncio.to_thread(shutil.rmtree, root, ignore_errors=True)
+    return _attach_workspace_snapshot(outcome, _snapshot_payload(BASE_COMMIT_SHA, None, reason="reference_run"))
+
+
+@env.template(id=f"{TASK_ID}_gold", description="Reference run graded on the pinned base with the gold patch.")
+async def reference_gold():
+    """Reference variant: grades the gold tree, never the agent's workspace."""
+    _ = yield REFERENCE_PROMPT
+    yield await _guarded(run_reference_grading("gold"))
+
+
+@env.template(id=f"{TASK_ID}_base", description="Reference run graded on the unchanged pinned base.")
+async def reference_base():
+    """Reference variant: grades the unchanged base tree, never the agent's workspace."""
+    _ = yield REFERENCE_PROMPT
+    yield await _guarded(run_reference_grading("base"))

harbor/sources/aioftp-v11/gold.diff

@@ -0,0 +1,795 @@
+diff --git a/src/aioftp/client.py b/src/aioftp/client.py
+index 9755aaf..ebb9c6e 100644
+--- a/src/aioftp/client.py
++++ b/src/aioftp/client.py
+@@ -97,6 +97,7 @@ class DataConnectionThrottleStreamIO(ThrottleStreamIO):
+             write_timeout=write_timeout,
+         )
+         self.client = client
++        self.final_reply_consumed = False
+ 
+     async def finish(
+         self,
+@@ -118,7 +119,18 @@ class DataConnectionThrottleStreamIO(ThrottleStreamIO):
+             :py:class:`str`
+         """
+         self.close()
+-        await self.client.command(None, expected_codes, wait_codes)
++        async with self.client._control_lock():
++            if self.final_reply_consumed:
++                # abort already read the transfer's final reply
++                return
++            # from here the reply is registered as owed before anything awaits, so a
++            # cancelled finish leaves it to be read by the next command instead of lost
++            self.final_reply_consumed = True
++            try:
++                await self.client._exchange(None, expected_codes, wait_codes, claim=self)
++            finally:
++                if self.client._transfer_stream is self:
++                    self.client._transfer_stream = None
+ 
+     async def __aexit__(
+         self,
+@@ -128,6 +140,15 @@ class DataConnectionThrottleStreamIO(ThrottleStreamIO):
+     ) -> None:
+         if exc is None:
+             await self.finish()
++        elif self.client._transfer_stream is self:
++            try:
++                await asyncio.shield(self.client.abort())
++            except Exception:
++                # the exception raised inside the block is the one that propagates: an abort
++                # that failed (an unexpected ABOR reply, a lost connection) must not replace it
++                logger.debug("abort after an exception in the stream block failed", exc_info=True)
++            finally:
++                self.close()
+         else:
+             self.close()
+ 
+@@ -214,6 +235,14 @@ class BaseClient:
+         self.path_io = path_io_factory(timeout=path_timeout)
+         self.encoding = encoding
+         self._stream: ThrottleStreamIO | None = None
++        self._transfer_stream: DataConnectionThrottleStreamIO | None = None
++        self._control_lock_instance: asyncio.Lock | None = None
++        self._write_lock_instance: asyncio.Lock | None = None
++        # replies written for but not yet read from the control connection, oldest first:
++        # (expected codes, wait codes, owner) where owner is the stream entitled to read it
++        self._owed_replies: collections.deque[tuple[tuple[Code, ...], tuple[Code, ...], object | None]] = (
++            collections.deque()
++        )
+         self.ssl = ssl
+         self.parse_list_line_custom = parse_list_line_custom
+         self.parse_list_line_custom_first = parse_list_line_custom_first
+@@ -252,6 +281,8 @@ class BaseClient:
+             throttles={"_": self.throttle},
+             timeout=self.socket_timeout,
+         )
++        self._transfer_stream = None
++        self._owed_replies.clear()
+ 
+     @property
+     def stream(self) -> ThrottleStreamIO:
+@@ -265,6 +296,20 @@ class BaseClient:
+         """
+         if self._stream is not None:
+             self._stream.close()
++        self._transfer_stream = None
++        self._owed_replies.clear()
++
++    def _control_lock(self) -> asyncio.Lock:
++        """Serializes readers of the control connection."""
++        if self._control_lock_instance is None:
++            self._control_lock_instance = asyncio.Lock()
++        return self._control_lock_instance
++
++    def _write_lock(self) -> asyncio.Lock:
++        """Keeps the ledger of owed replies in wire order across concurrent writers."""
++        if self._write_lock_instance is None:
++            self._write_lock_instance = asyncio.Lock()
++        return self._write_lock_instance
+ 
+     async def parse_line(self) -> tuple[Code, str]:
+         """
+@@ -381,28 +426,210 @@ class BaseClient:
+             when logging
+         :type censor_after: :py:class:`None` or :py:class:`int`
+         """
++        async with self._control_lock():
++            if command and command.split(None, 1)[0].upper() != "ABOR":
++                # a transfer nobody finished or aborted must not leak its reply into this exchange
++                await self._settle_transfer()
++            return await self._exchange(command, expected_codes, wait_codes, censor_after)
++
++    async def _exchange(
++        self,
++        command: str | None,
++        expected_codes: str | tuple[str, ...] = (),
++        wait_codes: str | tuple[str, ...] = (),
++        censor_after: int | None = None,
++        *,
++        claim: object | None = None,
++    ) -> tuple[Code, list[str]] | None:
++        """
++        :py:func:`asyncio.coroutine`
++
++        One control exchange; the caller holds the control lock. Replies owed
++        from earlier exchanges are consumed before this one is read. `claim`
++        names a stream whose reply may already be owed (a transfer aborted
++        with `wait=False`); that reply is then read here instead of drained.
++        """
+         expected_codes = wrap_into_codes(wrap_with_container(expected_codes))
+         wait_codes = wrap_into_codes(wrap_with_container(wait_codes))
+-        if command:
+-            if "\r" in command or "\n" in command:
+-                raise errors.InvalidCommand("Command must not contain CR/LF")
+-            if censor_after:
+-                # Censor the user's command
+-                raw = command[:censor_after]
+-                stars = "*" * len(command[censor_after:])
+-                logger.debug("%s%s", raw, stars)
++        entry = None
++        if claim is not None:
++            entry = next((owed for owed in self._owed_replies if owed[2] is claim), None)
++        if entry is None and (expected_codes or wait_codes):
++            entry = (expected_codes, wait_codes, claim)
++            if command:
++                # nothing is written while replies are still owed: no pipelining behind them
++                await self._drain_owed_replies()
++                await self._submit(command, [entry], censor_after)
+             else:
+-                logger.debug(command)
+-            message = command + END_OF_LINE
+-            await self.stream.write(message.encode(encoding=self.encoding))
+-        if expected_codes or wait_codes:
++                # the reply is already in flight: owed from now on, even if we are cancelled
++                self._owed_replies.append(entry)
++        elif command:
++            await self._drain_owed_replies()
++            await self._submit(command, [], censor_after)
++        if entry is None:
++            return None
++        # another writer (an abort from a second task) may have slipped in ahead of us
++        await self._drain_owed_replies(until=entry)
++        code, info = await self._collect_reply()
++        if expected_codes:
++            self.check_codes(expected_codes, code, info)
++        return code, info
++
++    async def _submit(
++        self,
++        command: str,
++        entries: list[tuple[tuple[Code, ...], tuple[Code, ...], object | None]],
++        censor_after: int | None = None,
++    ) -> None:
++        """
++        :py:func:`asyncio.coroutine`
++
++        Register the replies `command` will earn and write it, as one step
++        with respect to other writers, so the ledger matches the wire order.
++        A failed write owes nothing.
++        """
++        async with self._write_lock():
++            self._owed_replies.extend(entries)
++            try:
++                await self._write_command(command, censor_after)
++            except BaseException:
++                for entry in reversed(entries):
++                    if self._owed_replies and self._owed_replies[-1] is entry:
++                        self._owed_replies.pop()
++                raise
++
++    async def _write_command(self, command: str, censor_after: int | None = None) -> None:
++        if "\r" in command or "\n" in command:
++            raise errors.InvalidCommand("Command must not contain CR/LF")
++        if censor_after:
++            # Censor the user's command
++            raw = command[:censor_after]
++            stars = "*" * len(command[censor_after:])
++            logger.debug("%s%s", raw, stars)
++        else:
++            logger.debug(command)
++        message = command + END_OF_LINE
++        await self.stream.write(message.encode(encoding=self.encoding))
++
++    async def _collect_reply(self) -> tuple[Code, list[str]]:
++        """
++        :py:func:`asyncio.coroutine`
++
++        Read the reply of the oldest owed exchange. The exchange leaves the
++        ledger only once its reply has actually been read, so a cancelled
++        reader leaves the reply owed rather than lost.
++        """
++        _, wait_codes, _ = self._owed_replies[0]
++        code, info = await self.parse_response()
++        while any(map(code.matches, wait_codes)):
+             code, info = await self.parse_response()
+-            while any(map(code.matches, wait_codes)):
+-                code, info = await self.parse_response()
+-            if expected_codes:
+-                self.check_codes(expected_codes, code, info)
+-            return code, info
+-        return None
++        self._owed_replies.popleft()
++        return code, info
++
++    async def _drain_owed_replies(self, until: object | None = None) -> None:
++        while self._owed_replies and self._owed_replies[0] is not until:
++            await self._collect_reply()
++
++    async def _settle_transfer(self) -> None:
++        """
++        :py:func:`asyncio.coroutine`
++
++        Abort a transfer that is still outstanding on the control connection
++        (a stream that was neither finished nor aborted) before a new command
++        is written; the caller holds the control lock.
++        """
++        transfer = self._transfer_stream
++        if transfer is None:
++            return
++        if transfer.final_reply_consumed:
++            self._transfer_stream = None
++            return
++        await self._abort_transfer(transfer, wait=True, locked=True)
++
++    async def _abort_transfer(
++        self,
++        transfer: DataConnectionThrottleStreamIO | None,
++        *,
++        wait: bool,
++        locked: bool = False,
++    ) -> None:
++        """
++        :py:func:`asyncio.coroutine`
++
++        Send `ABOR` and, with `wait`, read its reply. When a transfer is
++        outstanding its final reply comes first, whatever its code; it is
++        registered as owed together with the abort reply before `ABOR` is
++        written, so both are still read later if this coroutine is cancelled
++        or `wait` is false. With `wait=False`, or while `finish()` is already
++        reading it, the transfer's final reply stays with the stream and
++        `finish()` checks it as usual. `ABOR` is written without the control
++        lock, so it can interrupt a `finish()` that is waiting on the server;
++        `locked` says the caller already holds the control lock for reading.
++        """
++        token = object()
++        entries: list[tuple[tuple[Code, ...], tuple[Code, ...], object | None]] = []
++        claimed = False
++        if transfer is not None:
++            registered = any(owed[2] is transfer for owed in self._owed_replies)
++            if not transfer.final_reply_consumed and not registered:
++                if wait:
++                    transfer.final_reply_consumed = True
++                    claimed = True
++                entries.append(((), wrap_into_codes(("1xx",)), None if wait else transfer))
++            if self._transfer_stream is transfer:
++                self._transfer_stream = None
++        abort_codes = wrap_into_codes(("225", "226"))
++        entries.append((abort_codes, wrap_into_codes(("1xx", "426")), token))
++        try:
++            await self._submit("ABOR", entries)
++        except BaseException:
++            if transfer is not None:
++                if claimed:
++                    transfer.final_reply_consumed = False
++                self._transfer_stream = transfer
++            raise
++        if not wait:
++            return
++        try:
++            if locked:
++                await self._read_owed(token, abort_codes)
++            else:
++                async with self._control_lock():
++                    await self._read_owed(token, abort_codes)
++        finally:
++            if transfer is not None:
++                transfer.close()
++
++    async def _read_owed(self, owner: object, expected_codes: tuple[Code, ...]) -> tuple[Code, list[str]]:
++        """
++        :py:func:`asyncio.coroutine`
++
++        Read everything owed before `owner`'s reply, then that reply; the
++        caller holds the control lock.
++        """
++        entry = next(owed for owed in self._owed_replies if owed[2] is owner)
++        await self._drain_owed_replies(until=entry)
++        code, info = await self._collect_reply()
++        self.check_codes(expected_codes, code, info)
++        return code, info
++
++    async def abort(self, *, wait: bool = True) -> None:
++        """
++        :py:func:`asyncio.coroutine`
++
++        Request data transfer abort.
++
++        If a transfer is in progress, its final reply (usually 426, or the
++        transfer's own completion reply if it finished first) is read before
++        the abort reply (225 or 226), so the control connection stays in sync.
++        With `wait=False` the abort reply is read by the next command instead.
++        `abort()` may be called from another task while the stream is being
++        finished; the calls serialize on the control connection.
++
++        :param wait: wait for abort response [426]→226 if `True`
++        :type wait: :py:class:`bool`
++        """
++        await self._abort_transfer(self._transfer_stream, wait=wait)
+ 
+     @staticmethod
+     def parse_epsv_response(s: str) -> tuple[None, int]:
+@@ -970,6 +1197,21 @@ class Client(BaseClient):
+                 )
+                 return cls
+ 
++            async def aclose(cls) -> None:
++                """
++                Stop listing early: abort the listing still in progress so the
++                control connection is ready for the next command.
++                """
++                stream, cls.stream = cls.stream, None
++                if hasattr(cls, "directories"):
++                    cls.directories.clear()
++                if stream is None:
++                    return
++                if self._transfer_stream is stream and not stream.final_reply_consumed:
++                    await self.abort()
++                else:
++                    stream.close()
++
+             async def __anext__(cls) -> PathWithInfo:
+                 if cls.stream is None:
+                     cls.stream = await cls._new_stream(path)
+@@ -1405,7 +1647,7 @@ class Client(BaseClient):
+         reader, writer = await self.get_passive_connection(conn_type)
+         if offset:
+             await self.command("REST " + str(offset), "350")
+-        await self.command(command, expected_codes, wait_codes, censor_after)
++        response = await self.command(command, expected_codes, wait_codes, censor_after)
+         stream = DataConnectionThrottleStreamIO(
+             self,
+             reader,
+@@ -1413,22 +1655,11 @@ class Client(BaseClient):
+             throttles={"_": self.throttle},
+             timeout=self.socket_timeout,
+         )
++        if command and (response is None or response[0].matches("1xx")):
++            # the transfer's final reply is still to come on the control connection
++            self._transfer_stream = stream
+         return stream
+ 
+-    async def abort(self, *, wait: bool = True) -> None:
+-        """
+-        :py:func:`asyncio.coroutine`
+-
+-        Request data transfer abort.
+-
+-        :param wait: wait for abort response [426]→226 if `True`
+-        :type wait: :py:class:`bool`
+-        """
+-        if wait:
+-            await self.command("ABOR", "226", "426")
+-        else:
+-            await self.command("ABOR")
+-
+     @classmethod
+     @contextlib.asynccontextmanager
+     async def context(
+diff --git a/src/aioftp/server.py b/src/aioftp/server.py
+index 076e702..defcd66 100644
+--- a/src/aioftp/server.py
++++ b/src/aioftp/server.py
+@@ -584,8 +584,21 @@ def worker(
+         try:
+             await f(cls, connection, rest)
+         except asyncio.CancelledError:
++            # an aborted transfer gives up its data path: whatever connects for it
++            # later must not be picked up by the next transfer
++            if connection.future.data_connection.done():
++                connection.data_connection.close()
++                del connection.data_connection
++            cls._release_passive_server(connection)
+             connection.response("426", "transfer aborted")
+-            connection.response("226", "abort successful")
++            if asyncio.current_task() in connection.aborting_workers:
++                connection.response("226", "abort successful")
++        except ConnectionError:
++            if connection.future.data_connection.done():
++                connection.data_connection.close()
++                del connection.data_connection
++            cls._release_passive_server(connection)
++            connection.response("426", "Connection closed; transfer aborted")
+ 
+     return wrapper  # type: ignore[return-value]
+ 
+@@ -902,6 +915,31 @@ class Server:
+                 await write(code + "-" + line)
+             await write(code + " " + tail)
+ 
++    @staticmethod
++    def _strip_telnet(b: bytes) -> bytes:
++        out = bytearray()
++        i = 0
++        n = len(b)
++        while i < n:
++            if b[i] == 0xFF:
++                if i + 1 < n:
++                    cmd = b[i + 1]
++                    if cmd == 0xFF:
++                        out.append(0xFF)
++                        i += 2
++                        continue
++                    elif 0xF0 <= cmd <= 0xFA:
++                        i += 2
++                        continue
++                    elif 0xFB <= cmd <= 0xFE:
++                        i += 3
++                        continue
++                i += 1
++                continue
++            out.append(b[i])
++            i += 1
++        return bytes(out)
++
+     async def parse_command(self, stream: StreamIO, censor_commands: tuple[str] = ("pass",)) -> tuple[str, str]:
+         """
+         :py:func:`asyncio.coroutine`
+@@ -920,7 +958,11 @@ class Server:
+         line = await stream.readline()
+         if not line:
+             raise ConnectionResetError
+-        s = line.decode(encoding=self.encoding).rstrip()
++        line = self._strip_telnet(line)
++        try:
++            s = line.decode(encoding=self.encoding).rstrip()
++        except UnicodeDecodeError:
++            return "undecodable", ""
+         cmd, _, rest = s.partition(" ")
+ 
+         if cmd.lower() in censor_commands:
+@@ -965,6 +1007,14 @@ class Server:
+         host, port, *_ = writer.transport.get_extra_info("peername", ("", ""))
+         current_server_host, *_ = writer.transport.get_extra_info("sockname")
+         logger.info("new connection from %s:%s", host, port)
++
++        sock = writer.transport.get_extra_info("socket")
++        if sock is not None:
++            try:
++                sock.setsockopt(socket.SOL_SOCKET, socket.SO_OOBINLINE, 1)
++            except Exception:
++                pass
++
+         key = stream = ThrottleStreamIO(
+             reader,
+             writer,
+@@ -990,6 +1040,8 @@ class Server:
+             path_io_factory=self.path_io_factory,
+             path_timeout=self.path_timeout,
+             extra_workers=set(),
++            aborting_workers=set(),
++            transferring_workers=set(),
+             response=lambda *args: response_queue.put_nowait(args),
+             acquired=False,
+             restart_offset=0,
+@@ -999,46 +1051,119 @@ class Server:
+             timeout=self.path_timeout,
+             connection=connection,
+         )
+-        pending = {
+-            asyncio.create_task(self.greeting(connection, "")),
+-            asyncio.create_task(self.response_writer(stream, response_queue)),
+-            asyncio.create_task(self.parse_command(stream)),
+-        }
+         self.connections[key] = connection
++
++        response_writer_task = asyncio.create_task(self.response_writer(stream, response_queue))
++        pending_cleanup = {response_writer_task}
++        command_queue: asyncio.Queue[tuple[str, str]] = asyncio.Queue()
++
++        async def command_reader() -> None:
++            while True:
++                try:
++                    res = await self.parse_command(stream)
++                except ConnectionResetError:
++                    await command_queue.put(("eof", ""))
++                    return
++                except Exception as exc:
++                    await command_queue.put(("error", str(exc)))
++                    return
++                await command_queue.put(res)
++
+         try:
++            greeting_ok = await self.greeting(connection, "")
++            if not greeting_ok:
++                await response_queue.join()
++                return
++
++            reader_task = asyncio.create_task(command_reader())
++            pending_cleanup.add(reader_task)
++            queue_get_task = asyncio.create_task(command_queue.get())
++            pending_cleanup.add(queue_get_task)
++
+             while True:
+-                done, pending = await asyncio.wait(
+-                    pending | connection.extra_workers,
++                wait_set = {queue_get_task, reader_task, response_writer_task} | connection.extra_workers
++                done, _ = await asyncio.wait(
++                    wait_set,
+                     return_when=asyncio.FIRST_COMPLETED,
+                 )
+-                connection.extra_workers -= done
+-                for task in done:
++
++                for worker_task in done & connection.extra_workers:
++                    connection.extra_workers.remove(worker_task)
+                     try:
+-                        result = task.result()
++                        worker_task.result()
+                     except errors.PathIOError:
+                         connection.response("451", "file system error")
++                    except asyncio.CancelledError:
++                        pass
++
++                if response_writer_task in done:
++                    response_writer_task.result()
++                    return
++
++                if reader_task in done and not reader_task.cancelled():
++                    exc = reader_task.exception()
++                    if exc is not None:
++                        raise exc
++
++                if queue_get_task in done:
++                    cmd, rest = queue_get_task.result()
++                    pending_cleanup.remove(queue_get_task)
++                    queue_get_task = asyncio.create_task(command_queue.get())
++                    pending_cleanup.add(queue_get_task)
++
++                    if cmd == "eof":
++                        return
++                    if cmd == "error":
++                        return
++
++                    if cmd == "undecodable":
++                        connection.response("500", "syntax error, command unrecognized")
+                         continue
+-                    # this is "command" result
+-                    if isinstance(result, bool):
+-                        if not result:
+-                            await response_queue.join()
+-                            return
+-                    # this is parse_command result
+-                    elif isinstance(result, tuple):
+-                        pending.add(
+-                            asyncio.create_task(self.parse_command(stream)),
+-                        )
+-                        cmd, rest = result
+-                        f = self.commands_mapping.get(cmd)
+-                        if f is not None:
+-                            pending.add(
+-                                asyncio.create_task(f(connection, rest)),  # type: ignore[arg-type]
++
++                    f = self.commands_mapping.get(cmd)
++                    if f is None:
++                        connection.response("502", f"{cmd!r} not implemented")
++                        if cmd not in ("retr", "stor", "appe", "rest"):
++                            connection.restart_offset = 0
++                        continue
++
++                    cmd_task = asyncio.create_task(f(connection, rest))
++                    pending_cleanup.add(cmd_task)
++                    try:
++                        while not cmd_task.done():
++                            wait_cmd_set = {cmd_task, response_writer_task} | connection.extra_workers
++                            cmd_done, _ = await asyncio.wait(
++                                wait_cmd_set,
++                                return_when=asyncio.FIRST_COMPLETED,
+                             )
+-                            if cmd not in ("retr", "stor", "appe"):
+-                                connection.restart_offset = 0
+-                        else:
+-                            message = f"{cmd!r} not implemented"
+-                            connection.response("502", message)
++                            for worker_task in cmd_done & connection.extra_workers:
++                                connection.extra_workers.remove(worker_task)
++                                try:
++                                    worker_task.result()
++                                except errors.PathIOError:
++                                    connection.response("451", "file system error")
++                                except asyncio.CancelledError:
++                                    pass
++
++                            if response_writer_task in cmd_done:
++                                response_writer_task.result()
++                                return
++
++                        result = cmd_task.result()
++                    except errors.PathIOError:
++                        connection.response("451", "file system error")
++                        result = True
++                    finally:
++                        pending_cleanup.discard(cmd_task)
++
++                    if cmd not in ("retr", "stor", "appe", "rest"):
++                        # a transfer's worker still has to read the REST offset
++                        connection.restart_offset = 0
++
++                    if isinstance(result, bool) and not result:
++                        await response_queue.join()
++                        return
++
+         except asyncio.CancelledError:
+             raise
+         except Exception:
+@@ -1047,7 +1172,7 @@ class Server:
+             logger.info("closing connection from %s:%s", host, port)
+             tasks_to_wait = []
+             if not asyncio.get_running_loop().is_closed():
+-                for task in pending | connection.extra_workers:
++                for task in pending_cleanup | connection.extra_workers:
+                     task.cancel()
+                     tasks_to_wait.append(task)
+                 if connection.future.passive_server.done():
+@@ -1065,7 +1190,7 @@ class Server:
+                     self.user_manager.notify_logout(connection.user),
+                 )
+                 tasks_to_wait.append(task)
+-            self.connections.pop(key)
++            self.connections.pop(key, None)
+             if tasks_to_wait:
+                 await asyncio.wait(tasks_to_wait)
+ 
+@@ -1162,6 +1287,22 @@ class Server:
+         return True
+ 
+     async def quit(self, connection: Connection, rest: str | PurePosixPath) -> bool:
++        # a transfer that owns its data connection finishes and replies first; one still waiting
++        # for its data connection is aborted with 426
++        waiting = []
++        open_workers = []
++        for w in list(connection.extra_workers):
++            if not w.done():
++                if w in connection.transferring_workers:
++                    open_workers.append(w)
++                else:
++                    waiting.append(w)
++        for w in waiting:
++            w.cancel()
++        if waiting:
++            await asyncio.wait(waiting)
++        if open_workers:
++            await asyncio.wait(open_workers)
+         connection.response("221", "bye")
+         return False
+ 
+@@ -1245,16 +1386,17 @@ class Server:
+     @PathConditions(PathConditions.path_must_exists)
+     @PathPermissions(PathPermissions.readable)
+     async def mlsd(self, connection: Connection, rest: str | PurePosixPath) -> bool:
++        @worker
+         @ConnectionConditions(
+             ConnectionConditions.data_connection_made,
+             wait=True,
+             fail_code="425",
+             fail_info="Can't open data connection",
+         )
+-        @worker
+         async def mlsd_worker(self: "Server", connection: Connection, rest: str | PurePosixPath) -> bool:
+             stream = connection.data_connection
+             del connection.data_connection
++            connection.transferring_workers.add(asyncio.current_task())
+             async with stream:
+                 async for path in connection.path_io.list(real_path):
+                     s = await self.build_mlsx_string(connection, path)
+@@ -1304,16 +1446,17 @@ class Server:
+     @PathConditions(PathConditions.path_must_exists)
+     @PathPermissions(PathPermissions.readable)
+     async def list(self, connection: Connection, rest: str | PurePosixPath) -> bool:
++        @worker
+         @ConnectionConditions(
+             ConnectionConditions.data_connection_made,
+             wait=True,
+             fail_code="425",
+             fail_info="Can't open data connection",
+         )
+-        @worker
+         async def list_worker(self: "Server", connection: Connection, rest: str | PurePosixPath) -> bool:
+             stream = connection.data_connection
+             del connection.data_connection
++            connection.transferring_workers.add(asyncio.current_task())
+             async with stream:
+                 async for path in connection.path_io.list(real_path):
+                     if not (await connection.path_io.exists(path)):
+@@ -1382,16 +1525,17 @@ class Server:
+     )
+     @PathPermissions(PathPermissions.writable)
+     async def stor(self, connection: Connection, rest: str | PurePosixPath, mode: str = "wb") -> bool:
++        @worker
+         @ConnectionConditions(
+             ConnectionConditions.data_connection_made,
+             wait=True,
+             fail_code="425",
+             fail_info="Can't open data connection",
+         )
+-        @worker
+         async def stor_worker(self: "Server", connection: Connection, rest: str | PurePosixPath) -> bool:
+             stream = connection.data_connection
+             del connection.data_connection
++            connection.transferring_workers.add(asyncio.current_task())
+             if connection.restart_offset:
+                 file_mode = "r+b"
+             else:
+@@ -1426,16 +1570,17 @@ class Server:
+     )
+     @PathPermissions(PathPermissions.readable)
+     async def retr(self, connection: Connection, rest: str | PurePosixPath) -> bool:
++        @worker
+         @ConnectionConditions(
+             ConnectionConditions.data_connection_made,
+             wait=True,
+             fail_code="425",
+             fail_info="Can't open data connection",
+         )
+-        @worker
+         async def retr_worker(self: "Server", connection: Connection, rest: str | PurePosixPath) -> bool:
+             stream = connection.data_connection
+             del connection.data_connection
++            connection.transferring_workers.add(asyncio.current_task())
+             file_in = connection.path_io.open(real_path, mode="rb")
+             async with file_in, stream:
+                 if connection.restart_offset:
+@@ -1514,10 +1659,18 @@ class Server:
+             )
+         return passive_server
+ 
++    def _release_passive_server(self, connection: Connection) -> None:
++        if not connection.future.passive_server.done():
++            return
++        connection.passive_server.close()
++        if self.available_data_ports is not None:
++            self.available_data_ports.put_nowait((0, connection.passive_server_port))
++        del connection.passive_server
++
+     @ConnectionConditions(ConnectionConditions.login_required)
+     async def pasv(self, connection: Connection, rest: str | PurePosixPath) -> bool:
+         async def handler(reader: asyncio.StreamReader, writer: asyncio.StreamWriter) -> None:
+-            if connection.future.data_connection.done():
++            if connection.future.data_connection.done() or not connection.future.passive_server.done():
+                 writer.close()
+             else:
+                 connection.data_connection = ThrottleStreamIO(
+@@ -1562,7 +1715,7 @@ class Server:
+     @ConnectionConditions(ConnectionConditions.login_required)
+     async def epsv(self, connection: Connection, rest: str | PurePosixPath) -> bool:
+         async def handler(reader: asyncio.StreamReader, writer: asyncio.StreamWriter) -> None:
+-            if connection.future.data_connection.done():
++            if connection.future.data_connection.done() or not connection.future.passive_server.done():
+                 writer.close()
+             else:
+                 connection.data_connection = ThrottleStreamIO(
+@@ -1601,11 +1754,19 @@ class Server:
+ 
+     @ConnectionConditions(ConnectionConditions.login_required)
+     async def abor(self, connection: Connection, rest: str | PurePosixPath) -> bool:
+-        if connection.extra_workers:
+-            for worker in connection.extra_workers:
++        connection.aborting_workers.intersection_update(connection.extra_workers)
++        active = [worker for worker in connection.extra_workers if not worker.done()]
++        fresh = [worker for worker in active if worker not in connection.aborting_workers]
++        if fresh:
++            # the cancelled workers answer 426 then 226 themselves
++            for worker in fresh:
++                connection.aborting_workers.add(worker)
+                 worker.cancel()
+-        else:
+-            connection.response("226", "nothing to abort")
++            return True
++        if active:
++            # a repeated ABOR is answered after the abort it repeats, never before
++            await asyncio.wait(active)
++        connection.response("226", "nothing to abort")
+         return True
+ 
+     async def appe(self, connection: Connection, rest: str | PurePosixPath) -> bool:

harbor/sources/aioftp-v11/grading/__init__.py

@@ -0,0 +1,18 @@
+"""Grading modules for aioftp_stream_abort_desynchronization.
+
+Submodules are exposed lazily (PEP 562). Importing this package must never pull
+in the grader chain, so that a missing or corrupted grader submodule surfaces as
+a fail-closed result from ``env.run_grading`` instead of an import-time crash
+that leaves the container with no JSON output.
+"""
+
+__all__ = ["GradingRunner"]
+
+
+def __getattr__(name: str):
+    """Import submodule attributes on first access, not at package import."""
+    if name == "GradingRunner":
+        from .runner import GradingRunner
+
+        return GradingRunner
+    raise AttributeError(f"module {__name__!r} has no attribute {name!r}")

harbor/sources/aioftp-v11/grading/adversarial-manifest.json

@@ -0,0 +1,2082 @@
+{
+  "schema_version": 7,
+  "task_id": "aioftp_stream_abort_desynchronization",
+  "gold_diff_sha256": "90b71765ae5d945b112d042be7044a1f152625dcdb70a14450d0ef32ec1036d3",
+  "comment_syntax": [
+    "#"
+  ],
+  "native_commands": {
+    "test": {
+      "argv": [
+        "/opt/aioftp-venv/bin/python",
+        "-I",
+        "-m",
+        "pytest",
+        "-p",
+        "no:cacheprovider",
+        "-o",
+        "pythonpath=src"
+      ],
+      "timeout_seconds": 600,
+      "measured_runtime_seconds": {
+        "gold": 19.733,
+        "base": 19.029
+      },
+      "may_cap": true,
+      "requires": [
+        "repo_python"
+      ],
+      "description": "The repository's whole pytest suite from the repository root with its own [tool.pytest.ini_options] (792 node IDs: -x, --durations 10, -p no:anyio, --cov, --import-mode=importlib), invoked directly through the repository interpreter. The grader runs the same argv on a tree whose tests/ and pyproject.toml are restored from grading/baseline; a failure lands on the 0.30 regression ceiling. Where IPv6 loopback is unavailable (the HUD sandbox), the image's aioftp_ipv6_loopback pytest plugin skips exactly the 212 ::1 node IDs and the grader requires that skip count; where ::1 binds, it requires zero skips."
+    }
+  },
+  "runtime_requirements": {
+    "executables": [
+      {
+        "id": "repo_python",
+        "executable": "/opt/aioftp-venv/bin/python",
+        "role": "interpreter",
+        "discovery": "dynamic",
+        "version_probe": {
+          "argv": [
+            "/opt/aioftp-venv/bin/python",
+            "--version"
+          ],
+          "regex": "Python (\\d+\\.\\d+\\.\\d+)"
+        },
+        "version_constraint": {
+          "major": 3,
+          "minimum": "3.11"
+        },
+        "required_by": [
+          "native_command:test",
+          "grader:grading/runner.py",
+          "grader:grading/regression.py",
+          "agent_prompt"
+        ],
+        "offline_assets": [
+          "/opt/aioftp-venv/lib/python3.11/site-packages/aioftp-0.28.3.dist-info/METADATA",
+          "/opt/aioftp-venv/lib/python3.11/site-packages/__editable__.aioftp-0.28.3.pth",
+          "/opt/aioftp-venv/lib/python3.11/site-packages/pytest_asyncio",
+          "/opt/aioftp-venv/lib/python3.11/site-packages/pytest_cov",
+          "/opt/aioftp-venv/lib/python3.11/site-packages/siosocks",
+          "/opt/aioftp-venv/lib/python3.11/site-packages/trustme",
+          "/opt/aioftp-venv/lib/python3.11/site-packages/async_timeout"
+        ],
+        "description": "CPython 3.11 with the repository's dev extra; spawned by the grader through AIOFTP_REPO_PYTHON, so static discovery cannot see it."
+      },
+      {
+        "id": "repo_pytest",
+        "executable": "/opt/aioftp-venv/bin/pytest",
+        "role": "quality",
+        "discovery": "dynamic",
+        "version_probe": {
+          "argv": [
+            "/opt/aioftp-venv/bin/pytest",
+            "--version"
+          ],
+          "regex": "pytest (\\d+\\.\\d+\\.\\d+)"
+        },
+        "required_by": [
+          "agent_prompt"
+        ],
+        "description": "The `pytest` the prompt tells the agent to run from the repository root (first on the agent PATH)."
+      },
+      {
+        "id": "git",
+        "executable": "git",
+        "role": "system",
+        "discovery": "static",
+        "version_probe": {
+          "argv": [
+            "git",
+            "--version"
+          ],
+          "regex": "git version (\\d+\\.\\d+\\.\\d+)"
+        },
+        "required_by": [
+          "grader:env.py",
+          "grader:tests/test_env.py"
+        ],
+        "description": "Reference-tree materialization and workspace snapshot in env.py."
+      },
+      {
+        "id": "setpriv",
+        "executable": "setpriv",
+        "role": "system",
+        "discovery": "dynamic",
+        "version_probe": {
+          "argv": [
+            "setpriv",
+            "--version"
+          ],
+          "regex": "(\\d+\\.\\d+(?:\\.\\d+)?)"
+        },
+        "required_by": [
+          "grader:env.py"
+        ],
+        "description": "UID wall: UidWallWorkspace.shell_argv prefixes the agent shell with setpriv."
+      },
+      {
+        "id": "hud",
+        "executable": "hud",
+        "role": "system",
+        "discovery": "dynamic",
+        "version_probe": {
+          "argv": [
+            "hud",
+            "--version"
+          ],
+          "regex": "(\\d+\\.\\d+\\.\\d+)"
+        },
+        "required_by": [
+          "grader:env.py"
+        ],
+        "description": "The HUD server the image CMD runs (`hud serve env:env`)."
+      }
+    ],
+    "services": []
+  },
+  "runtime_build_inputs": {
+    "base_images": [
+      {
+        "stage": "runtime-contract",
+        "reference": "ubuntu:24.04@sha256:224a1869083a311ef3f13648a154ba79832fbef6364d31493642ca03082da254"
+      },
+      {
+        "stage": "runtime-contract",
+        "reference": "ghcr.io/astral-sh/uv:0.12.3@sha256:2d890623d310b57771ce840f0da5eed5fc6d657da05ffaa45d82797b53fa3abc"
+      }
+    ],
+    "package_repositories": [
+      {
+        "manager": "apt",
+        "snapshot": "20260915T000000Z"
+      },
+      {
+        "manager": "pip",
+        "snapshot": "runtime/aioftp-dev-requirements.txt (uv pip compile --generate-hashes --exclude-newer 2026-09-15T00:00:00Z, installed --require-hashes) and uv.lock (uv sync --frozen)"
+      }
+    ],
+    "artifacts": [
+      {
+        "id": "cpython_3_11_15",
+        "version": "3.11.15+20260807",
+        "source": "https://releases.astral.sh/github/python-build-standalone/releases/download/20260807/cpython-3.11.15%2B20260807-x86_64-unknown-linux-gnu-install_only_stripped.tar.gz",
+        "sha256": "69dfac9d0f15a0b9281a38486f212cbf76421609228c184dc0d34a0533d57ba6"
+      }
+    ]
+  },
+  "deployment_context": {
+    "required_files": [
+      ".dockerignore",
+      ".gitignore",
+      "Dockerfile.hud",
+      "env.py",
+      "gold.diff",
+      "grading/__init__.py",
+      "grading/adversarial-manifest.json",
+      "grading/baseline/pyproject.toml",
+      "grading/baseline/tests/conftest.py",
+      "grading/baseline/tests/test_abort.py",
+      "grading/baseline/tests/test_client_side_socks.py",
+      "grading/baseline/tests/test_connection.py",
+      "grading/baseline/tests/test_corner_cases.py",
+      "grading/baseline/tests/test_current_directory.py",
+      "grading/baseline/tests/test_directory_actions.py",
+      "grading/baseline/tests/test_extra.py",
+      "grading/baseline/tests/test_file.py",
+      "grading/baseline/tests/test_list_fallback.py",
+      "grading/baseline/tests/test_login.py",
+      "grading/baseline/tests/test_maximum_connections.py",
+      "grading/baseline/tests/test_passive.py",
+      "grading/baseline/tests/test_pathio.py",
+      "grading/baseline/tests/test_permissions.py",
+      "grading/baseline/tests/test_restart.py",
+      "grading/baseline/tests/test_simple_functions.py",
+      "grading/baseline/tests/test_throttle.py",
+      "grading/baseline/tests/test_tls.py",
+      "grading/baseline/tests/test_user.py",
+      "grading/graders.py",
+      "grading/integrity.py",
+      "grading/probes/client_probe.py",
+      "grading/probes/hudprobe_common.py",
+      "grading/probes/hudprobe_fixture.py",
+      "grading/probes/server_probe.py",
+      "grading/regression.py",
+      "grading/runner.py",
+      "pyproject.toml",
+      "runtime/aioftp-dev-requirements.txt",
+      "task/__init__.py",
+      "task/prompt.md",
+      "task/rubric.md",
+      "tasks.py",
+      "uv.lock"
+    ],
+    "forbidden_files": [
+      "artifact-manifest.json",
+      "scaffold-manifest.json"
+    ],
+    "forbidden_directories": [
+      "grading/adversarial",
+      "tests",
+      ".venv"
+    ]
+  },
+  "safety_gate_mutants": [
+    {
+      "id": "g1_abort_takes_completion_as_reply",
+      "target_gate_id": "G1_stale_reply_after_abort",
+      "target_criteria": [
+        "client_abort_after_completed_transfer"
+      ],
+      "cue_preserving": true,
+      "cue_tokens_required_present": [
+        "abort",
+        "ABOR",
+        "225",
+        "226",
+        "426",
+        "1xx",
+        "final reply",
+        "_transfer_stream"
+      ],
+      "patch_path": "grading/adversarial/patches/g1_abort_takes_completion_as_reply.diff",
+      "applies_to": "gold",
+      "description": "Gold whose abort() reads the transfer's final reply first and, when it is a 2xx, takes it as the ABOR reply itself (dropping the ABOR reply from the owed ledger); the error exit keeps an independent drain-any-code path. The cue tokens stay in code and comments. After a completed transfer the 226 Transfer complete is taken as the ABOR reply and the successor PWD reads the stale ABOR reply: client_abort_after_completed_transfer is lost, and client_abort_idle_rejected with it (the 500 ABOR reply after a completed transfer is never read, so abort() does not raise). Raw 0.9120, capped at 0.60.",
+      "expected": {
+        "target_raw_leaf_max": 0.0,
+        "gate_statuses": {
+          "G1_stale_reply_after_abort": false,
+          "G2_stale_reply_after_error_exit": true,
+          "G3_stale_data_connection_after_abort": true,
+          "G4_stale_reply_after_cancelled_finish": true
+        },
+        "safety_gates_failed": true,
+        "build_failed": false,
+        "raw_score": 0.8786,
+        "final_score": 0.6,
+        "final_score_cap": 0.6,
+        "unaffected_criteria": [
+          "server_early_abort_session",
+          "client_stream_exit_after_abort",
+          "client_abort_idle_225",
+          "client_abort_after_failed_transfer",
+          "client_error_exit_mid_transfer",
+          "client_error_exit_after_completed_transfer",
+          "client_upload_stream_abort",
+          "client_abort_concurrent",
+          "client_cancelled_abort_or_finish",
+          "client_abort_during_finish",
+          "client_abandoned_listing_resync",
+          "server_listener_release_after_abort",
+          "server_repeated_abor_order",
+          "server_pipelined_command_order",
+          "server_real_client_abor",
+          "server_quit_during_transfer",
+          "server_data_connection_loss",
+          "server_undecodable_command_syntax",
+          "client_abort_after_cancelled_finish",
+          "server_pipelined_abort_open_data",
+          "server_pipelined_quit_open_data",
+          "client_nowait_abort_owed_replies",
+          "client_command_during_abort"
+        ]
+      }
+    },
+    {
+      "id": "g2_error_exit_drain_expects_426",
+      "target_gate_id": "G2_stale_reply_after_error_exit",
+      "target_criteria": [
+        "client_error_exit_after_completed_transfer"
+      ],
+      "cue_preserving": true,
+      "cue_tokens_required_present": [
+        "__aexit__",
+        "abort",
+        "ABOR",
+        "226",
+        "225",
+        "Transfer complete",
+        "_transfer_stream"
+      ],
+      "patch_path": "grading/adversarial/patches/g2_error_exit_drain_expects_426.diff",
+      "applies_to": "gold",
+      "description": "Gold whose __aexit__ exception branch closes the data connection and finishes expecting 426; the cue tokens stay in code and comments. After a completed transfer the drain meets 226 Transfer complete and raises, replacing the caller's exception: only client_error_exit_after_completed_transfer is lost, raw 0.9932, capped at 0.60.",
+      "expected": {
+        "target_raw_leaf_max": 0.0,
+        "gate_statuses": {
+          "G1_stale_reply_after_abort": true,
+          "G2_stale_reply_after_error_exit": false,
+          "G3_stale_data_connection_after_abort": true,
+          "G4_stale_reply_after_cancelled_finish": true
+        },
+        "safety_gates_failed": true,
+        "build_failed": false,
+        "raw_score": 0.9393,
+        "final_score": 0.6,
+        "final_score_cap": 0.6,
+        "unaffected_criteria": [
+          "server_early_abort_session",
+          "client_stream_exit_after_abort",
+          "client_abort_after_completed_transfer",
+          "client_abort_idle_225",
+          "client_abort_after_failed_transfer",
+          "client_error_exit_mid_transfer",
+          "client_abort_idle_rejected",
+          "client_upload_stream_abort",
+          "client_abort_concurrent",
+          "client_cancelled_abort_or_finish",
+          "client_abort_during_finish",
+          "client_abandoned_listing_resync",
+          "server_listener_release_after_abort",
+          "server_repeated_abor_order",
+          "server_pipelined_command_order",
+          "server_real_client_abor",
+          "server_quit_during_transfer",
+          "server_data_connection_loss",
+          "server_undecodable_command_syntax",
+          "client_abort_after_cancelled_finish",
+          "server_pipelined_abort_open_data",
+          "server_pipelined_quit_open_data",
+          "client_nowait_abort_owed_replies",
+          "client_command_during_abort"
+        ]
+      }
+    },
+    {
+      "id": "g3_listener_kept_after_abort",
+      "target_gate_id": "G3_stale_data_connection_after_abort",
+      "target_criteria": [
+        "server_listener_release_after_abort"
+      ],
+      "cue_preserving": true,
+      "cue_tokens_required_present": [
+        "passive",
+        "data_ports",
+        "_release_passive_server",
+        "426",
+        "226",
+        "abort"
+      ],
+      "patch_path": "grading/adversarial/patches/g3_listener_kept_after_abort.diff",
+      "applies_to": "gold",
+      "description": "Gold whose cancelled transfer worker keeps the passive listener (and its data_ports entry) for the next transfer instead of releasing it; the release helper and the cue tokens stay in code and comments. A data connection arriving late for the aborted transfer is still accepted and kept: only server_listener_release_after_abort is lost, raw 0.9188, capped at 0.60.",
+      "expected": {
+        "target_raw_leaf_max": 0.0,
+        "gate_statuses": {
+          "G1_stale_reply_after_abort": true,
+          "G2_stale_reply_after_error_exit": true,
+          "G3_stale_data_connection_after_abort": false,
+          "G4_stale_reply_after_cancelled_finish": true
+        },
+        "safety_gates_failed": true,
+        "build_failed": false,
+        "raw_score": 0.9835,
+        "final_score": 0.6,
+        "final_score_cap": 0.6,
+        "unaffected_criteria": [
+          "server_early_abort_session",
+          "client_stream_exit_after_abort",
+          "client_abort_after_completed_transfer",
+          "client_abort_idle_225",
+          "client_abort_after_failed_transfer",
+          "client_error_exit_mid_transfer",
+          "client_error_exit_after_completed_transfer",
+          "client_abort_idle_rejected",
+          "client_upload_stream_abort",
+          "client_abort_concurrent",
+          "client_cancelled_abort_or_finish",
+          "client_abort_during_finish",
+          "client_abandoned_listing_resync",
+          "server_repeated_abor_order",
+          "server_pipelined_command_order",
+          "server_real_client_abor",
+          "server_quit_during_transfer",
+          "server_data_connection_loss",
+          "server_undecodable_command_syntax",
+          "client_abort_after_cancelled_finish",
+          "server_pipelined_abort_open_data",
+          "server_pipelined_quit_open_data",
+          "client_nowait_abort_owed_replies",
+          "client_command_during_abort"
+        ]
+      }
+    },
+    {
+      "id": "g4_abort_settles_cancelled_finish_first",
+      "target_gate_id": "G4_stale_reply_after_cancelled_finish",
+      "target_criteria": [
+        "client_abort_after_cancelled_finish"
+      ],
+      "cue_preserving": true,
+      "cue_tokens_required_present": [
+        "abort",
+        "finish",
+        "ABOR",
+        "_owed_replies",
+        "_drain_owed_replies",
+        "cancel"
+      ],
+      "patch_path": "grading/adversarial/patches/g4_abort_settles_cancelled_finish_first.diff",
+      "applies_to": "gold",
+      "description": "Gold whose abort(), when a cancelled finish() left a transfer's final reply owed and nothing else holds the control lock, first reads that owed reply and only then writes ABOR. Against a server that sends the transfer's final reply only once ABOR arrives, abort() blocks permanently (the mechanism of 9065889c candidates c397 and c7e9). CASE-12 still passes because a running finish() holds the lock. Only client_abort_after_cancelled_finish is lost: raw 0.8000, capped at 0.60 by G4.",
+      "expected": {
+        "target_raw_leaf_max": 0.0,
+        "gate_statuses": {
+          "G1_stale_reply_after_abort": true,
+          "G2_stale_reply_after_error_exit": true,
+          "G3_stale_data_connection_after_abort": true,
+          "G4_stale_reply_after_cancelled_finish": false
+        },
+        "safety_gates_failed": true,
+        "build_failed": false,
+        "raw_score": 0.9737,
+        "final_score": 0.6,
+        "final_score_cap": 0.6,
+        "unaffected_criteria": [
+          "server_early_abort_session",
+          "client_stream_exit_after_abort",
+          "client_abort_after_completed_transfer",
+          "client_abort_idle_225",
+          "client_abort_after_failed_transfer",
+          "client_error_exit_mid_transfer",
+          "client_error_exit_after_completed_transfer",
+          "client_abort_idle_rejected",
+          "client_upload_stream_abort",
+          "client_abort_concurrent",
+          "client_cancelled_abort_or_finish",
+          "client_abort_during_finish",
+          "client_abandoned_listing_resync",
+          "server_repeated_abor_order",
+          "server_pipelined_command_order",
+          "server_real_client_abor",
+          "server_quit_during_transfer",
+          "server_data_connection_loss",
+          "server_undecodable_command_syntax",
+          "server_listener_release_after_abort",
+          "server_pipelined_abort_open_data",
+          "server_pipelined_quit_open_data",
+          "client_nowait_abort_owed_replies",
+          "client_command_during_abort"
+        ]
+      }
+    }
+  ],
+  "keyword_stuffing_fixtures": [
+    {
+      "id": "keyword_stuffing_all_cues",
+      "patch_path": "grading/adversarial/patches/keyword_stuffing_all_cues.diff",
+      "target_files": [
+        "src/aioftp/client.py",
+        "src/aioftp/server.py"
+      ],
+      "applies_to": "base",
+      "description": "Base plus comments carrying every Negative-controls cue token of the fifteen criteria next to Client.abort and Server.abor, the two files the probes import from the graded copy. No behaviour changes.",
+      "expected": {
+        "final_score": 0.0,
+        "raw_score": 0.0,
+        "build_failed": false
+      }
+    }
+  ],
+  "quality_bypass_fixtures": [
+    {
+      "id": "quality_bypass_stub_with_weakened_settings",
+      "patch_path": "grading/adversarial/patches/quality_bypass_stub_with_weakened_settings.diff",
+      "applies_to": "base",
+      "description": "The stub/echo control plus a pyproject.toml addopts entry deselecting tests/test_abort.py and a tests/conftest.py collection hook that skips every abort test. The grader restores the pristine settings, so tests/test_abort.py fails, the 0.30 regression ceiling fires, and the diagnostic names addopts (weakened) and tests/conftest.py (changed).",
+      "expected": {
+        "build_failed": true,
+        "raw_score": 0.0,
+        "final_score": 0.0,
+        "final_score_cap": 0.3
+      }
+    }
+  ],
+  "broken_regression_fixtures": [
+    {
+      "id": "broken_regression_no_wait_reads_reply",
+      "patch_path": "grading/adversarial/patches/broken_regression_no_wait_reads_reply.diff",
+      "applies_to": "gold",
+      "description": "Gold with abort(wait=False) changed to read a reply, which breaks tests/test_abort.py::test_abort_retr_no_wait. Every probe still passes: raw 1.0000, final 0.30.",
+      "expected": {
+        "build_failed": true,
+        "native_checks_pass": false,
+        "safety_gates_failed": false,
+        "raw_score": 1.0,
+        "final_score": 0.3,
+        "final_score_cap": 0.3
+      }
+    }
+  ],
+  "no_op_fixtures": [
+    {
+      "id": "no_op_comment_in_client",
+      "patch_path": "grading/adversarial/patches/no_op_comment_in_client.diff",
+      "applies_to": "base",
+      "description": "A comment-only edit to src/aioftp/client.py, the client module every client probe imports.",
+      "expected": {
+        "final_score": 0.0,
+        "raw_score": 0.0,
+        "build_failed": false
+      }
+    }
+  ],
+  "lazy_fix_fixtures": [
+    {
+      "id": "lazy_fix_decorators_and_225",
+      "patch_path": "grading/adversarial/patches/lazy_fix_decorators_and_225.diff",
+      "applies_to": "base",
+      "description": "Own patch against base: the server's @worker decorator moved outermost on the four transfer handlers, and 225 added to abort()'s accepted codes. Nothing else. Approved Solution-level profile: server_early_abort_session and client_abort_idle_225 1.0, every other criterion 0.0 (its base abor handler cancels the closing worker again on a repeated ABOR, so CASE-15 is lost too); raw 0.0826, final 0.0826, inside [0.02, 0.30]. The gated children score 0, so all four gates are recorded as fired, but their shared 0.60 ceiling does not bind (final equals raw): the score is partial credit alone.",
+      "expected": {
+        "criteria_profile": {
+          "server_early_abort_session": 1.0,
+          "client_stream_exit_after_abort": 0.0,
+          "client_abort_after_completed_transfer": 0.0,
+          "client_abort_idle_225": 1.0,
+          "client_abort_after_failed_transfer": 0.0,
+          "client_error_exit_mid_transfer": 0.0,
+          "client_error_exit_after_completed_transfer": 0.0,
+          "client_abort_idle_rejected": 0.0,
+          "client_upload_stream_abort": 0.0,
+          "client_abort_concurrent": 0.0,
+          "client_cancelled_abort_or_finish": 0.0,
+          "client_abort_during_finish": 0.0,
+          "client_abandoned_listing_resync": 0.0,
+          "server_listener_release_after_abort": 0.0,
+          "server_repeated_abor_order": 0.0,
+          "server_pipelined_command_order": 0.0,
+          "server_real_client_abor": 0.0,
+          "server_quit_during_transfer": 0.0,
+          "server_data_connection_loss": 0.0,
+          "server_undecodable_command_syntax": 0.0,
+          "client_abort_after_cancelled_finish": 0.0,
+          "server_pipelined_abort_open_data": 0.0,
+          "server_pipelined_quit_open_data": 0.0,
+          "client_nowait_abort_owed_replies": 0.0,
+          "client_command_during_abort": 0.0
+        },
+        "final_score_min": 0.02,
+        "final_score_max": 0.3,
+        "raw_score": 0.0266,
+        "final_score": 0.0266,
+        "build_failed": false,
+        "safety_gates_failed": true
+      }
+    }
+  ],
+  "competent_partial_fixtures": [
+    {
+      "id": "competent_partial_naive_abort",
+      "patch_path": "grading/adversarial/patches/competent_partial_naive_abort.diff",
+      "applies_to": "base",
+      "description": "The semantic viability screen's shortest_competent_solution, written from the prompt: accept 225, swap the server decorators, make finish() return after abort(), and abort on an exception exit (about twenty lines). Band = the `Projected raw mean: 0.476` of the fifteen-case refit in intake/difficulty-rationale.md, +/- 0.15. Measured: it earns CASE-01, 02, 04, 06, 09 and CASE-21 (its abort() writes ABOR at once and finish() returns after abort(), so the cancelled-finish scenario passes): raw 0.2908, inside the band.",
+      "expected": {
+        "final_score_min": 0.068,
+        "final_score_max": 0.432
+      }
+    }
+  ],
+  "stub_echo_fixtures": [
+    {
+      "id": "stub_echo_constant_entrypoints",
+      "patch_path": "grading/adversarial/patches/stub_echo_constant_entrypoints.diff",
+      "applies_to": "base",
+      "description": "Client.abort() returns None without I/O, finish() and __aexit__ only close, and Server.abor returns True without replying.",
+      "expected": {
+        "final_score": 0.0,
+        "raw_score": 0.0,
+        "all_behavioural_children_zero": true
+      }
+    }
+  ],
+  "alternate_correct_fixtures": [
+    {
+      "id": "alternate_a_counter_and_cancel_guard",
+      "patch_path": "grading/adversarial/patches/alternate_a_counter_and_cancel_guard.diff",
+      "applies_to": "base",
+      "independence": "structural",
+      "purpose": "Independent implementation A. Every exchange takes an integer turn number in wire order; replies are read in turn order under a reader lock and a reply read on behalf of a cancelled or later owner is kept for it (_early_replies). Owed transfer replies are a per-client integer count (_pending_final_replies) plus a turn reserved by abort(wait=False); the owed reply is read through a private Client method (_drain_outstanding_reply) and abort stays on Client. An abandoned transfer is aborted by the next command and by the lister's aclose(). The server answers an abort that lands before the data connection by catching CancelledError around each transfer worker task (releasing the listener and its port there), refuses late data connections in the passive handlers, and answers a repeated ABOR after the running one through Task.cancelling(). Rules out coupling to gold's stream reference, its Boolean marker, its owed-reply deque, abort's placement on BaseClient, the write/control lock pair, the aborting_workers set and the decorator order.",
+      "materiality": {
+        "min_non_comment_changed_lines": 120,
+        "min_files_touched": 2,
+        "max_gold_added_line_overlap_ratio": 0.3,
+        "min_gold_delta_lines": 120,
+        "forbidden_added_symbols": [
+          "_transfer_stream",
+          "final_reply_consumed",
+          "_owed_replies",
+          "aborting_workers"
+        ]
+      },
+      "independence_axes": {
+        "callable_surface": {
+          "rationale": "The owed reply is read by the private Client._drain_outstanding_reply and abort() stays on Client; gold moves abort to BaseClient and reads through _exchange/_read_owed."
+        },
+        "persistence_layout": {
+          "rationale": "State is BaseClient._pending_final_replies (a counter) plus integer turn counters and an _early_replies map, instead of gold's BaseClient._transfer_stream stream reference and _owed_replies deque of tuples.",
+          "gold_literals_absent": [
+            "_transfer_stream",
+            "_owed_replies"
+          ]
+        },
+        "marker_representation": {
+          "rationale": "Integer turn numbers and a count replace gold's per-stream Boolean final_reply_consumed.",
+          "gold_literals_absent": [
+            "final_reply_consumed"
+          ]
+        }
+      },
+      "expected": {
+        "final_score": 1.0,
+        "raw_score": 1.0,
+        "build_failed": false,
+        "safety_gates_failed": false
+      }
+    },
+    {
+      "id": "alternate_b_ledger_module_and_drain_exit",
+      "patch_path": "grading/adversarial/patches/alternate_b_ledger_module_and_drain_exit.diff",
+      "applies_to": "base",
+      "independence": "structural",
+      "purpose": "Independent implementation B. Reply ownership lives in a new private module (src/aioftp/_replies.py: ReplyLedger with Slip records, tickets, reservation and an ordered collector) used by client.py; finish() is idempotent through a None ticket sentinel; the stream's exception exit, the lister's aclose() and the next command after an abandoned transfer all close the data connection and drain the owed final reply with any code without sending ABOR; the server waits for the data connection inside the abortable worker (dropping the listener and returning its port when cancelled there) and its abor handler counts live workers, answering a repeated ABOR after the running one. Rules out coupling to gold's module composition and to abort() as the error-exit and settle seam.",
+      "materiality": {
+        "min_non_comment_changed_lines": 150,
+        "min_files_touched": 3,
+        "max_gold_added_line_overlap_ratio": 0.2,
+        "min_gold_delta_lines": 120,
+        "forbidden_added_symbols": [
+          "_transfer_stream",
+          "final_reply_consumed",
+          "_owed_replies",
+          "aborting_workers"
+        ]
+      },
+      "independence_axes": {
+        "module_composition": {
+          "rationale": "Reply bookkeeping is a new private module src/aioftp/_replies.py imported by client.py; gold keeps everything inside client.py and server.py."
+        },
+        "injection_seam": {
+          "rationale": "The exception exit, aclose() and the settle of an abandoned transfer close the data connection and drain the owed final reply with any code, without calling abort() or sending ABOR; gold's exit, aclose() and settle call abort()."
+        }
+      },
+      "expected": {
+        "final_score": 1.0,
+        "raw_score": 1.0,
+        "build_failed": false,
+        "safety_gates_failed": false
+      }
+    }
+  ],
+  "malformed_output_fault_injections": [
+    {
+      "id": "delete_client_probe",
+      "kind": "delete_file",
+      "target": "grading/probes/client_probe.py",
+      "expected": {
+        "is_error": true,
+        "reward": 0.0
+      }
+    },
+    {
+      "id": "corrupt_probe_fixture",
+      "kind": "corrupt_file",
+      "target": "grading/probes/hudprobe_fixture.py",
+      "expected": {
+        "is_error": true,
+        "reward": 0.0
+      }
+    },
+    {
+      "id": "corrupt_scoring_module",
+      "kind": "corrupt_file",
+      "target": "grading/graders.py",
+      "expected": {
+        "is_error": true,
+        "reward": 0.0
+      }
+    }
+  ],
+  "keyword_stuffing_gold_fixtures": [
+    {
+      "id": "ksg_server_decorators_restored",
+      "patch_path": "grading/adversarial/patches/ksg_server_decorators_restored.diff",
+      "applies_to": "gold",
+      "description": "Gold with ConnectionConditions restored outermost on the four transfer handlers; the listener release and the cue tokens stay in code and comments. A transfer cancelled before its data connection (ABOR or QUIT) is not answered and the session drops: server_early_abort_session, server_listener_release_after_abort, server_pipelined_command_order, server_quit_during_transfer are lost (G3 fires), raw 0.7830, final 0.60.",
+      "cue_tokens_required_present": [
+        "426",
+        "226",
+        "transfer aborted",
+        "abort successful",
+        "worker"
+      ],
+      "expected": {
+        "target_criteria": [
+          "server_early_abort_session",
+          "server_listener_release_after_abort",
+          "server_pipelined_command_order",
+          "server_quit_during_transfer"
+        ],
+        "target_raw_leaf_max": 0.0,
+        "criteria_profile": {
+          "server_early_abort_session": 0.0,
+          "client_stream_exit_after_abort": 1.0,
+          "client_abort_after_completed_transfer": 1.0,
+          "client_abort_idle_225": 1.0,
+          "client_abort_after_failed_transfer": 1.0,
+          "client_error_exit_mid_transfer": 1.0,
+          "client_error_exit_after_completed_transfer": 1.0,
+          "client_abort_idle_rejected": 1.0,
+          "client_upload_stream_abort": 1.0,
+          "client_abort_concurrent": 1.0,
+          "client_cancelled_abort_or_finish": 1.0,
+          "client_abort_during_finish": 1.0,
+          "client_abandoned_listing_resync": 1.0,
+          "server_listener_release_after_abort": 0.0,
+          "server_repeated_abor_order": 1.0,
+          "server_pipelined_command_order": 0.0,
+          "server_real_client_abor": 1.0,
+          "server_quit_during_transfer": 0.0,
+          "server_data_connection_loss": 1.0,
+          "server_undecodable_command_syntax": 1.0,
+          "client_abort_after_cancelled_finish": 1.0,
+          "server_pipelined_abort_open_data": 1.0,
+          "server_pipelined_quit_open_data": 1.0,
+          "client_nowait_abort_owed_replies": 1.0,
+          "client_command_during_abort": 1.0
+        },
+        "raw_score": 0.9557,
+        "final_score": 0.6,
+        "unaffected_criteria": [
+          "client_stream_exit_after_abort",
+          "client_abort_after_completed_transfer",
+          "client_abort_idle_225",
+          "client_abort_after_failed_transfer",
+          "client_error_exit_mid_transfer",
+          "client_error_exit_after_completed_transfer",
+          "client_abort_idle_rejected",
+          "client_upload_stream_abort",
+          "client_abort_concurrent",
+          "client_cancelled_abort_or_finish",
+          "client_abort_during_finish",
+          "client_abandoned_listing_resync",
+          "server_repeated_abor_order",
+          "server_real_client_abor",
+          "server_data_connection_loss",
+          "server_undecodable_command_syntax",
+          "client_abort_after_cancelled_finish",
+          "server_pipelined_abort_open_data",
+          "server_pipelined_quit_open_data",
+          "client_nowait_abort_owed_replies",
+          "client_command_during_abort"
+        ],
+        "build_failed": false,
+        "safety_gates_failed": true,
+        "gate_statuses": {
+          "G1_stale_reply_after_abort": true,
+          "G2_stale_reply_after_error_exit": true,
+          "G3_stale_data_connection_after_abort": false,
+          "G4_stale_reply_after_cancelled_finish": true
+        }
+      }
+    },
+    {
+      "id": "ksg_finish_unguarded",
+      "patch_path": "grading/adversarial/patches/ksg_finish_unguarded.diff",
+      "applies_to": "gold",
+      "description": "Gold whose finish() always reads a reply; cue tokens finish, abort, consumed and __aexit__ stay in code and comments. A block exit after abort() waits for a reply that never comes: client_stream_exit_after_abort, client_abort_concurrent, client_cancelled_abort_or_finish and client_abort_after_cancelled_finish are lost (G4 fires), raw 0.5474.",
+      "cue_tokens_required_present": [
+        "finish",
+        "abort",
+        "consumed",
+        "__aexit__"
+      ],
+      "expected": {
+        "target_criteria": [
+          "client_stream_exit_after_abort",
+          "client_abort_concurrent",
+          "client_cancelled_abort_or_finish",
+          "client_abort_after_cancelled_finish"
+        ],
+        "target_raw_leaf_max": 0.0,
+        "criteria_profile": {
+          "server_early_abort_session": 1.0,
+          "client_stream_exit_after_abort": 0.0,
+          "client_abort_after_completed_transfer": 1.0,
+          "client_abort_idle_225": 1.0,
+          "client_abort_after_failed_transfer": 1.0,
+          "client_error_exit_mid_transfer": 1.0,
+          "client_error_exit_after_completed_transfer": 1.0,
+          "client_abort_idle_rejected": 1.0,
+          "client_upload_stream_abort": 1.0,
+          "client_abort_concurrent": 0.0,
+          "client_cancelled_abort_or_finish": 0.0,
+          "client_abort_during_finish": 1.0,
+          "client_abandoned_listing_resync": 1.0,
+          "server_listener_release_after_abort": 1.0,
+          "server_repeated_abor_order": 1.0,
+          "server_pipelined_command_order": 1.0,
+          "server_real_client_abor": 1.0,
+          "server_quit_during_transfer": 1.0,
+          "server_data_connection_loss": 1.0,
+          "server_undecodable_command_syntax": 1.0,
+          "client_abort_after_cancelled_finish": 0.0,
+          "server_pipelined_abort_open_data": 1.0,
+          "server_pipelined_quit_open_data": 1.0,
+          "client_nowait_abort_owed_replies": 1.0,
+          "client_command_during_abort": 1.0
+        },
+        "raw_score": 0.9367,
+        "final_score": 0.6,
+        "unaffected_criteria": [
+          "server_early_abort_session",
+          "client_abort_after_completed_transfer",
+          "client_abort_idle_225",
+          "client_abort_after_failed_transfer",
+          "client_error_exit_mid_transfer",
+          "client_error_exit_after_completed_transfer",
+          "client_abort_idle_rejected",
+          "client_upload_stream_abort",
+          "client_abort_during_finish",
+          "client_abandoned_listing_resync",
+          "server_listener_release_after_abort",
+          "server_repeated_abor_order",
+          "server_pipelined_command_order",
+          "server_real_client_abor",
+          "server_quit_during_transfer",
+          "server_data_connection_loss",
+          "server_undecodable_command_syntax",
+          "server_pipelined_abort_open_data",
+          "server_pipelined_quit_open_data",
+          "client_nowait_abort_owed_replies",
+          "client_command_during_abort"
+        ],
+        "build_failed": false,
+        "safety_gates_failed": true,
+        "gate_statuses": {
+          "G1_stale_reply_after_abort": true,
+          "G2_stale_reply_after_error_exit": true,
+          "G3_stale_data_connection_after_abort": true,
+          "G4_stale_reply_after_cancelled_finish": false
+        }
+      }
+    },
+    {
+      "id": "ksg_idle_abort_rejects_225",
+      "patch_path": "grading/adversarial/patches/ksg_idle_abort_rejects_225.diff",
+      "applies_to": "gold",
+      "description": "Gold whose idle abort() accepts only 226; cue tokens 225, 226 and ABOR stay in code and comments. Only client_abort_idle_225 is lost, raw 0.9986.",
+      "cue_tokens_required_present": [
+        "225",
+        "226",
+        "ABOR"
+      ],
+      "expected": {
+        "target_criteria": [
+          "client_abort_idle_225"
+        ],
+        "target_raw_leaf_max": 0.0,
+        "criteria_profile": {
+          "server_early_abort_session": 1.0,
+          "client_stream_exit_after_abort": 1.0,
+          "client_abort_after_completed_transfer": 1.0,
+          "client_abort_idle_225": 0.0,
+          "client_abort_after_failed_transfer": 1.0,
+          "client_error_exit_mid_transfer": 1.0,
+          "client_error_exit_after_completed_transfer": 1.0,
+          "client_abort_idle_rejected": 1.0,
+          "client_upload_stream_abort": 1.0,
+          "client_abort_concurrent": 1.0,
+          "client_cancelled_abort_or_finish": 1.0,
+          "client_abort_during_finish": 1.0,
+          "client_abandoned_listing_resync": 1.0,
+          "server_listener_release_after_abort": 1.0,
+          "server_repeated_abor_order": 1.0,
+          "server_pipelined_command_order": 1.0,
+          "server_real_client_abor": 1.0,
+          "server_quit_during_transfer": 1.0,
+          "server_data_connection_loss": 1.0,
+          "server_undecodable_command_syntax": 1.0,
+          "client_abort_after_cancelled_finish": 1.0,
+          "server_pipelined_abort_open_data": 1.0,
+          "server_pipelined_quit_open_data": 1.0,
+          "client_nowait_abort_owed_replies": 1.0,
+          "client_command_during_abort": 1.0
+        },
+        "raw_score": 0.99,
+        "final_score": 0.99,
+        "unaffected_criteria": [
+          "server_early_abort_session",
+          "client_stream_exit_after_abort",
+          "client_abort_after_completed_transfer",
+          "client_abort_after_failed_transfer",
+          "client_error_exit_mid_transfer",
+          "client_error_exit_after_completed_transfer",
+          "client_abort_idle_rejected",
+          "client_upload_stream_abort",
+          "client_abort_concurrent",
+          "client_cancelled_abort_or_finish",
+          "client_abort_during_finish",
+          "client_abandoned_listing_resync",
+          "server_listener_release_after_abort",
+          "server_repeated_abor_order",
+          "server_pipelined_command_order",
+          "server_real_client_abor",
+          "server_quit_during_transfer",
+          "server_data_connection_loss",
+          "server_undecodable_command_syntax",
+          "client_abort_after_cancelled_finish",
+          "server_pipelined_abort_open_data",
+          "server_pipelined_quit_open_data",
+          "client_nowait_abort_owed_replies",
+          "client_command_during_abort"
+        ],
+        "build_failed": false,
+        "safety_gates_failed": false,
+        "gate_statuses": {
+          "G1_stale_reply_after_abort": true,
+          "G2_stale_reply_after_error_exit": true,
+          "G3_stale_data_connection_after_abort": true,
+          "G4_stale_reply_after_cancelled_finish": true
+        }
+      }
+    },
+    {
+      "id": "ksg_failed_transfer_code_checked",
+      "patch_path": "grading/adversarial/patches/ksg_failed_transfer_code_checked.diff",
+      "applies_to": "gold",
+      "description": "Gold whose abort() validates the owed final reply as 2xx or 426 (code-checked tracking); cue tokens abort, ABOR, 451, final reply and any code stay in code and comments. Only client_abort_after_failed_transfer is lost, raw 0.9932.",
+      "cue_tokens_required_present": [
+        "abort",
+        "ABOR",
+        "451",
+        "final reply",
+        "any code"
+      ],
+      "expected": {
+        "target_criteria": [
+          "client_abort_after_failed_transfer"
+        ],
+        "target_raw_leaf_max": 0.0,
+        "criteria_profile": {
+          "server_early_abort_session": 1.0,
+          "client_stream_exit_after_abort": 1.0,
+          "client_abort_after_completed_transfer": 1.0,
+          "client_abort_idle_225": 1.0,
+          "client_abort_after_failed_transfer": 0.0,
+          "client_error_exit_mid_transfer": 1.0,
+          "client_error_exit_after_completed_transfer": 1.0,
+          "client_abort_idle_rejected": 1.0,
+          "client_upload_stream_abort": 1.0,
+          "client_abort_concurrent": 1.0,
+          "client_cancelled_abort_or_finish": 1.0,
+          "client_abort_during_finish": 1.0,
+          "client_abandoned_listing_resync": 1.0,
+          "server_listener_release_after_abort": 1.0,
+          "server_repeated_abor_order": 1.0,
+          "server_pipelined_command_order": 1.0,
+          "server_real_client_abor": 1.0,
+          "server_quit_during_transfer": 1.0,
+          "server_data_connection_loss": 1.0,
+          "server_undecodable_command_syntax": 1.0,
+          "client_abort_after_cancelled_finish": 1.0,
+          "server_pipelined_abort_open_data": 1.0,
+          "server_pipelined_quit_open_data": 1.0,
+          "client_nowait_abort_owed_replies": 1.0,
+          "client_command_during_abort": 1.0
+        },
+        "raw_score": 0.9393,
+        "final_score": 0.9393,
+        "unaffected_criteria": [
+          "server_early_abort_session",
+          "client_stream_exit_after_abort",
+          "client_abort_after_completed_transfer",
+          "client_abort_idle_225",
+          "client_error_exit_mid_transfer",
+          "client_error_exit_after_completed_transfer",
+          "client_abort_idle_rejected",
+          "client_upload_stream_abort",
+          "client_abort_concurrent",
+          "client_cancelled_abort_or_finish",
+          "client_abort_during_finish",
+          "client_abandoned_listing_resync",
+          "server_listener_release_after_abort",
+          "server_repeated_abor_order",
+          "server_pipelined_command_order",
+          "server_real_client_abor",
+          "server_quit_during_transfer",
+          "server_data_connection_loss",
+          "server_undecodable_command_syntax",
+          "client_abort_after_cancelled_finish",
+          "server_pipelined_abort_open_data",
+          "server_pipelined_quit_open_data",
+          "client_nowait_abort_owed_replies",
+          "client_command_during_abort"
+        ],
+        "build_failed": false,
+        "safety_gates_failed": false,
+        "gate_statuses": {
+          "G1_stale_reply_after_abort": true,
+          "G2_stale_reply_after_error_exit": true,
+          "G3_stale_data_connection_after_abort": true,
+          "G4_stale_reply_after_cancelled_finish": true
+        }
+      }
+    },
+    {
+      "id": "ksg_error_exit_mid_transfer_finish_replaces",
+      "patch_path": "grading/adversarial/patches/ksg_error_exit_mid_transfer_finish_replaces.diff",
+      "applies_to": "gold",
+      "description": "Gold whose __aexit__ exception branch finishes the stream instead of aborting; cue tokens __aexit__, abort, Exception and close stay in code and comments. Mid-transfer the finish meets 426 and its StatusCodeError replaces the caller's exception: client_error_exit_mid_transfer and client_upload_stream_abort are lost, raw 0.9938.",
+      "cue_tokens_required_present": [
+        "__aexit__",
+        "abort",
+        "Exception",
+        "close"
+      ],
+      "expected": {
+        "target_criteria": [
+          "client_error_exit_mid_transfer",
+          "client_upload_stream_abort"
+        ],
+        "target_raw_leaf_max": 0.0,
+        "criteria_profile": {
+          "server_early_abort_session": 1.0,
+          "client_stream_exit_after_abort": 1.0,
+          "client_abort_after_completed_transfer": 1.0,
+          "client_abort_idle_225": 1.0,
+          "client_abort_after_failed_transfer": 1.0,
+          "client_error_exit_mid_transfer": 0.0,
+          "client_error_exit_after_completed_transfer": 1.0,
+          "client_abort_idle_rejected": 1.0,
+          "client_upload_stream_abort": 0.0,
+          "client_abort_concurrent": 1.0,
+          "client_cancelled_abort_or_finish": 1.0,
+          "client_abort_during_finish": 1.0,
+          "client_abandoned_listing_resync": 1.0,
+          "server_listener_release_after_abort": 1.0,
+          "server_repeated_abor_order": 1.0,
+          "server_pipelined_command_order": 1.0,
+          "server_real_client_abor": 1.0,
+          "server_quit_during_transfer": 1.0,
+          "server_data_connection_loss": 1.0,
+          "server_undecodable_command_syntax": 1.0,
+          "client_abort_after_cancelled_finish": 1.0,
+          "server_pipelined_abort_open_data": 1.0,
+          "server_pipelined_quit_open_data": 1.0,
+          "client_nowait_abort_owed_replies": 1.0,
+          "client_command_during_abort": 1.0
+        },
+        "raw_score": 0.989,
+        "final_score": 0.989,
+        "unaffected_criteria": [
+          "server_early_abort_session",
+          "client_stream_exit_after_abort",
+          "client_abort_after_completed_transfer",
+          "client_abort_idle_225",
+          "client_abort_after_failed_transfer",
+          "client_error_exit_after_completed_transfer",
+          "client_abort_idle_rejected",
+          "client_abort_concurrent",
+          "client_cancelled_abort_or_finish",
+          "client_abort_during_finish",
+          "client_abandoned_listing_resync",
+          "server_listener_release_after_abort",
+          "server_repeated_abor_order",
+          "server_pipelined_command_order",
+          "server_real_client_abor",
+          "server_quit_during_transfer",
+          "server_data_connection_loss",
+          "server_undecodable_command_syntax",
+          "client_abort_after_cancelled_finish",
+          "server_pipelined_abort_open_data",
+          "server_pipelined_quit_open_data",
+          "client_nowait_abort_owed_replies",
+          "client_command_during_abort"
+        ],
+        "build_failed": false,
+        "safety_gates_failed": false,
+        "gate_statuses": {
+          "G1_stale_reply_after_abort": true,
+          "G2_stale_reply_after_error_exit": true,
+          "G3_stale_data_connection_after_abort": true,
+          "G4_stale_reply_after_cancelled_finish": true
+        }
+      }
+    },
+    {
+      "id": "ksg_rejected_abor_accepts_5xx",
+      "patch_path": "grading/adversarial/patches/ksg_rejected_abor_accepts_5xx.diff",
+      "applies_to": "gold",
+      "description": "Gold whose abort() accepts a 5xx as the ABOR reply; cue tokens 225, 226, ABOR and 5xx stay in code and comments. abort() does not raise on 500: only client_abort_idle_rejected is lost, raw 0.9188.",
+      "cue_tokens_required_present": [
+        "225",
+        "226",
+        "ABOR",
+        "5xx"
+      ],
+      "expected": {
+        "target_criteria": [
+          "client_abort_idle_rejected"
+        ],
+        "target_raw_leaf_max": 0.0,
+        "criteria_profile": {
+          "server_early_abort_session": 1.0,
+          "client_stream_exit_after_abort": 1.0,
+          "client_abort_after_completed_transfer": 1.0,
+          "client_abort_idle_225": 1.0,
+          "client_abort_after_failed_transfer": 1.0,
+          "client_error_exit_mid_transfer": 1.0,
+          "client_error_exit_after_completed_transfer": 1.0,
+          "client_abort_idle_rejected": 0.0,
+          "client_upload_stream_abort": 1.0,
+          "client_abort_concurrent": 1.0,
+          "client_cancelled_abort_or_finish": 1.0,
+          "client_abort_during_finish": 1.0,
+          "client_abandoned_listing_resync": 1.0,
+          "server_listener_release_after_abort": 1.0,
+          "server_repeated_abor_order": 1.0,
+          "server_pipelined_command_order": 1.0,
+          "server_real_client_abor": 1.0,
+          "server_quit_during_transfer": 1.0,
+          "server_data_connection_loss": 1.0,
+          "server_undecodable_command_syntax": 1.0,
+          "client_abort_after_cancelled_finish": 1.0,
+          "server_pipelined_abort_open_data": 1.0,
+          "server_pipelined_quit_open_data": 1.0,
+          "client_nowait_abort_owed_replies": 1.0,
+          "client_command_during_abort": 1.0
+        },
+        "raw_score": 0.9393,
+        "final_score": 0.9393,
+        "unaffected_criteria": [
+          "server_early_abort_session",
+          "client_stream_exit_after_abort",
+          "client_abort_after_completed_transfer",
+          "client_abort_idle_225",
+          "client_abort_after_failed_transfer",
+          "client_error_exit_mid_transfer",
+          "client_error_exit_after_completed_transfer",
+          "client_upload_stream_abort",
+          "client_abort_concurrent",
+          "client_cancelled_abort_or_finish",
+          "client_abort_during_finish",
+          "client_abandoned_listing_resync",
+          "server_listener_release_after_abort",
+          "server_repeated_abor_order",
+          "server_pipelined_command_order",
+          "server_real_client_abor",
+          "server_quit_during_transfer",
+          "server_data_connection_loss",
+          "server_undecodable_command_syntax",
+          "client_abort_after_cancelled_finish",
+          "server_pipelined_abort_open_data",
+          "server_pipelined_quit_open_data",
+          "client_nowait_abort_owed_replies",
+          "client_command_during_abort"
+        ],
+        "build_failed": false,
+        "safety_gates_failed": false,
+        "gate_statuses": {
+          "G1_stale_reply_after_abort": true,
+          "G2_stale_reply_after_error_exit": true,
+          "G3_stale_data_connection_after_abort": true,
+          "G4_stale_reply_after_cancelled_finish": true
+        }
+      }
+    },
+    {
+      "id": "ksg_upload_exit_finishes",
+      "patch_path": "grading/adversarial/patches/ksg_upload_exit_finishes.diff",
+      "applies_to": "gold",
+      "description": "Gold whose __aexit__ exception branch finishes an upload stream instead of aborting it (a marker set by upload_stream/append_stream); cue tokens upload_stream, STOR, __aexit__ and abort stay in code and comments. Only client_upload_stream_abort is lost, raw 0.9952.",
+      "cue_tokens_required_present": [
+        "upload_stream",
+        "STOR",
+        "__aexit__",
+        "abort"
+      ],
+      "expected": {
+        "target_criteria": [
+          "client_upload_stream_abort"
+        ],
+        "target_raw_leaf_max": 0.0,
+        "criteria_profile": {
+          "server_early_abort_session": 1.0,
+          "client_stream_exit_after_abort": 1.0,
+          "client_abort_after_completed_transfer": 1.0,
+          "client_abort_idle_225": 1.0,
+          "client_abort_after_failed_transfer": 1.0,
+          "client_error_exit_mid_transfer": 1.0,
+          "client_error_exit_after_completed_transfer": 1.0,
+          "client_abort_idle_rejected": 1.0,
+          "client_upload_stream_abort": 0.0,
+          "client_abort_concurrent": 1.0,
+          "client_cancelled_abort_or_finish": 1.0,
+          "client_abort_during_finish": 1.0,
+          "client_abandoned_listing_resync": 1.0,
+          "server_listener_release_after_abort": 1.0,
+          "server_repeated_abor_order": 1.0,
+          "server_pipelined_command_order": 1.0,
+          "server_real_client_abor": 1.0,
+          "server_quit_during_transfer": 1.0,
+          "server_data_connection_loss": 1.0,
+          "server_undecodable_command_syntax": 1.0,
+          "client_abort_after_cancelled_finish": 1.0,
+          "server_pipelined_abort_open_data": 1.0,
+          "server_pipelined_quit_open_data": 1.0,
+          "client_nowait_abort_owed_replies": 1.0,
+          "client_command_during_abort": 1.0
+        },
+        "raw_score": 0.999,
+        "final_score": 0.999,
+        "unaffected_criteria": [
+          "server_early_abort_session",
+          "client_stream_exit_after_abort",
+          "client_abort_after_completed_transfer",
+          "client_abort_idle_225",
+          "client_abort_after_failed_transfer",
+          "client_error_exit_mid_transfer",
+          "client_error_exit_after_completed_transfer",
+          "client_abort_idle_rejected",
+          "client_abort_concurrent",
+          "client_cancelled_abort_or_finish",
+          "client_abort_during_finish",
+          "client_abandoned_listing_resync",
+          "server_listener_release_after_abort",
+          "server_repeated_abor_order",
+          "server_pipelined_command_order",
+          "server_real_client_abor",
+          "server_quit_during_transfer",
+          "server_data_connection_loss",
+          "server_undecodable_command_syntax",
+          "client_abort_after_cancelled_finish",
+          "server_pipelined_abort_open_data",
+          "server_pipelined_quit_open_data",
+          "client_nowait_abort_owed_replies",
+          "client_command_during_abort"
+        ],
+        "build_failed": false,
+        "safety_gates_failed": false,
+        "gate_statuses": {
+          "G1_stale_reply_after_abort": true,
+          "G2_stale_reply_after_error_exit": true,
+          "G3_stale_data_connection_after_abort": true,
+          "G4_stale_reply_after_cancelled_finish": true
+        }
+      }
+    },
+    {
+      "id": "ksg_concurrent_abort_rejected",
+      "patch_path": "grading/adversarial/patches/ksg_concurrent_abort_rejected.diff",
+      "applies_to": "gold",
+      "description": "Gold whose abort() raises RuntimeError while another abort() is in flight; cue tokens concurrent, serialized, abort and ABOR stay in code and comments. Only client_abort_concurrent is lost, raw 0.9932.",
+      "cue_tokens_required_present": [
+        "concurrent",
+        "serialized",
+        "abort",
+        "ABOR"
+      ],
+      "expected": {
+        "target_criteria": [
+          "client_abort_concurrent"
+        ],
+        "target_raw_leaf_max": 0.0,
+        "criteria_profile": {
+          "server_early_abort_session": 1.0,
+          "client_stream_exit_after_abort": 1.0,
+          "client_abort_after_completed_transfer": 1.0,
+          "client_abort_idle_225": 1.0,
+          "client_abort_after_failed_transfer": 1.0,
+          "client_error_exit_mid_transfer": 1.0,
+          "client_error_exit_after_completed_transfer": 1.0,
+          "client_abort_idle_rejected": 1.0,
+          "client_upload_stream_abort": 1.0,
+          "client_abort_concurrent": 0.0,
+          "client_cancelled_abort_or_finish": 1.0,
+          "client_abort_during_finish": 1.0,
+          "client_abandoned_listing_resync": 1.0,
+          "server_listener_release_after_abort": 1.0,
+          "server_repeated_abor_order": 1.0,
+          "server_pipelined_command_order": 1.0,
+          "server_real_client_abor": 1.0,
+          "server_quit_during_transfer": 1.0,
+          "server_data_connection_loss": 1.0,
+          "server_undecodable_command_syntax": 1.0,
+          "client_abort_after_cancelled_finish": 1.0,
+          "server_pipelined_abort_open_data": 1.0,
+          "server_pipelined_quit_open_data": 1.0,
+          "client_nowait_abort_owed_replies": 1.0,
+          "client_command_during_abort": 1.0
+        },
+        "raw_score": 0.9986,
+        "final_score": 0.9986,
+        "unaffected_criteria": [
+          "server_early_abort_session",
+          "client_stream_exit_after_abort",
+          "client_abort_after_completed_transfer",
+          "client_abort_idle_225",
+          "client_abort_after_failed_transfer",
+          "client_error_exit_mid_transfer",
+          "client_error_exit_after_completed_transfer",
+          "client_abort_idle_rejected",
+          "client_upload_stream_abort",
+          "client_cancelled_abort_or_finish",
+          "client_abort_during_finish",
+          "client_abandoned_listing_resync",
+          "server_listener_release_after_abort",
+          "server_repeated_abor_order",
+          "server_pipelined_command_order",
+          "server_real_client_abor",
+          "server_quit_during_transfer",
+          "server_data_connection_loss",
+          "server_undecodable_command_syntax",
+          "client_abort_after_cancelled_finish",
+          "server_pipelined_abort_open_data",
+          "server_pipelined_quit_open_data",
+          "client_nowait_abort_owed_replies",
+          "client_command_during_abort"
+        ],
+        "build_failed": false,
+        "safety_gates_failed": false,
+        "gate_statuses": {
+          "G1_stale_reply_after_abort": true,
+          "G2_stale_reply_after_error_exit": true,
+          "G3_stale_data_connection_after_abort": true,
+          "G4_stale_reply_after_cancelled_finish": true
+        }
+      }
+    },
+    {
+      "id": "ksg_cancelled_wait_forgets_replies",
+      "patch_path": "grading/adversarial/patches/ksg_cancelled_wait_forgets_replies.diff",
+      "applies_to": "gold",
+      "description": "Gold whose abort() and finish() drop their owed replies from the ledger when cancelled; cue tokens cancel, owed, CancelledError, abort and finish stay in code and comments. The next command reads the stale 426 or 226: client_cancelled_abort_or_finish and client_abort_after_cancelled_finish are lost (G4 fires), raw 0.5562.",
+      "cue_tokens_required_present": [
+        "cancel",
+        "owed",
+        "CancelledError",
+        "abort",
+        "finish"
+      ],
+      "expected": {
+        "target_criteria": [
+          "client_cancelled_abort_or_finish",
+          "client_abort_after_cancelled_finish"
+        ],
+        "target_raw_leaf_max": 0.0,
+        "criteria_profile": {
+          "server_early_abort_session": 1.0,
+          "client_stream_exit_after_abort": 1.0,
+          "client_abort_after_completed_transfer": 1.0,
+          "client_abort_idle_225": 1.0,
+          "client_abort_after_failed_transfer": 1.0,
+          "client_error_exit_mid_transfer": 1.0,
+          "client_error_exit_after_completed_transfer": 1.0,
+          "client_abort_idle_rejected": 1.0,
+          "client_upload_stream_abort": 1.0,
+          "client_abort_concurrent": 1.0,
+          "client_cancelled_abort_or_finish": 0.0,
+          "client_abort_during_finish": 1.0,
+          "client_abandoned_listing_resync": 1.0,
+          "server_listener_release_after_abort": 1.0,
+          "server_repeated_abor_order": 1.0,
+          "server_pipelined_command_order": 1.0,
+          "server_real_client_abor": 1.0,
+          "server_quit_during_transfer": 1.0,
+          "server_data_connection_loss": 1.0,
+          "server_undecodable_command_syntax": 1.0,
+          "client_abort_after_cancelled_finish": 0.0,
+          "server_pipelined_abort_open_data": 1.0,
+          "server_pipelined_quit_open_data": 1.0,
+          "client_nowait_abort_owed_replies": 1.0,
+          "client_command_during_abort": 1.0
+        },
+        "raw_score": 0.9385,
+        "final_score": 0.6,
+        "unaffected_criteria": [
+          "server_early_abort_session",
+          "client_stream_exit_after_abort",
+          "client_abort_after_completed_transfer",
+          "client_abort_idle_225",
+          "client_abort_after_failed_transfer",
+          "client_error_exit_mid_transfer",
+          "client_error_exit_after_completed_transfer",
+          "client_abort_idle_rejected",
+          "client_upload_stream_abort",
+          "client_abort_concurrent",
+          "client_abort_during_finish",
+          "client_abandoned_listing_resync",
+          "server_listener_release_after_abort",
+          "server_repeated_abor_order",
+          "server_pipelined_command_order",
+          "server_real_client_abor",
+          "server_quit_during_transfer",
+          "server_data_connection_loss",
+          "server_undecodable_command_syntax",
+          "server_pipelined_abort_open_data",
+          "server_pipelined_quit_open_data",
+          "client_nowait_abort_owed_replies",
+          "client_command_during_abort"
+        ],
+        "build_failed": false,
+        "safety_gates_failed": true,
+        "gate_statuses": {
+          "G1_stale_reply_after_abort": true,
+          "G2_stale_reply_after_error_exit": true,
+          "G3_stale_data_connection_after_abort": true,
+          "G4_stale_reply_after_cancelled_finish": false
+        }
+      }
+    },
+    {
+      "id": "ksg_abort_waits_for_control_lock",
+      "patch_path": "grading/adversarial/patches/ksg_abort_waits_for_control_lock.diff",
+      "applies_to": "gold",
+      "description": "Gold whose abort() writes ABOR only under the control lock, like every other command; cue tokens control lock, ABOR, abort and finish stay in code and comments. While finish() waits for the transfer's final reply the ABOR is never sent and both calls hang: only client_abort_during_finish is lost, raw 0.9932.",
+      "cue_tokens_required_present": [
+        "control lock",
+        "ABOR",
+        "abort",
+        "finish"
+      ],
+      "expected": {
+        "target_criteria": [
+          "client_abort_during_finish",
+          "client_command_during_abort"
+        ],
+        "target_raw_leaf_max": 0.0,
+        "criteria_profile": {
+          "server_early_abort_session": 1.0,
+          "client_stream_exit_after_abort": 1.0,
+          "client_abort_after_completed_transfer": 1.0,
+          "client_abort_idle_225": 1.0,
+          "client_abort_after_failed_transfer": 1.0,
+          "client_error_exit_mid_transfer": 1.0,
+          "client_error_exit_after_completed_transfer": 1.0,
+          "client_abort_idle_rejected": 1.0,
+          "client_upload_stream_abort": 1.0,
+          "client_abort_concurrent": 1.0,
+          "client_cancelled_abort_or_finish": 1.0,
+          "client_abort_during_finish": 0.0,
+          "client_abandoned_listing_resync": 1.0,
+          "server_listener_release_after_abort": 1.0,
+          "server_repeated_abor_order": 1.0,
+          "server_pipelined_command_order": 1.0,
+          "server_real_client_abor": 1.0,
+          "server_quit_during_transfer": 1.0,
+          "server_data_connection_loss": 1.0,
+          "server_undecodable_command_syntax": 1.0,
+          "client_abort_after_cancelled_finish": 1.0,
+          "server_pipelined_abort_open_data": 1.0,
+          "server_pipelined_quit_open_data": 1.0,
+          "client_nowait_abort_owed_replies": 1.0,
+          "client_command_during_abort": 0.0
+        },
+        "raw_score": 0.9379,
+        "final_score": 0.9379,
+        "unaffected_criteria": [
+          "server_early_abort_session",
+          "client_stream_exit_after_abort",
+          "client_abort_after_completed_transfer",
+          "client_abort_idle_225",
+          "client_abort_after_failed_transfer",
+          "client_error_exit_mid_transfer",
+          "client_error_exit_after_completed_transfer",
+          "client_abort_idle_rejected",
+          "client_upload_stream_abort",
+          "client_abort_concurrent",
+          "client_cancelled_abort_or_finish",
+          "client_abandoned_listing_resync",
+          "server_listener_release_after_abort",
+          "server_repeated_abor_order",
+          "server_pipelined_command_order",
+          "server_real_client_abor",
+          "server_quit_during_transfer",
+          "server_data_connection_loss",
+          "server_undecodable_command_syntax",
+          "client_abort_after_cancelled_finish",
+          "server_pipelined_abort_open_data",
+          "server_pipelined_quit_open_data",
+          "client_nowait_abort_owed_replies"
+        ],
+        "build_failed": false,
+        "safety_gates_failed": false,
+        "gate_statuses": {
+          "G1_stale_reply_after_abort": true,
+          "G2_stale_reply_after_error_exit": true,
+          "G3_stale_data_connection_after_abort": true,
+          "G4_stale_reply_after_cancelled_finish": true
+        }
+      }
+    },
+    {
+      "id": "ksg_abandoned_transfer_not_settled",
+      "patch_path": "grading/adversarial/patches/ksg_abandoned_transfer_not_settled.diff",
+      "applies_to": "gold",
+      "description": "Gold whose command() no longer settles an outstanding transfer and whose aclose() only closes the listing's data connection; cue tokens outstanding, settle, aclose, ABOR and get_stream stay in code and comments. Only client_abandoned_listing_resync is lost, raw 0.9945.",
+      "cue_tokens_required_present": [
+        "outstanding",
+        "settle",
+        "aclose",
+        "ABOR",
+        "get_stream"
+      ],
+      "expected": {
+        "target_criteria": [
+          "client_abandoned_listing_resync"
+        ],
+        "target_raw_leaf_max": 0.0,
+        "criteria_profile": {
+          "server_early_abort_session": 1.0,
+          "client_stream_exit_after_abort": 1.0,
+          "client_abort_after_completed_transfer": 1.0,
+          "client_abort_idle_225": 1.0,
+          "client_abort_after_failed_transfer": 1.0,
+          "client_error_exit_mid_transfer": 1.0,
+          "client_error_exit_after_completed_transfer": 1.0,
+          "client_abort_idle_rejected": 1.0,
+          "client_upload_stream_abort": 1.0,
+          "client_abort_concurrent": 1.0,
+          "client_cancelled_abort_or_finish": 1.0,
+          "client_abort_during_finish": 1.0,
+          "client_abandoned_listing_resync": 0.0,
+          "server_listener_release_after_abort": 1.0,
+          "server_repeated_abor_order": 1.0,
+          "server_pipelined_command_order": 1.0,
+          "server_real_client_abor": 1.0,
+          "server_quit_during_transfer": 1.0,
+          "server_data_connection_loss": 1.0,
+          "server_undecodable_command_syntax": 1.0,
+          "client_abort_after_cancelled_finish": 1.0,
+          "server_pipelined_abort_open_data": 1.0,
+          "server_pipelined_quit_open_data": 1.0,
+          "client_nowait_abort_owed_replies": 1.0,
+          "client_command_during_abort": 1.0
+        },
+        "raw_score": 0.9989,
+        "final_score": 0.9989,
+        "unaffected_criteria": [
+          "server_early_abort_session",
+          "client_stream_exit_after_abort",
+          "client_abort_after_completed_transfer",
+          "client_abort_idle_225",
+          "client_abort_after_failed_transfer",
+          "client_error_exit_mid_transfer",
+          "client_error_exit_after_completed_transfer",
+          "client_abort_idle_rejected",
+          "client_upload_stream_abort",
+          "client_abort_concurrent",
+          "client_cancelled_abort_or_finish",
+          "client_abort_during_finish",
+          "server_listener_release_after_abort",
+          "server_repeated_abor_order",
+          "server_pipelined_command_order",
+          "server_real_client_abor",
+          "server_quit_during_transfer",
+          "server_data_connection_loss",
+          "server_undecodable_command_syntax",
+          "client_abort_after_cancelled_finish",
+          "server_pipelined_abort_open_data",
+          "server_pipelined_quit_open_data",
+          "client_nowait_abort_owed_replies",
+          "client_command_during_abort"
+        ],
+        "build_failed": false,
+        "safety_gates_failed": false,
+        "gate_statuses": {
+          "G1_stale_reply_after_abort": true,
+          "G2_stale_reply_after_error_exit": true,
+          "G3_stale_data_connection_after_abort": true,
+          "G4_stale_reply_after_cancelled_finish": true
+        }
+      }
+    },
+    {
+      "id": "ksg_repeated_abor_immediate_226",
+      "patch_path": "grading/adversarial/patches/ksg_repeated_abor_immediate_226.diff",
+      "applies_to": "gold",
+      "description": "Gold whose abor handler answers a repeated ABOR at once instead of after the abort it repeats; cue tokens aborting_workers, 226, nothing to abort and ABOR stay in code and comments. The wire order becomes 226, 426, 226: only server_repeated_abor_order is lost, raw 0.9980.",
+      "cue_tokens_required_present": [
+        "aborting_workers",
+        "226",
+        "nothing to abort",
+        "ABOR"
+      ],
+      "expected": {
+        "target_criteria": [
+          "server_repeated_abor_order"
+        ],
+        "target_raw_leaf_max": 0.0,
+        "criteria_profile": {
+          "server_early_abort_session": 1.0,
+          "client_stream_exit_after_abort": 1.0,
+          "client_abort_after_completed_transfer": 1.0,
+          "client_abort_idle_225": 1.0,
+          "client_abort_after_failed_transfer": 1.0,
+          "client_error_exit_mid_transfer": 1.0,
+          "client_error_exit_after_completed_transfer": 1.0,
+          "client_abort_idle_rejected": 1.0,
+          "client_upload_stream_abort": 1.0,
+          "client_abort_concurrent": 1.0,
+          "client_cancelled_abort_or_finish": 1.0,
+          "client_abort_during_finish": 1.0,
+          "client_abandoned_listing_resync": 1.0,
+          "server_listener_release_after_abort": 1.0,
+          "server_repeated_abor_order": 0.0,
+          "server_pipelined_command_order": 1.0,
+          "server_real_client_abor": 1.0,
+          "server_quit_during_transfer": 1.0,
+          "server_data_connection_loss": 1.0,
+          "server_undecodable_command_syntax": 1.0,
+          "client_abort_after_cancelled_finish": 1.0,
+          "server_pipelined_abort_open_data": 1.0,
+          "server_pipelined_quit_open_data": 1.0,
+          "client_nowait_abort_owed_replies": 1.0,
+          "client_command_during_abort": 1.0
+        },
+        "raw_score": 0.99,
+        "final_score": 0.99,
+        "unaffected_criteria": [
+          "server_early_abort_session",
+          "client_stream_exit_after_abort",
+          "client_abort_after_completed_transfer",
+          "client_abort_idle_225",
+          "client_abort_after_failed_transfer",
+          "client_error_exit_mid_transfer",
+          "client_error_exit_after_completed_transfer",
+          "client_abort_idle_rejected",
+          "client_upload_stream_abort",
+          "client_abort_concurrent",
+          "client_cancelled_abort_or_finish",
+          "client_abort_during_finish",
+          "client_abandoned_listing_resync",
+          "server_listener_release_after_abort",
+          "server_pipelined_command_order",
+          "server_real_client_abor",
+          "server_quit_during_transfer",
+          "server_data_connection_loss",
+          "server_undecodable_command_syntax",
+          "client_abort_after_cancelled_finish",
+          "server_pipelined_abort_open_data",
+          "server_pipelined_quit_open_data",
+          "client_nowait_abort_owed_replies",
+          "client_command_during_abort"
+        ],
+        "build_failed": false,
+        "safety_gates_failed": false,
+        "gate_statuses": {
+          "G1_stale_reply_after_abort": true,
+          "G2_stale_reply_after_error_exit": true,
+          "G3_stale_data_connection_after_abort": true,
+          "G4_stale_reply_after_cancelled_finish": true
+        }
+      }
+    },
+    {
+      "id": "ksg_reply_order_fixed_state_racing",
+      "applies_to": "gold",
+      "cue_tokens_required_present": [
+        "response",
+        "dispatcher",
+        "create_task",
+        "restart_offset"
+      ],
+      "description": "Gold running commands concurrently as tasks without executor queue synchronization; pipelined state mutations race against subsequent commands on slow storage: server_pipelined_command_order is lost, raw 0.9693.",
+      "patch_path": "grading/adversarial/patches/ksg_reply_order_fixed_state_racing.diff",
+      "expected": {
+        "build_failed": false,
+        "criteria_profile": {
+          "server_early_abort_session": 1.0,
+          "client_stream_exit_after_abort": 1.0,
+          "client_abort_after_completed_transfer": 1.0,
+          "client_abort_idle_225": 1.0,
+          "client_abort_after_failed_transfer": 1.0,
+          "client_error_exit_mid_transfer": 1.0,
+          "client_error_exit_after_completed_transfer": 1.0,
+          "client_abort_idle_rejected": 1.0,
+          "client_upload_stream_abort": 1.0,
+          "client_abort_concurrent": 1.0,
+          "client_cancelled_abort_or_finish": 1.0,
+          "client_abort_during_finish": 1.0,
+          "client_abandoned_listing_resync": 1.0,
+          "server_listener_release_after_abort": 1.0,
+          "server_repeated_abor_order": 1.0,
+          "server_pipelined_command_order": 0.0,
+          "server_real_client_abor": 1.0,
+          "server_quit_during_transfer": 1.0,
+          "server_data_connection_loss": 1.0,
+          "server_undecodable_command_syntax": 1.0,
+          "client_abort_after_cancelled_finish": 1.0,
+          "server_pipelined_abort_open_data": 0.0,
+          "server_pipelined_quit_open_data": 0.0,
+          "client_nowait_abort_owed_replies": 1.0,
+          "client_command_during_abort": 1.0
+        },
+        "final_score": 0.6294,
+        "gate_statuses": {
+          "G1_stale_reply_after_abort": true,
+          "G2_stale_reply_after_error_exit": true,
+          "G3_stale_data_connection_after_abort": true,
+          "G4_stale_reply_after_cancelled_finish": true
+        },
+        "raw_score": 0.6294,
+        "safety_gates_failed": false,
+        "target_criteria": [
+          "server_pipelined_command_order",
+          "server_pipelined_abort_open_data",
+          "server_pipelined_quit_open_data"
+        ],
+        "target_raw_leaf_max": 0.0,
+        "unaffected_criteria": [
+          "server_early_abort_session",
+          "client_stream_exit_after_abort",
+          "client_abort_after_completed_transfer",
+          "client_abort_idle_225",
+          "client_abort_after_failed_transfer",
+          "client_error_exit_mid_transfer",
+          "client_error_exit_after_completed_transfer",
+          "client_abort_idle_rejected",
+          "client_upload_stream_abort",
+          "client_abort_concurrent",
+          "client_cancelled_abort_or_finish",
+          "client_abort_during_finish",
+          "client_abandoned_listing_resync",
+          "server_listener_release_after_abort",
+          "server_repeated_abor_order",
+          "server_real_client_abor",
+          "server_quit_during_transfer",
+          "server_data_connection_loss",
+          "server_undecodable_command_syntax",
+          "client_abort_after_cancelled_finish",
+          "client_nowait_abort_owed_replies",
+          "client_command_during_abort"
+        ]
+      }
+    },
+    {
+      "id": "ksg_telnet_stripped_no_urgent",
+      "applies_to": "gold",
+      "cue_tokens_required_present": [
+        "SO_OOBINLINE",
+        "IAC",
+        "parse_command",
+        "setsockopt"
+      ],
+      "description": "Gold stripping Telnet escapes in parse_command but omitting SO_OOBINLINE on the control socket; urgent-data ABOR sequences from real FTP clients corrupt the control stream or drop the session: server_real_client_abor is lost, raw 0.9744.",
+      "patch_path": "grading/adversarial/patches/ksg_telnet_stripped_no_urgent.diff",
+      "expected": {
+        "build_failed": false,
+        "criteria_profile": {
+          "server_early_abort_session": 1.0,
+          "client_stream_exit_after_abort": 1.0,
+          "client_abort_after_completed_transfer": 1.0,
+          "client_abort_idle_225": 1.0,
+          "client_abort_after_failed_transfer": 1.0,
+          "client_error_exit_mid_transfer": 1.0,
+          "client_error_exit_after_completed_transfer": 1.0,
+          "client_abort_idle_rejected": 1.0,
+          "client_upload_stream_abort": 1.0,
+          "client_abort_concurrent": 1.0,
+          "client_cancelled_abort_or_finish": 1.0,
+          "client_abort_during_finish": 1.0,
+          "client_abandoned_listing_resync": 1.0,
+          "server_listener_release_after_abort": 1.0,
+          "server_repeated_abor_order": 1.0,
+          "server_pipelined_command_order": 1.0,
+          "server_real_client_abor": 0.0,
+          "server_quit_during_transfer": 1.0,
+          "server_data_connection_loss": 1.0,
+          "server_undecodable_command_syntax": 1.0,
+          "client_abort_after_cancelled_finish": 1.0,
+          "server_pipelined_abort_open_data": 1.0,
+          "server_pipelined_quit_open_data": 1.0,
+          "client_nowait_abort_owed_replies": 1.0,
+          "client_command_during_abort": 1.0
+        },
+        "final_score": 0.9948,
+        "gate_statuses": {
+          "G1_stale_reply_after_abort": true,
+          "G2_stale_reply_after_error_exit": true,
+          "G3_stale_data_connection_after_abort": true,
+          "G4_stale_reply_after_cancelled_finish": true
+        },
+        "raw_score": 0.9948,
+        "safety_gates_failed": false,
+        "target_criteria": [
+          "server_real_client_abor"
+        ],
+        "target_raw_leaf_max": 0.0,
+        "unaffected_criteria": [
+          "server_early_abort_session",
+          "client_stream_exit_after_abort",
+          "client_abort_after_completed_transfer",
+          "client_abort_idle_225",
+          "client_abort_after_failed_transfer",
+          "client_error_exit_mid_transfer",
+          "client_error_exit_after_completed_transfer",
+          "client_abort_idle_rejected",
+          "client_upload_stream_abort",
+          "client_abort_concurrent",
+          "client_cancelled_abort_or_finish",
+          "client_abort_during_finish",
+          "client_abandoned_listing_resync",
+          "server_listener_release_after_abort",
+          "server_repeated_abor_order",
+          "server_pipelined_command_order",
+          "server_quit_during_transfer",
+          "server_data_connection_loss",
+          "server_undecodable_command_syntax",
+          "client_abort_after_cancelled_finish",
+          "server_pipelined_abort_open_data",
+          "server_pipelined_quit_open_data",
+          "client_nowait_abort_owed_replies",
+          "client_command_during_abort"
+        ]
+      }
+    },
+    {
+      "id": "ksg_quit_cancels_transfers_with_426",
+      "applies_to": "gold",
+      "cue_tokens_required_present": [
+        "quit",
+        "221",
+        "extra_workers",
+        "cancel"
+      ],
+      "description": "Gold whose quit() immediately cancels active data transfers with 426 instead of awaiting completion of open transfers with 226: server_quit_during_transfer is lost, raw 0.9761.",
+      "patch_path": "grading/adversarial/patches/ksg_quit_cancels_transfers_with_426.diff",
+      "expected": {
+        "build_failed": false,
+        "criteria_profile": {
+          "server_early_abort_session": 1.0,
+          "client_stream_exit_after_abort": 1.0,
+          "client_abort_after_completed_transfer": 1.0,
+          "client_abort_idle_225": 1.0,
+          "client_abort_after_failed_transfer": 1.0,
+          "client_error_exit_mid_transfer": 1.0,
+          "client_error_exit_after_completed_transfer": 1.0,
+          "client_abort_idle_rejected": 1.0,
+          "client_upload_stream_abort": 1.0,
+          "client_abort_concurrent": 1.0,
+          "client_cancelled_abort_or_finish": 1.0,
+          "client_abort_during_finish": 1.0,
+          "client_abandoned_listing_resync": 1.0,
+          "server_listener_release_after_abort": 1.0,
+          "server_repeated_abor_order": 1.0,
+          "server_pipelined_command_order": 1.0,
+          "server_real_client_abor": 1.0,
+          "server_quit_during_transfer": 0.0,
+          "server_data_connection_loss": 1.0,
+          "server_undecodable_command_syntax": 1.0,
+          "client_abort_after_cancelled_finish": 1.0,
+          "server_pipelined_abort_open_data": 1.0,
+          "server_pipelined_quit_open_data": 0.0,
+          "client_nowait_abort_owed_replies": 1.0,
+          "client_command_during_abort": 1.0
+        },
+        "final_score": 0.8737,
+        "gate_statuses": {
+          "G1_stale_reply_after_abort": true,
+          "G2_stale_reply_after_error_exit": true,
+          "G3_stale_data_connection_after_abort": true,
+          "G4_stale_reply_after_cancelled_finish": true
+        },
+        "raw_score": 0.8737,
+        "safety_gates_failed": false,
+        "target_criteria": [
+          "server_quit_during_transfer",
+          "server_pipelined_quit_open_data"
+        ],
+        "target_raw_leaf_max": 0.0,
+        "unaffected_criteria": [
+          "server_early_abort_session",
+          "client_stream_exit_after_abort",
+          "client_abort_after_completed_transfer",
+          "client_abort_idle_225",
+          "client_abort_after_failed_transfer",
+          "client_error_exit_mid_transfer",
+          "client_error_exit_after_completed_transfer",
+          "client_abort_idle_rejected",
+          "client_upload_stream_abort",
+          "client_abort_concurrent",
+          "client_cancelled_abort_or_finish",
+          "client_abort_during_finish",
+          "client_abandoned_listing_resync",
+          "server_listener_release_after_abort",
+          "server_repeated_abor_order",
+          "server_pipelined_command_order",
+          "server_real_client_abor",
+          "server_data_connection_loss",
+          "server_undecodable_command_syntax",
+          "client_abort_after_cancelled_finish",
+          "server_pipelined_abort_open_data",
+          "client_nowait_abort_owed_replies",
+          "client_command_during_abort"
+        ]
+      }
+    },
+    {
+      "id": "ksg_data_loss_no_listener_release",
+      "applies_to": "gold",
+      "cue_tokens_required_present": [
+        "ConnectionError",
+        "426",
+        "_release_passive_server",
+        "data_ports"
+      ],
+      "description": "Gold whose worker catches ConnectionError on the data stream and sends 426 but fails to release the passive listener and return the port to data_ports: server_data_connection_loss is lost, raw 0.9812.",
+      "patch_path": "grading/adversarial/patches/ksg_data_loss_no_listener_release.diff",
+      "expected": {
+        "build_failed": false,
+        "criteria_profile": {
+          "server_early_abort_session": 1.0,
+          "client_stream_exit_after_abort": 1.0,
+          "client_abort_after_completed_transfer": 1.0,
+          "client_abort_idle_225": 1.0,
+          "client_abort_after_failed_transfer": 1.0,
+          "client_error_exit_mid_transfer": 1.0,
+          "client_error_exit_after_completed_transfer": 1.0,
+          "client_abort_idle_rejected": 1.0,
+          "client_upload_stream_abort": 1.0,
+          "client_abort_concurrent": 1.0,
+          "client_cancelled_abort_or_finish": 1.0,
+          "client_abort_during_finish": 1.0,
+          "client_abandoned_listing_resync": 1.0,
+          "server_listener_release_after_abort": 1.0,
+          "server_repeated_abor_order": 1.0,
+          "server_pipelined_command_order": 1.0,
+          "server_real_client_abor": 1.0,
+          "server_quit_during_transfer": 1.0,
+          "server_data_connection_loss": 0.0,
+          "server_undecodable_command_syntax": 1.0,
+          "client_abort_after_cancelled_finish": 1.0,
+          "server_pipelined_abort_open_data": 1.0,
+          "server_pipelined_quit_open_data": 1.0,
+          "client_nowait_abort_owed_replies": 1.0,
+          "client_command_during_abort": 1.0
+        },
+        "final_score": 0.9962,
+        "gate_statuses": {
+          "G1_stale_reply_after_abort": true,
+          "G2_stale_reply_after_error_exit": true,
+          "G3_stale_data_connection_after_abort": true,
+          "G4_stale_reply_after_cancelled_finish": true
+        },
+        "raw_score": 0.9962,
+        "safety_gates_failed": false,
+        "target_criteria": [
+          "server_data_connection_loss"
+        ],
+        "target_raw_leaf_max": 0.0,
+        "unaffected_criteria": [
+          "server_early_abort_session",
+          "client_stream_exit_after_abort",
+          "client_abort_after_completed_transfer",
+          "client_abort_idle_225",
+          "client_abort_after_failed_transfer",
+          "client_error_exit_mid_transfer",
+          "client_error_exit_after_completed_transfer",
+          "client_abort_idle_rejected",
+          "client_upload_stream_abort",
+          "client_abort_concurrent",
+          "client_cancelled_abort_or_finish",
+          "client_abort_during_finish",
+          "client_abandoned_listing_resync",
+          "server_listener_release_after_abort",
+          "server_repeated_abor_order",
+          "server_pipelined_command_order",
+          "server_real_client_abor",
+          "server_quit_during_transfer",
+          "server_undecodable_command_syntax",
+          "client_abort_after_cancelled_finish",
+          "server_pipelined_abort_open_data",
+          "server_pipelined_quit_open_data",
+          "client_nowait_abort_owed_replies",
+          "client_command_during_abort"
+        ]
+      }
+    },
+    {
+      "id": "ksg_abor_queued_behind_transfer",
+      "applies_to": "gold",
+      "cue_tokens_required_present": [
+        "abor",
+        "extra_workers",
+        "wait",
+        "restart_offset"
+      ],
+      "description": "Gold whose command executor awaits active transfer workers before accepting the next command; an ABOR sent while a transfer is pending or running is not executed until that transfer ends, so it is answered late or never: server_early_abort_session, server_listener_release_after_abort, server_repeated_abor_order, server_pipelined_command_order, server_real_client_abor, server_quit_during_transfer are lost (G3 fires), raw 0.7554, final 0.60.",
+      "patch_path": "grading/adversarial/patches/ksg_abor_queued_behind_transfer.diff",
+      "expected": {
+        "build_failed": false,
+        "criteria_profile": {
+          "server_early_abort_session": 0.0,
+          "client_stream_exit_after_abort": 1.0,
+          "client_abort_after_completed_transfer": 1.0,
+          "client_abort_idle_225": 1.0,
+          "client_abort_after_failed_transfer": 1.0,
+          "client_error_exit_mid_transfer": 1.0,
+          "client_error_exit_after_completed_transfer": 1.0,
+          "client_abort_idle_rejected": 1.0,
+          "client_upload_stream_abort": 1.0,
+          "client_abort_concurrent": 1.0,
+          "client_cancelled_abort_or_finish": 1.0,
+          "client_abort_during_finish": 1.0,
+          "client_abandoned_listing_resync": 1.0,
+          "server_listener_release_after_abort": 0.0,
+          "server_repeated_abor_order": 0.0,
+          "server_pipelined_command_order": 0.0,
+          "server_real_client_abor": 0.0,
+          "server_quit_during_transfer": 0.0,
+          "server_data_connection_loss": 1.0,
+          "server_undecodable_command_syntax": 1.0,
+          "client_abort_after_cancelled_finish": 1.0,
+          "server_pipelined_abort_open_data": 0.0,
+          "server_pipelined_quit_open_data": 1.0,
+          "client_nowait_abort_owed_replies": 1.0,
+          "client_command_during_abort": 1.0
+        },
+        "final_score": 0.6,
+        "gate_statuses": {
+          "G1_stale_reply_after_abort": true,
+          "G2_stale_reply_after_error_exit": true,
+          "G3_stale_data_connection_after_abort": false,
+          "G4_stale_reply_after_cancelled_finish": true
+        },
+        "raw_score": 0.6976,
+        "safety_gates_failed": true,
+        "target_criteria": [
+          "server_early_abort_session",
+          "server_listener_release_after_abort",
+          "server_repeated_abor_order",
+          "server_pipelined_command_order",
+          "server_real_client_abor",
+          "server_quit_during_transfer",
+          "server_pipelined_abort_open_data"
+        ],
+        "target_raw_leaf_max": 0.0,
+        "unaffected_criteria": [
+          "client_stream_exit_after_abort",
+          "client_abort_after_completed_transfer",
+          "client_abort_idle_225",
+          "client_abort_after_failed_transfer",
+          "client_error_exit_mid_transfer",
+          "client_error_exit_after_completed_transfer",
+          "client_abort_idle_rejected",
+          "client_upload_stream_abort",
+          "client_abort_concurrent",
+          "client_cancelled_abort_or_finish",
+          "client_abort_during_finish",
+          "client_abandoned_listing_resync",
+          "server_data_connection_loss",
+          "server_undecodable_command_syntax",
+          "client_abort_after_cancelled_finish",
+          "server_pipelined_quit_open_data",
+          "client_nowait_abort_owed_replies",
+          "client_command_during_abort"
+        ]
+      }
+    }
+  ]
+}

harbor/sources/aioftp-v11/grading/baseline/pyproject.toml

@@ -0,0 +1,117 @@
+[project]
+name = "aioftp"
+version = "0.28.3"
+description = "ftp client/server for asyncio"
+readme = "README.rst"
+requires-python = ">= 3.10"
+license = {file = "license.txt"}
+authors = [
+    {name = "pohmelie", email = "multisosnooley@gmail.com"},
+    {name = "yieyu"},
+    {name = "rsichnyi"},
+    {name = "jw4js"},
+    {name = "asvetlov", email = "andrew.svetlov@gmail.com"},
+    {name = "decaz", email = "decaz89@gmail.com"},
+    {name = "oleksandr-kuzmenko"},
+    {name = "ndhansen"},
+    {name = "janneronkko", email="janne.ronkko@iki.fi"},
+    {name = "thirtyseven", email="ted@shlashdot.org"},
+    {name = "modelmat"},
+    {name = "greut"},
+    {name = "ported-pw", email="contact@ported.pw"},
+    {name = "PonyPC"},
+    {name = "jacobtomlinson"},
+    {name = "Olegt0rr", email="t0rr@mail.ru"},
+    {name = "michalc", email="michal@charemza.name"},
+    {name = "bachya"},
+    {name = "ch3pjw", email="paul@concertdaw.co.uk"},
+    {name = "puddly", email="puddly3@gmail.com"},
+    {name = "CrafterKolyan"},
+    {name = "jkr78"},
+    {name = "AMDmi3", email="amdmi3@amdmi3.ru"},
+    {name = "webknjaz", email="webknjaz+github/profile@redhat.com"},
+    {name = "rcfox", email="ryan@rcfox.ca"},
+    {name = "bellini666", email="thiago@bellini.dev"},
+    {name = "sakosha"},
+    {name = "rwakulszowa"},
+]
+classifiers = [
+    "Programming Language :: Python",
+    "Programming Language :: Python :: 3",
+    "Development Status :: 5 - Production/Stable",
+    "Topic :: Internet :: File Transfer Protocol (FTP)",
+]
+
+# TODO: remove `typing-extensions` when requires-python = ">= 3.11"
+dependencies = [
+    "typing-extensions>=4.13.2; python_version < '3.11'",  # 3.9-10 typing support
+]
+
+[project.urls]
+Github = "https://github.com/aio-libs/aioftp"
+Documentation = "https://aioftp.rtfd.io"
+
+[project.optional-dependencies]
+socks = [
+    "siosocks >= 0.2.0",
+]
+dev = [
+    # tests
+    "async_timeout >= 4.0.0",
+    "pytest-asyncio",
+    "pytest-cov",
+    "pytest-mock",
+    "pytest",
+    "siosocks",
+    "trustme",
+
+    # linters
+    "pre-commit",
+    "ruff",
+    "mypy",
+]
+
+[build-system]
+requires = ["setuptools", "wheel"]
+build-backend = "setuptools.build_meta"
+
+[tool.setuptools]
+packages.find.where = ["src"]
+package-data.aioftp = ["py.typed"]
+
+# tools
+[tool.ruff]
+line-length = 120
+target-version = "py310"
+lint.select = ["E", "W", "F", "Q", "UP", "I", "ASYNC"]
+lint.ignore = ["ASYNC240"]
+src = ["src"]
+exclude = ["ftpbench.py"]
+
+
+[tool.mypy]
+files = "src/aioftp"
+pretty = true
+strict = true
+ignore_missing_imports = true
+warn_unreachable = true
+
+[tool.coverage]
+run.source = ["./src/aioftp"]
+run.omit = ["./src/aioftp/__main__.py"]
+report.show_missing = true
+report.precision = 2
+
+[tool.pytest.ini_options]
+addopts = [
+    "-x",
+    "--durations", "10",
+    "-p", "no:anyio",
+    "--cov",
+    "--import-mode=importlib",
+]
+testpaths = "tests"
+log_format = "%(asctime)s.%(msecs)03d %(name)-20s %(levelname)-8s %(filename)-15s %(lineno)-4d %(message)s"
+log_date_format = "%H:%M:%S"
+log_level = "DEBUG"
+asyncio_mode = "strict"

harbor/sources/aioftp-v11/grading/baseline/tests/conftest.py

@@ -0,0 +1,232 @@
+import asyncio
+import collections
+import contextlib
+import functools
+import math
+import socket
+import ssl
+import tempfile
+import time
+from pathlib import Path
+
+import pytest
+import pytest_asyncio
+import trustme
+from async_timeout import timeout
+from siosocks.io.asyncio import socks_server_handler
+
+import aioftp
+
+# No ssl tests since https://bugs.python.org/issue36098
+ca = trustme.CA()
+server_cert = ca.issue_server_cert("127.0.0.1", "::1")
+
+ssl_server = ssl.create_default_context(ssl.Purpose.CLIENT_AUTH)
+server_cert.configure_cert(ssl_server)
+
+ssl_client = ssl.create_default_context(ssl.Purpose.SERVER_AUTH)
+ca.configure_trust(ssl_client)
+
+
+class Container:
+    def __init__(self, *args, **kwargs):
+        self.args = args
+        self.kwargs = kwargs
+
+
+@pytest.fixture
+def Client():
+    return Container
+
+
+@pytest.fixture
+def Server():
+    return Container
+
+
+def _wrap_with_defaults(kwargs):
+    test_defaults = dict(
+        path_io_factory=aioftp.MemoryPathIO,
+    )
+    return collections.ChainMap(kwargs, test_defaults)
+
+
+@pytest.fixture(params=["127.0.0.1", "::1"])
+def pair_factory(request):
+    class Factory:
+        def __init__(
+            self,
+            client=None,
+            server=None,
+            *,
+            connected=True,
+            logged=True,
+            do_quit=True,
+            host=request.param,
+            server_factory=aioftp.Server,
+            client_factory=aioftp.Client,
+        ):
+            if client is None:
+                client = Container()
+            self.client = client_factory(*client.args, **_wrap_with_defaults(client.kwargs))
+            if server is None:
+                server = Container()
+            self.server = server_factory(*server.args, **_wrap_with_defaults(server.kwargs))
+            self.connected = connected
+            self.logged = logged
+            self.do_quit = do_quit
+            self.host = host
+            self.timeout = timeout(1)
+
+        async def make_server_files(self, *paths, size=None, atom=b"-"):
+            if size is None:
+                size = aioftp.DEFAULT_BLOCK_SIZE * 3
+            data = atom * size
+            for p in paths:
+                await self.client.make_directory(Path(p).parent)
+                async with self.client.upload_stream(p) as stream:
+                    await stream.write(data)
+
+        async def make_client_files(self, *paths, size=None, atom=b"-"):
+            if size is None:
+                size = aioftp.DEFAULT_BLOCK_SIZE * 3
+            data = atom * size
+            for p in map(Path, paths):
+                await self.client.path_io.mkdir(
+                    p.parent,
+                    parents=True,
+                    exist_ok=True,
+                )
+                async with self.client.path_io.open(p, mode="wb") as f:
+                    await f.write(data)
+
+        async def server_paths_exists(self, *paths):
+            values = []
+            for p in paths:
+                values.append(await self.client.exists(p))
+            if all(values):
+                return True
+            if any(values):
+                raise ValueError("Mixed exists/not exists list")
+            return False
+
+        async def client_paths_exists(self, *paths):
+            values = []
+            for p in paths:
+                values.append(await self.client.path_io.exists(Path(p)))
+            if all(values):
+                return True
+            if any(values):
+                raise ValueError("Mixed exists/not exists list")
+            return False
+
+        async def __aenter__(self):
+            await self.timeout.__aenter__()
+            await self.server.start(host=self.host)
+            if self.connected:
+                await self.client.connect(
+                    self.server.server_host,
+                    self.server.server_port,
+                )
+                if self.logged:
+                    await self.client.login()
+            return self
+
+        async def __aexit__(self, *exc_info):
+            if self.connected and self.do_quit:
+                await self.client.quit()
+            self.client.close()
+            await self.server.close()
+            await self.timeout.__aexit__(*exc_info)
+
+    return Factory
+
+
+@pytest.fixture
+def expect_codes_in_exception():
+    @contextlib.contextmanager
+    def context(*codes):
+        try:
+            yield
+        except aioftp.StatusCodeError as e:
+            assert set(e.received_codes) == set(codes)
+        else:
+            raise RuntimeError("There was no exception")
+
+    return context
+
+
+@pytest.fixture(
+    params=[
+        aioftp.MemoryPathIO,
+        aioftp.PathIO,
+        aioftp.AsyncPathIO,
+    ],
+)
+def path_io(request):
+    return request.param()
+
+
+@pytest.fixture
+def temp_dir(path_io):
+    if isinstance(path_io, aioftp.MemoryPathIO):
+        yield Path("/")
+    else:
+        with tempfile.TemporaryDirectory() as name:
+            yield Path(name)
+
+
+class Sleep:
+    def __init__(self):
+        self.delay = 0
+        self.first_sleep = None
+
+    async def sleep(self, delay, result=None, **kwargs):
+        if self.first_sleep is None:
+            self.first_sleep = time.monotonic()
+        delay = (time.monotonic() - self.first_sleep) + delay
+        self.delay = max(self.delay, delay)
+        return result
+
+    def is_close(self, delay, *, rel_tol=0.05, abs_tol=0.5):
+        ok = math.isclose(self.delay, delay, rel_tol=rel_tol, abs_tol=abs_tol)
+        if not ok:
+            print(
+                f"latest sleep: {self.delay}; expected delay: {delay}; rel: {rel_tol}",
+            )
+        return ok
+
+
+@pytest.fixture
+def skip_sleep(monkeypatch):
+    with monkeypatch.context() as m:
+        sleeper = Sleep()
+        m.setattr(asyncio, "sleep", sleeper.sleep)
+        yield sleeper
+
+
+@pytest_asyncio.fixture(
+    params=[
+        ("127.0.0.1", socket.AF_INET),
+        ("::1", socket.AF_INET6),
+    ],
+)
+async def socks(request, unused_tcp_port):
+    handler = functools.partial(
+        socks_server_handler,
+        allowed_versions={5},
+        username="foo",
+        password="bar",
+    )
+    Socks = collections.namedtuple("Socks", "host port server")
+    host, family = request.param
+    port = unused_tcp_port
+    server = await asyncio.start_server(
+        handler,
+        host=host,
+        port=port,
+        family=family,
+    )
+    yield Socks(host, port, server)
+    server.close()
+    await server.wait_closed()

harbor/sources/aioftp-v11/grading/baseline/tests/test_abort.py

@@ -0,0 +1,96 @@
+import asyncio
+import itertools
+import pathlib
+
+import pytest
+
+import aioftp
+
+
+@pytest.mark.asyncio
+async def test_abort_stor(pair_factory):
+    async with pair_factory() as pair:
+        stream = await pair.client.upload_stream("test.txt")
+        with pytest.raises((ConnectionResetError, BrokenPipeError)):
+            while True:
+                await stream.write(b"-" * aioftp.DEFAULT_BLOCK_SIZE)
+                await pair.client.abort()
+
+
+class SlowReadMemoryPathIO(aioftp.MemoryPathIO):
+    async def read(self, *args, **kwargs):
+        await asyncio.sleep(0.01)
+        return await super().read(*args, **kwargs)
+
+
+@pytest.mark.asyncio
+async def test_abort_retr(pair_factory, Server):
+    s = Server(path_io_factory=SlowReadMemoryPathIO)
+    async with pair_factory(None, s) as pair:
+        await pair.make_server_files("test.txt")
+        stream = await pair.client.download_stream("test.txt")
+        for i in itertools.count():
+            data = await stream.read(aioftp.DEFAULT_BLOCK_SIZE)
+            if not data:
+                stream.close()
+                break
+            if i == 0:
+                await pair.client.abort()
+
+
+@pytest.mark.asyncio
+async def test_abort_retr_no_wait(
+    pair_factory,
+    Server,
+    expect_codes_in_exception,
+):
+    s = Server(path_io_factory=SlowReadMemoryPathIO)
+    async with pair_factory(None, s) as pair:
+        await pair.make_server_files("test.txt")
+        stream = await pair.client.download_stream("test.txt")
+        with expect_codes_in_exception("426"):
+            for i in itertools.count():
+                data = await stream.read(aioftp.DEFAULT_BLOCK_SIZE)
+                if not data:
+                    await stream.finish()
+                    break
+                if i == 0:
+                    await pair.client.abort(wait=False)
+
+
+@pytest.mark.asyncio
+async def test_nothing_to_abort(pair_factory):
+    async with pair_factory() as pair:
+        await pair.client.abort()
+
+
+class SlowListMemoryPathIO(aioftp.MemoryPathIO):
+    async def is_file(self, *a, **kw):
+        return True
+
+    def list(self, *args, **kwargs):
+        class Lister(aioftp.AbstractAsyncLister):
+            async def __anext__(cls):
+                await asyncio.sleep(0.01)
+                return pathlib.PurePath("/test.txt")
+
+        return Lister()
+
+    async def stat(self, *a, **kw):
+        class Stat:
+            st_size = 0
+            st_mtime = 0
+            st_ctime = 0
+
+        return Stat
+
+
+@pytest.mark.asyncio
+async def test_mlsd_abort(pair_factory, Server):
+    s = Server(path_io_factory=SlowListMemoryPathIO)
+    async with pair_factory(None, s) as pair:
+        cwd = await pair.client.get_current_directory()
+        assert cwd == pathlib.PurePosixPath("/")
+        async for path, info in pair.client.list():
+            await pair.client.abort()
+            break

harbor/sources/aioftp-v11/grading/baseline/tests/test_client_side_socks.py

@@ -0,0 +1,29 @@
+import pytest
+from siosocks.exceptions import SocksException
+
+
+@pytest.mark.asyncio
+async def test_socks_success(pair_factory, Client, socks):
+    client = Client(
+        socks_host=socks.host,
+        socks_port=socks.port,
+        socks_version=5,
+        username="foo",
+        password="bar",
+    )
+    async with pair_factory(client):
+        pass
+
+
+@pytest.mark.asyncio
+async def test_socks_fail(pair_factory, Client, socks):
+    client = Client(
+        socks_host=socks.host,
+        socks_port=socks.port,
+        socks_version=5,
+        username="bar",
+        password="bar",
+    )
+    with pytest.raises(SocksException):
+        async with pair_factory(client):
+            pass

harbor/sources/aioftp-v11/grading/baseline/tests/test_connection.py

@@ -0,0 +1,336 @@
+import asyncio
+import ipaddress
+
+import pytest
+
+import aioftp
+
+
+@pytest.mark.asyncio
+async def test_client_without_server(pair_factory, unused_tcp_port_factory):
+    f = pair_factory(connected=False, logged=False, do_quit=False)
+    async with f as pair:
+        pass
+    with pytest.raises(OSError):
+        await pair.client.connect("127.0.0.1", unused_tcp_port_factory())
+
+
+@pytest.mark.asyncio
+async def test_connection(pair_factory):
+    async with pair_factory(connected=True, logged=False, do_quit=False):
+        pass
+
+
+@pytest.mark.asyncio
+async def test_quit(pair_factory):
+    async with pair_factory(connected=True, logged=False, do_quit=True):
+        pass
+
+
+@pytest.mark.asyncio
+async def test_not_implemented(pair_factory, expect_codes_in_exception):
+    async with pair_factory() as pair:
+        with expect_codes_in_exception("502"):
+            await pair.client.command("FOOBAR", "2xx", "1xx")
+
+
+@pytest.mark.asyncio
+async def test_type_success(pair_factory, expect_codes_in_exception):
+    async with pair_factory() as pair:
+        await pair.client.get_passive_connection("A")
+
+
+@pytest.mark.asyncio
+async def test_custom_passive_commands(pair_factory):
+    async with pair_factory(host="127.0.0.1") as pair:
+        pair.client._passive_commands = None
+        await pair.client.get_passive_connection(
+            "A",
+            commands=["pasv", "epsv"],
+        )
+
+
+@pytest.mark.asyncio
+async def test_extra_pasv_connection(pair_factory):
+    async with pair_factory() as pair:
+        r, w = await pair.client.get_passive_connection()
+        er, ew = await pair.client.get_passive_connection()
+        with pytest.raises((ConnectionResetError, BrokenPipeError)):
+            while True:
+                w.write(b"-" * aioftp.DEFAULT_BLOCK_SIZE)
+                await w.drain()
+
+
+@pytest.mark.parametrize("method", ["epsv", "pasv"])
+@pytest.mark.asyncio
+async def test_closing_passive_connection(pair_factory, method):
+    async with pair_factory(host="127.0.0.1") as pair:
+        r, w = await pair.client.get_passive_connection(commands=[method])
+        host, port, *_ = w.transport.get_extra_info("peername")
+        nr, nw = await asyncio.open_connection(host, port)
+        with pytest.raises((ConnectionResetError, BrokenPipeError)):
+            while True:
+                nw.write(b"-" * aioftp.DEFAULT_BLOCK_SIZE)
+                await nw.drain()
+
+
+@pytest.mark.asyncio
+async def test_pasv_connection_ports_not_added(pair_factory):
+    async with pair_factory() as pair:
+        r, w = await pair.client.get_passive_connection()
+        assert pair.server.available_data_ports is None
+
+
+@pytest.mark.asyncio
+async def test_pasv_connection_ports(
+    pair_factory,
+    Server,
+    unused_tcp_port_factory,
+):
+    ports = [unused_tcp_port_factory(), unused_tcp_port_factory()]
+    async with pair_factory(None, Server(data_ports=ports)) as pair:
+        r, w = await pair.client.get_passive_connection()
+        host, port, *_ = w.transport.get_extra_info("peername")
+        assert port in ports
+        assert pair.server.available_data_ports.qsize() == 1
+
+
+@pytest.mark.asyncio
+async def test_data_ports_remains_empty(pair_factory, Server):
+    async with pair_factory(None, Server(data_ports=[])) as pair:
+        assert pair.server.available_data_ports.qsize() == 0
+
+
+@pytest.mark.asyncio
+async def test_pasv_connection_port_reused(
+    pair_factory,
+    Server,
+    unused_tcp_port,
+):
+    s = Server(data_ports=[unused_tcp_port])
+    async with pair_factory(None, s) as pair:
+        r, w = await pair.client.get_passive_connection()
+        host, port, *_ = w.transport.get_extra_info("peername")
+        assert port == unused_tcp_port
+        assert pair.server.available_data_ports.qsize() == 0
+        w.close()
+        await pair.client.quit()
+        pair.client.close()
+        assert pair.server.available_data_ports.qsize() == 1
+        await pair.client.connect(
+            pair.server.server_host,
+            pair.server.server_port,
+        )
+        await pair.client.login()
+        r, w = await pair.client.get_passive_connection()
+        host, port, *_ = w.transport.get_extra_info("peername")
+        assert port == unused_tcp_port
+        assert pair.server.available_data_ports.qsize() == 0
+
+
+@pytest.mark.asyncio
+async def test_pasv_connection_pasv_forced_response_address(pair_factory, Client, Server):
+    def ipv4_used():
+        try:
+            ipaddress.IPv4Address(pair.host)
+            return True
+        except ValueError:
+            return False
+
+    # using TEST-NET-1 address
+    ipv4_address = "192.0.2.1"
+    async with pair_factory(
+        client=Client(trust_server_pasv_ipv4_address=True),
+        server=Server(ipv4_pasv_forced_response_address=ipv4_address),
+    ) as pair:
+        assert pair.server.ipv4_pasv_forced_response_address == ipv4_address
+
+        if ipv4_used():
+            # The connection should fail here because the server starts to listen for
+            # the passive connections on the host (IPv4 address) that is used
+            # by the control channel. In reality, if the server is behind NAT,
+            # the server is reached with the defined external IPv4 address,
+            # i.e. we can check that the connection to
+            # pair.server.ipv4_pasv_forced_response_address failed to know that
+            # the server returned correct external IP
+            # ...
+            # but we can't use check like this:
+            # with pytest.raises(OSError):
+            #     await pair.client.get_passive_connection(commands=["pasv"])
+            # because there is no such ipv4 which will be non-routable, so
+            # we can only check `PASV` response
+            ip, _ = await pair.client._do_pasv()
+            assert ip == ipv4_address
+
+        # With epsv the connection should open as that does not use the
+        # external IPv4 address but just tells the client the port to connect
+        # to
+        await pair.client.get_passive_connection(commands=["epsv"])
+
+
+@pytest.mark.parametrize("method", ["epsv", "pasv"])
+@pytest.mark.asyncio
+async def test_pasv_connection_no_free_port(
+    pair_factory,
+    Server,
+    expect_codes_in_exception,
+    method,
+):
+    s = Server(data_ports=[])
+    async with pair_factory(None, s, do_quit=False, host="127.0.0.1") as pair:
+        assert pair.server.available_data_ports.qsize() == 0
+        with expect_codes_in_exception("421"):
+            await pair.client.get_passive_connection(commands=[method])
+
+
+@pytest.mark.asyncio
+async def test_pasv_connection_busy_port(
+    pair_factory,
+    Server,
+    unused_tcp_port_factory,
+):
+    ports = [unused_tcp_port_factory(), unused_tcp_port_factory()]
+    async with pair_factory(None, Server(data_ports=ports)) as pair:
+        conflicting_server = await asyncio.start_server(
+            lambda r, w: w.close(),
+            host=pair.server.server_host,
+            port=ports[0],
+        )
+        r, w = await pair.client.get_passive_connection()
+        host, port, *_ = w.transport.get_extra_info("peername")
+        assert port == ports[1]
+        assert pair.server.available_data_ports.qsize() == 1
+    conflicting_server.close()
+    await conflicting_server.wait_closed()
+
+
+@pytest.mark.asyncio
+async def test_pasv_connection_busy_port2(
+    pair_factory,
+    Server,
+    unused_tcp_port_factory,
+    expect_codes_in_exception,
+):
+    ports = [unused_tcp_port_factory()]
+    s = Server(data_ports=ports)
+    async with pair_factory(None, s, do_quit=False) as pair:
+        conflicting_server = await asyncio.start_server(
+            lambda r, w: w.close(),
+            host=pair.server.server_host,
+            port=ports[0],
+        )
+        with expect_codes_in_exception("421"):
+            await pair.client.get_passive_connection()
+    conflicting_server.close()
+    await conflicting_server.wait_closed()
+
+
+@pytest.mark.asyncio
+async def test_server_shutdown(pair_factory):
+    async with pair_factory(do_quit=False) as pair:
+        await pair.client.list()
+        await pair.server.close()
+        with pytest.raises(ConnectionResetError):
+            await pair.client.list()
+
+
+@pytest.mark.asyncio
+async def test_client_session_context_manager(pair_factory):
+    async with pair_factory(connected=False) as pair:
+        async with aioftp.Client.context(*pair.server.address) as client:
+            await client.list()
+
+
+@pytest.mark.asyncio
+async def test_long_login_sequence_fail(
+    pair_factory,
+    expect_codes_in_exception,
+):
+    class CustomServer(aioftp.Server):
+        def __init__(self, *args, **kwargs):
+            super().__init__(*args, **kwargs)
+            self.commands_mapping["acct"] = self.acct
+
+        async def user(self, connection, rest):
+            connection.response("331")
+            return True
+
+        async def pass_(self, connection, rest):
+            connection.response("332")
+            return True
+
+        async def acct(self, connection, rest):
+            connection.response("333")
+            return True
+
+    factory = pair_factory(
+        logged=False,
+        server_factory=CustomServer,
+        do_quit=False,
+    )
+    async with factory as pair:
+        with expect_codes_in_exception("333"):
+            await pair.client.login()
+
+
+@pytest.mark.asyncio
+async def test_bad_sublines_seq(pair_factory, expect_codes_in_exception):
+    class CustomServer(aioftp.Server):
+        async def write_response(self, stream, code, lines="", list=False):
+            import functools
+
+            lines = aioftp.wrap_with_container(lines)
+            write = functools.partial(self.write_line, stream)
+            *body, tail = lines
+            for line in body:
+                await write(code + "-" + line)
+            await write(str(int(code) + 1) + "-" + tail)
+            await write(code + " " + tail)
+
+    factory = pair_factory(connected=False, server_factory=CustomServer)
+    async with factory as pair:
+        with expect_codes_in_exception("220"):
+            await pair.client.connect(
+                pair.server.server_host,
+                pair.server.server_port,
+            )
+            await pair.client.login()
+
+
+@pytest.mark.asyncio
+async def test_welcome_message(pair_factory, Server):
+    welcome_message = "Welcome to the FTP server!"
+    async with pair_factory(
+        server=Server(welcome_message=welcome_message),
+        connected=False,
+    ) as pair:
+        lines = await pair.client.connect(
+            pair.server.server_host,
+            pair.server.server_port,
+        )
+        assert lines == [" " + welcome_message]
+
+
+@pytest.mark.asyncio
+async def test_system_type(pair_factory, Server):
+    system_type = "My custom OS"
+    async with pair_factory(
+        server=Server(system_type=system_type),
+    ) as pair:
+        resp = await pair.client.command("SYST", expected_codes=(215,))
+        assert resp == ("215", [" " + system_type])
+
+
+@pytest.mark.asyncio
+async def test_client_timeout_passive_connection(pair_factory, Client, monkeypatch):
+    open_connection = asyncio.open_connection
+
+    async def slow_open_connection(host, port, **kwargs):
+        await asyncio.sleep(1)
+        return await open_connection(host, port, **kwargs)
+
+    async with pair_factory(Client(connection_timeout=0.5), logged=True) as pair:
+        with monkeypatch.context() as m:
+            m.setattr("asyncio.open_connection", slow_open_connection)
+            with pytest.raises(asyncio.TimeoutError):
+                await pair.client.get_passive_connection()

harbor/sources/aioftp-v11/grading/baseline/tests/test_corner_cases.py

@@ -0,0 +1,77 @@
+from pathlib import Path
+
+import pytest
+
+import aioftp
+
+
+@pytest.mark.asyncio
+async def test_server_side_exception(pair_factory):
+    class CustomServer(aioftp.Server):
+        def __init__(self, *args, **kwargs):
+            super().__init__(*args, **kwargs)
+            self.commands_mapping["custom"] = self.custom
+
+        async def custom(*args, **kwargs):
+            raise RuntimeError("Test error")
+
+    factory = pair_factory(server_factory=CustomServer, do_quit=False)
+    async with factory as pair:
+        with pytest.raises(ConnectionResetError):
+            await pair.client.command("custom", "200")
+
+
+@pytest.mark.asyncio
+async def test_bad_type_value(pair_factory, expect_codes_in_exception):
+    async with pair_factory() as pair:
+        with expect_codes_in_exception("502"):
+            await pair.client.command("type FOO", "200")
+
+
+@pytest.mark.asyncio
+async def test_pbsz(pair_factory):
+    async with pair_factory() as pair:
+        await pair.client.command("pbsz", "200")
+
+
+@pytest.mark.asyncio
+async def test_prot(pair_factory, expect_codes_in_exception):
+    async with pair_factory() as pair:
+        await pair.client.command("prot P", "200")
+        with expect_codes_in_exception("502"):
+            await pair.client.command("prot foo", "200")
+
+
+@pytest.mark.asyncio
+async def test_server_ipv6_pasv(pair_factory, expect_codes_in_exception):
+    async with pair_factory(host="::1", do_quit=False) as pair:
+        with expect_codes_in_exception("503"):
+            await pair.client.get_passive_connection(commands=["pasv"])
+
+
+@pytest.mark.asyncio
+async def test_epsv_extra_arg(pair_factory, expect_codes_in_exception):
+    async with pair_factory(do_quit=False) as pair:
+        with expect_codes_in_exception("522"):
+            await pair.client.command("epsv foo", "229")
+
+
+@pytest.mark.asyncio
+async def test_bad_server_path_io(
+    pair_factory,
+    Server,
+    expect_codes_in_exception,
+):
+    class BadPathIO(aioftp.MemoryPathIO):
+        async def is_file(*a, **kw):
+            return False
+
+        async def is_dir(*a, **kw):
+            return False
+
+    s = Server(path_io_factory=BadPathIO)
+    async with pair_factory(None, s) as pair:
+        pio = pair.server.path_io_factory()
+        async with pio.open(Path("/foo"), "wb"):
+            pass
+        await pair.client.list()

harbor/sources/aioftp-v11/grading/baseline/tests/test_current_directory.py

@@ -0,0 +1,50 @@
+import pathlib
+
+import pytest
+
+import aioftp
+
+
+@pytest.mark.asyncio
+async def test_current_directory_simple(pair_factory):
+    async with pair_factory() as pair:
+        cwd = await pair.client.get_current_directory()
+        assert cwd == pathlib.PurePosixPath("/")
+
+
+@pytest.mark.asyncio
+async def test_current_directory_not_default(pair_factory, Server):
+    s = Server([aioftp.User(home_path="/home")])
+    async with pair_factory(None, s) as pair:
+        cwd = await pair.client.get_current_directory()
+        assert cwd == pathlib.PurePosixPath("/home")
+
+
+@pytest.mark.asyncio
+async def test_mlsd(pair_factory):
+    async with pair_factory() as pair:
+        await pair.make_server_files("test.txt")
+        (path, stat), *_ = files = await pair.client.list()
+        assert len(files) == 1
+        assert path == pathlib.PurePosixPath("test.txt")
+        assert stat["type"] == "file"
+
+
+@pytest.mark.asyncio
+async def test_resolving_double_dots(pair_factory):
+    async with pair_factory() as pair:
+        await pair.make_server_files("test.txt")
+
+        async def f():
+            cwd = await pair.client.get_current_directory()
+            assert cwd == pathlib.PurePosixPath("/")
+            (path, stat), *_ = files = await pair.client.list()
+            assert len(files) == 1
+            assert path == pathlib.PurePosixPath("test.txt")
+            assert stat["type"] == "file"
+
+        await f()
+        await pair.client.change_directory("../../../")
+        await f()
+        await pair.client.change_directory("/a/../b/..")
+        await f()