system stringlengths 1.23k 1.54k | user stringlengths 861 9.59k | think stringlengths 63 3.06k | answer stringlengths 21 6.29k |
|---|---|---|---|
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the required specs and checking the existing seed catalog.
Now let me read the Wave 6 catalog and its wave-wide laws.
Clean results: `-NonInteractive` mandatory-miss gives exit 1, empty stdout, one stderr line; parameter-set ambiguity reproduces for both call shapes. Next: switch semantics and `... | Wave 6, chunk 6P1 complete — all 6 catalog items authored, all gates green. These are the first PowerShell seeds in the repo (no `ps-*` collisions existed; pwsh 7.6.3 and the `powershell→pwsh` preflight mapping were already in place).
**ps-windowcast** — build-cli — `[CmdletBinding()]`/`param()` maintenance-window car... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-8 enterprise-integration seed authoring preamble
You are the sole seed author for one Wave-8 chunk. Repo root is the current
directory. These instructions are mandatory and supplement `tasks/SEED_SPEC.md`.
## Read before writing
1. Read all of `tasks/SEED_SPEC.md`.
2. Read the Wave-8 laws, evidence map, and a... | I'll start by reading the spec and catalog, then check for existing 8D seeds.
The layout is clear. Now the research phase — I'll search for the current official Confluent docs for all four interfaces.
Research is complete for all four seeds. Now let me inspect the existing Wave-8 C# seed for the xunit project layout ... | All gates are green. Wave-8 chunk 8D is complete — final report:
## Chunk 8D — 4/4 seeds authored, all gates passed
Every seed follows the Wave-8 laws: research-driven protected contract (`docs/contract.json` + `docs/official_sources.json` in `test_files`), hermetic loopback mocks with dummy credentials, injected sle... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-8 enterprise-integration seed authoring preamble
You are the sole seed author for one Wave-8 chunk. Repo root is the current
directory. These instructions are mandatory and supplement `tasks/SEED_SPEC.md`.
## Read before writing
1. Read all of `tasks/SEED_SPEC.md`.
2. Read the Wave-8 laws, evidence map, and a... | I'm the Wave-8 chunk 8A seed author. Let me start by reading the spec, the Wave-8 catalog, and checking the existing seeds.
Now let me check if any other Wave-8 seeds have already been authored (to see a wave-8-style layout), and look at existing project seed structure.
I'm the first Wave-8 author. Let me inspect one... | All four Wave-8 chunk 8A seeds are authored, audited `[complete]`, smoke-failing on baseline, and `[valid]` under `validate_seeds.py` (reference patch applies, verify passes, protected files untouched). No pre-existing seeds were modified; the only repo additions are the four new seed directories and their workspaces.
... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the required specs and checking the current state of the seed catalog.
Deprecation and generics lints behave as expected. Now checking Thread.stop availability, fallthrough/finally/serial lint texts, and this-escape triggers:
CS1998 didn't fire even in isolation — that's unexpected and matters f... | All 6 catalog items for chunk 5W4 are authored, self-checked, and gated. Final report:
**java-woolbale** — warnfix-java (5W4 #1): `-Xlint:all -Werror` generics retrofit of a wool-bale intake registry — raw-type sites ×5, unchecked casts/conversions, and a generic-varargs heap-pollution declaration whose three frozen c... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the required specs and checking the existing seed catalog.
Let me check the wave-4 Part I chunk details and the go-dashfeed example for how networked go seeds are structured.
Writing the go-lifegrid fixtures and reference implementation:
Now the task.json with the user-voice prompt:
go-lifegri... | Final report — wave 7, chunk 7G2, all 6 catalog items authored:
**go-lifegrid** — build-game — Conway's Life package with bounded and toroidal modes, `.cells` fixture loading, N-step evolution, and exact pinned ASCII frames (blinker/toad oscillation, glider torus lap in 24 steps, glider dying against a bounded corner ... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-8 enterprise-integration seed authoring preamble
You are the sole seed author for one Wave-8 chunk. Repo root is the current
directory. These instructions are mandatory and supplement `tasks/SEED_SPEC.md`.
## Read before writing
1. Read all of `tasks/SEED_SPEC.md`.
2. Read the Wave-8 laws, evidence map, and a... | I'm the Wave-8 chunk 8L author. Let me start by reading the spec, the catalog, and checking the existing seed IDs.
No existing jira seeds — all 4 items are missing. Now let me read the Wave-8 laws and chunk 8L catalog.
Now let me inspect a recently-authored Wave-8 seed for layout conventions.
The pattern is clear. L... | All four Wave-8 chunk 8L seeds are authored, validated, and gated. Final report:
## Chunk 8L — Atlassian Jira Cloud integrations (4/4 complete)
**Research.** No `jira-*` ID collisions existed; all catalog IDs used as-is. The Atlassian api-group reference pages are too large for WebFetch (truncated), so exact operatio... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the required specs and checking the existing seed catalog.
All six target ids are new. Now let me read the wave constraints and an example C++ seed.
Now let me look at an existing wave-7 C++ build seed as the formatting example.
Cribbage oracle values verified by hand-checking several. Now writ... | All six chunk-7G8 seeds are authored, self-checked, and gated.
**cpp-chessperft** — build-game — full-rules chess move generator (FEN load, sorted UCI legal-move lists, castling/EP/promotion/pins) verified by pinned perft counts at depths 1–4 over six fixture positions.
Smoke: `[ok ] cpp-chessperft: pre-fix verify fa... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-8 enterprise-integration seed authoring preamble
You are the sole seed author for one Wave-8 chunk. Repo root is the current
directory. These instructions are mandatory and supplement `tasks/SEED_SPEC.md`.
## Read before writing
1. Read all of `tasks/SEED_SPEC.md`.
2. Read the Wave-8 laws, evidence map, and a... | I'll start by reading the required spec and catalog files, then check existing seed IDs.
No ID collisions. Now reading the Wave-8 laws and chunk 8Z catalog.
Now I'll inspect exactly one existing Wave-8 seed for layout.
Let me look at the docs fixtures and a bit of the test harness in this same seed to understand the... | All gates pass. The other modified/untracked paths in git status belong to parallel authors and pre-existing work — my changes are confined to the four seed directories and their workspaces (which `validate_seeds.py` reset to baseline after proving the patches). No probes or temp artifacts remain, and per the integrity... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the required specs and checking the current state of the seed catalog.
Deprecation and generics lints behave as expected. Now checking Thread.stop availability, fallthrough/finally/serial lint texts, and this-escape triggers:
CS1998 didn't fire even in isolation — that's unexpected and matters f... | All 6 catalog items for chunk 5W4 are authored, self-checked, and gated. Final report:
**java-woolbale** — warnfix-java (5W4 #1): `-Xlint:all -Werror` generics retrofit of a wool-bale intake registry — raw-type sites ×5, unchecked casts/conversions, and a generic-varargs heap-pollution declaration whose three frozen c... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-8 enterprise-integration seed authoring preamble
You are the sole seed author for one Wave-8 chunk. Repo root is the current
directory. These instructions are mandatory and supplement `tasks/SEED_SPEC.md`.
## Read before writing
1. Read all of `tasks/SEED_SPEC.md`.
2. Read the Wave-8 laws, evidence map, and a... | I'll start by reading the required spec and catalog files, and checking existing seed IDs.
No ID collisions — none of the four Meraki IDs exist. Now let me read the Wave-8 catalog and laws.
Now I'll inspect exactly one existing Wave-8 seed for layout (a Go project one, relevant to my chunk).
Let me look at the rest ... | All gates are green. Final report:
# Wave-8 chunk 8Y — Cisco Meraki integrations: 4/4 seeds authored
All four catalog items are authored, smoke-checked, reference-proven, and validated (`validate_seeds.py --only …` run exactly once: **4 valid, 0 invalid**). No pre-existing seeds, pipeline code, catalogs, or traces we... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the required specs and checking the existing seed catalog.
All six target ids are new. Now let me read the wave constraints and an example C++ seed.
Now let me look at an existing wave-7 C++ build seed as the formatting example.
Cribbage oracle values verified by hand-checking several. Now writ... | All six chunk-7G8 seeds are authored, self-checked, and gated.
**cpp-chessperft** — build-game — full-rules chess move generator (FEN load, sorted UCI legal-move lists, castling/EP/promotion/pins) verified by pinned perft counts at depths 1–4 over six fixture positions.
Smoke: `[ok ] cpp-chessperft: pre-fix verify fa... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'm authoring the 6 x86-64 assembly seeds for wave 7 chunk 7A1. First, let me read the required docs and survey existing seeds.
The pipeline already supports `asm` (preflight checks gcc). Let me look at a recent wave-7 C build seed for format conventions.
Blackjack pins done (order-sensitive rollhash noted). Now the ... | All six seeds audit `[complete]`. Work is done — final report below.
## Wave 7 / Chunk 7A1 — final report
| id | kind-category | description | smoke | reference verify |
|---|---|---|---|---|
| asm-lifestep | build-game | One Life generation over a row-major byte grid, toroidal wrap with the multi-count tiny-grid law... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the required specs and checking the existing seed catalog.
Let me check the wave-4 Part I chunk details and the go-dashfeed example for how networked go seeds are structured.
Writing the go-lifegrid fixtures and reference implementation:
Now the task.json with the user-voice prompt:
go-lifegri... | Final report — wave 7, chunk 7G2, all 6 catalog items authored:
**go-lifegrid** — build-game — Conway's Life package with bounded and toroidal modes, `.cells` fixture loading, N-step evolution, and exact pinned ASCII frames (blinker/toad oscillation, glider torus lap in 24 steps, glider dying against a bounded corner ... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the required docs and checking the existing seed catalog.
JDK 26 is available. Let me set up task tracking and look at examples of java build seeds and a full wave-7 game prompt.
Now the test contract:
One correction — my "landing in opponent's empty pit" test needs 4 stones in pit 6 (store + b... | All 6 catalog items for chunk 7G6 are authored, self-checked, and round-trip verified.
**Final report — wave 7, chunk 7G6 (java ×6, all `build-game`)**
| id | idx | description | smoke | reference verify |
|---|---|---|---|---|
| java-adventure | 1 | OO text-adventure engine: plain-text room-graph format, typed Room/... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | Data-platform ticket. Our telemetry archiver pushes nightly rollup blobs to S3 with a single put_object call, and last night a 3.4 GB rollup timed out four times and the job gave up — we resend the whole object every time anything hiccups. I want a proper multipart uploader package, `s3mpu`, built on the official AWS S... | I'll start by reading the workspace fixtures and tests, then do the required official-docs research before writing any code.
Fixtures and tests reviewed. Now the required research — loading WebSearch/WebFetch and reading the current official AWS docs before writing any code.
The live docs check out against the fixtur... | Done — `bash run_verify.sh` is green (14 passed), and the review finding that caused the last rejection is fixed and verified with an adversarial probe.
**What was built.** A new `s3mpu/` package, split into `constants.py` (the pinned documented values), `planning.py` (`select_strategy` and `plan_parts` with the 1-bas... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-8 enterprise-integration seed authoring preamble
You are the sole seed author for one Wave-8 chunk. Repo root is the current
directory. These instructions are mandatory and supplement `tasks/SEED_SPEC.md`.
## Read before writing
1. Read all of `tasks/SEED_SPEC.md`.
2. Read the Wave-8 laws, evidence map, and a... | I'm the Wave-8 chunk 8R author. Let me start by reading the spec, catalog, and checking the existing seeds.
All 4 chunk 8R items are missing. Now let me read the Wave-8 laws and evidence map at the top of the catalog.
All four 8R slots are open. Let me set up task tracking and inspect representative Wave-8 seeds in e... | # Wave-8 chunk 8R final report — all 4 seeds authored and validated
All four missing catalog items are authored, smoke-failing on baseline, reference-proven, and `validate_seeds.py` reports **4 valid, 0 invalid**. No existing seeds, pipeline code, or catalogs were touched; no traces were generated.
**Research (shared... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | Feature request for our Jira retention tooling. The existing ProjectAdminClient does stage one fine — it looks up project state and archives projects that hit end-of-retention (getProject/archiveProject, both in production). Stage two is still a human with a browser: when legal signs off on final purge, someone clicks ... | I'll start by exploring the repo to understand the existing code and contract fixtures, and load the research tools.
I have the full picture of the code and pinned contract. Now the required research phase — I'll search for and read the current official Atlassian docs before touching any source.
All 8 tests pass. Cle... | **Done — `java TestMain.java` passes all 8 tests**, covering both the pinned `getProject`/`archiveProject` behavior and the new purge flow. The only change to the tree is the new `ProjectRetirement.java`; TestMain.java and docs/ are untouched, and all scratch artifacts (downloaded spec, fetched intro page) were removed... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | Ingest-platform ticket. Our camera-frame pipeline pulls from Pub/Sub with a hand-rolled loop somebody wrote years ago: it acks every message the moment it arrives (so a crashed decoder silently eats frames), it has no flow control (a backlog OOMs the worker), long decodes blow past the ack deadline and get redelivered ... | I'll start by looking at the workspace — the test contract, fixtures, and requirements — then do the required official-docs research before writing any code.
Now let me look at the protected fixtures in docs/, and kick off the venv setup in the background.
Two more official pages to confirm the hard modifyAckDeadline... | Done — `bash run_verify.sh` passes all 13 acceptance tests from the pinned venv (google-cloud-pubsub 2.39.0). The only new artifact is the `pubstream/` package; all protected files (test_stream.py, run_verify.sh, requirements.txt, docs/) are untouched.
**Research result:** I read the live official docs before writing ... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-8 enterprise-integration seed authoring preamble
You are the sole seed author for one Wave-8 chunk. Repo root is the current
directory. These instructions are mandatory and supplement `tasks/SEED_SPEC.md`.
## Read before writing
1. Read all of `tasks/SEED_SPEC.md`.
2. Read the Wave-8 laws, evidence map, and a... | I'm the Wave-8 chunk 8G author. Let me start by reading the spec, the Wave-8 catalog, and checking the current seed inventory.
Now fetching the official docs pages in parallel — the API overview plus the four operation pages.
Workspace materialized. Writing the reference implementation (4 source files: client, config... | All four Wave-8 chunk 8G seeds are authored, validated, and gate-clean. Every gate passed: baseline smoke fails on all four (checked twice), `validate_seeds.py` reports all four valid (reference patch applies and verify passes), `audit_seeds.py` reports all four complete, and the fail-closed prompt audit confirms the l... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-8 enterprise-integration seed authoring preamble
You are the sole seed author for one Wave-8 chunk. Repo root is the current
directory. These instructions are mandatory and supplement `tasks/SEED_SPEC.md`.
## Read before writing
1. Read all of `tasks/SEED_SPEC.md`.
2. Read the Wave-8 laws, evidence map, and a... | I'm the Wave-8 chunk 8F author. Let me start by reading the spec, the Wave-8 catalog, and checking the current state of the seeds directory.
The four IDs are free. Now let me study existing Wave-8 seeds in the same languages as templates (layout only).
Let me look at the docs fixtures and test-harness patterns from t... | Wave-8 chunk 8F is complete: all four missing catalog items are authored, validated, and gated. Every seed shipped with researched provenance, protected `docs/contract.json` + `docs/official_sources.json` fixtures (research object byte-equal to `task.json`), a loopback-mock acceptance harness, a failing baseline, and a... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the required specs and checking the existing seed catalog.
Now let me read the wave-4 additional constraints and the wave-6 catalog chunk.
Let me check which chunk assignments existing ps-* seeds already occupy, and look at an existing ps seed as a formatting example.
No 6P8 seeds exist yet — I... | All six 6P8 seeds authored, self-checked, and validated. Final report:
## Wave 6, chunk 6P8 — final report
| id | kind-category | description | smoke | reference verify |
|---|---|---|---|---|
| **ps-shiftbrief** | build / perf-strings | Feasibility gate: shift-log→single-string brief renderer over a 500k-line fixtur... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the required docs and checking the existing seed catalog.
JDK 26 is available. Let me set up task tracking and look at examples of java build seeds and a full wave-7 game prompt.
Now the test contract:
One correction — my "landing in opponent's empty pit" test needs 4 stones in pit 6 (store + b... | All 6 catalog items for chunk 7G6 are authored, self-checked, and round-trip verified.
**Final report — wave 7, chunk 7G6 (java ×6, all `build-game`)**
| id | idx | description | smoke | reference verify |
|---|---|---|---|---|
| java-adventure | 1 | OO text-adventure engine: plain-text room-graph format, typed Room/... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-8 enterprise-integration seed authoring preamble
You are the sole seed author for one Wave-8 chunk. Repo root is the current
directory. These instructions are mandatory and supplement `tasks/SEED_SPEC.md`.
## Read before writing
1. Read all of `tasks/SEED_SPEC.md`.
2. Read the Wave-8 laws, evidence map, and a... | I'm the Wave-8 chunk 8L author. Let me start by reading the spec, the catalog, and checking the existing seed IDs.
No existing jira seeds — all 4 items are missing. Now let me read the Wave-8 laws and chunk 8L catalog.
Now let me inspect a recently-authored Wave-8 seed for layout conventions.
The pattern is clear. L... | All four seeds pass full validation. Final housekeeping: assertion counts, scratch cleanup, and a scope check: |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-8 enterprise-integration seed authoring preamble
You are the sole seed author for one Wave-8 chunk. Repo root is the current
directory. These instructions are mandatory and supplement `tasks/SEED_SPEC.md`.
## Read before writing
1. Read all of `tasks/SEED_SPEC.md`.
2. Read the Wave-8 laws, evidence map, and a... | I'm the Wave-8 chunk 8A seed author. Let me start by reading the spec, the Wave-8 catalog, and checking the existing seeds.
Now let me check if any other Wave-8 seeds have already been authored (to see a wave-8-style layout), and look at existing project seed structure.
I'm the first Wave-8 author. Let me inspect one... | All four seeds now pass the audit. Final combined smoke plus the patch-based validation: |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-8 enterprise-integration seed authoring preamble
You are the sole seed author for one Wave-8 chunk. Repo root is the current
directory. These instructions are mandatory and supplement `tasks/SEED_SPEC.md`.
## Read before writing
1. Read all of `tasks/SEED_SPEC.md`.
2. Read the Wave-8 laws, evidence map, and a... | I'm the Wave-8 chunk 8A seed author. Let me start by reading the spec, the Wave-8 catalog, and checking the existing seeds.
Now let me check if any other Wave-8 seeds have already been authored (to see a wave-8-style layout), and look at existing project seed structure.
I'm the first Wave-8 author. Let me inspect one... | All four validate. Final scope check and assertion counts for the report: |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-8 enterprise-integration seed authoring preamble
You are the sole seed author for one Wave-8 chunk. Repo root is the current
directory. These instructions are mandatory and supplement `tasks/SEED_SPEC.md`.
## Read before writing
1. Read all of `tasks/SEED_SPEC.md`.
2. Read the Wave-8 laws, evidence map, and a... | I'm the Wave-8 chunk 8F author. Let me start by reading the spec, the Wave-8 catalog, and checking the current state of the seeds directory.
The four IDs are free. Now let me study existing Wave-8 seeds in the same languages as templates (layout only).
Let me look at the docs fixtures and test-harness patterns from t... | All four seeds audit as complete. Final consistency sweep on the patches and metadata. |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-8 enterprise-integration seed authoring preamble
You are the sole seed author for one Wave-8 chunk. Repo root is the current
directory. These instructions are mandatory and supplement `tasks/SEED_SPEC.md`.
## Read before writing
1. Read all of `tasks/SEED_SPEC.md`.
2. Read the Wave-8 laws, evidence map, and a... | I'm the Wave-8 chunk 8A seed author. Let me start by reading the spec, the Wave-8 catalog, and checking the existing seeds.
Now let me check if any other Wave-8 seeds have already been authored (to see a wave-8-style layout), and look at existing project seed structure.
I'm the first Wave-8 author. Let me inspect one... | The audit wants specific wording in prompts. Checking its exact requirements: |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-8 enterprise-integration seed authoring preamble
You are the sole seed author for one Wave-8 chunk. Repo root is the current
directory. These instructions are mandatory and supplement `tasks/SEED_SPEC.md`.
## Read before writing
1. Read all of `tasks/SEED_SPEC.md`.
2. Read the Wave-8 laws, evidence map, and a... | I'm the Wave-8 chunk 8A seed author. Let me start by reading the spec, the Wave-8 catalog, and checking the existing seeds.
Now let me check if any other Wave-8 seeds have already been authored (to see a wave-8-style layout), and look at existing project seed structure.
I'm the first Wave-8 author. Let me inspect one... | The audit requires literal phrases ("websearch", "webfetch", "before changing code") in prompts. Updating all four: |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-8 enterprise-integration seed authoring preamble
You are the sole seed author for one Wave-8 chunk. Repo root is the current
directory. These instructions are mandatory and supplement `tasks/SEED_SPEC.md`.
## Read before writing
1. Read all of `tasks/SEED_SPEC.md`.
2. Read the Wave-8 laws, evidence map, and a... | I'm the Wave-8 chunk 8A seed author. Let me start by reading the spec, the Wave-8 catalog, and checking the existing seeds.
Now let me check if any other Wave-8 seeds have already been authored (to see a wave-8-style layout), and look at existing project seed structure.
I'm the first Wave-8 author. Let me inspect one... | Now the csharp task.json: |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-8 enterprise-integration seed authoring preamble
You are the sole seed author for one Wave-8 chunk. Repo root is the current
directory. These instructions are mandatory and supplement `tasks/SEED_SPEC.md`.
## Read before writing
1. Read all of `tasks/SEED_SPEC.md`.
2. Read the Wave-8 laws, evidence map, and a... | I'm the Wave-8 chunk 8L author. Let me start by reading the spec, the catalog, and checking the existing seed IDs.
No existing jira seeds — all 4 items are missing. Now let me read the Wave-8 laws and chunk 8L catalog.
Now let me inspect a recently-authored Wave-8 seed for layout conventions.
The pattern is clear. L... | All 6 C# tests pass. Saving the patch and running the final gate checks: |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-8 enterprise-integration seed authoring preamble
You are the sole seed author for one Wave-8 chunk. Repo root is the current
directory. These instructions are mandatory and supplement `tasks/SEED_SPEC.md`.
## Read before writing
1. Read all of `tasks/SEED_SPEC.md`.
2. Read the Wave-8 laws, evidence map, and a... | I'm the Wave-8 chunk 8A seed author. Let me start by reading the spec, the Wave-8 catalog, and checking the existing seeds.
Now let me check if any other Wave-8 seeds have already been authored (to see a wave-8-style layout), and look at existing project seed structure.
I'm the first Wave-8 author. Let me inspect one... | Baseline fails on the missing implementation. Now the reference implementation: |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-8 enterprise-integration seed authoring preamble
You are the sole seed author for one Wave-8 chunk. Repo root is the current
directory. These instructions are mandatory and supplement `tasks/SEED_SPEC.md`.
## Read before writing
1. Read all of `tasks/SEED_SPEC.md`.
2. Read the Wave-8 laws, evidence map, and a... | I'm the Wave-8 chunk 8L author. Let me start by reading the spec, the catalog, and checking the existing seed IDs.
No existing jira seeds — all 4 items are missing. Now let me read the Wave-8 laws and chunk 8L catalog.
Now let me inspect a recently-authored Wave-8 seed for layout conventions.
The pattern is clear. L... | All patches are clean (source files only). Running the mandatory re-smoke (step 3) and checking for a seed validator: |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-8 enterprise-integration seed authoring preamble
You are the sole seed author for one Wave-8 chunk. Repo root is the current
directory. These instructions are mandatory and supplement `tasks/SEED_SPEC.md`.
## Read before writing
1. Read all of `tasks/SEED_SPEC.md`.
2. Read the Wave-8 laws, evidence map, and a... | I'll start by reading the spec, the Wave-8 catalog/laws, and checking existing seeds for collisions.
Now let me look at the contract fixture format and the C# wave-8 seed layout (for `dotnet test` conventions).
Let me set up task tracking, then look at the C# csproj and remaining conventions, and load the web researc... | All four missing Chunk 8H seeds are authored, validated, and gated. Every gate passed: baseline smoke fails for the right reason, `validate_seeds.py` reports all four valid (patch applies, verify passes, protected files untouched), re-smoke still fails on fresh baselines, and the audit lists all four as complete. No ID... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the required spec files and checking the existing seed catalog.
Now let me read the wave-4 constraints and the wave-5 catalog chunk.
The file contains bytes that make grep treat it as binary. Let me read it with `grep -a`.
The plan is set. Let me check `materialize` to understand workspace mech... | All gates pass. Final report:
## Wave 5, chunk 5S3 — 6/6 seeds authored
| # | id | kind-category | description |
|---|----|----|----|
| 1 | `java-lotsheet` | debug / syntax-java | Anonymous-class `Comparator` brace interleave off by one in an auction lot sheet — parse error lands 55 lines later at the class-closing b... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the required specs and checking the existing seed catalog.
The chunk heading grep came up empty — let me search the wave-5 catalog differently.
The grep for "5E" found nothing even though the intro mentions it — possibly unusual characters. Let me look at the file structure directly.
Now let me... | Wave 5, chunk 5E1 complete — all 6 catalog items authored, self-checked, and gated. Final report:
**py-gaugelog** — debug, `escape-regex` (catalog index 1)
River-gauge log parser whose regex patterns live in plain strings (`"\d+"`, `"\s*#"`); runs fine under plain python3, dies at import under the CI's `python3 -W err... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the required specs and checking the existing seed catalog.
The chunk 5C2 grep didn't hit — let me check WAVE5_USECASES.md directly.
Odd — the heading is visible in `head` but grep missed it. Possibly unusual characters. Let me check.
Now the task.json (placeholder prompt until I capture the rea... | Chunk 5C2 complete — all 6 catalog items authored, self-checked, and validated (`validate_seeds.py`: 6 valid, 0 invalid; `audit_seeds.py`: 568 complete, 0 partial; nothing outside my seed dirs touched).
## Final report — wave 5, chunk 5C2
**ts-bikedock** (index 1) — debug / compilefix-ts — Bike-share dock rebalancing... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'm authoring wave-5 chunk 5S4 (6 seeds: bash×2, zsh×1, C×2, C++×1). Starting by reading the spec and catalog.
Good pattern. Now let me probe toolchain versions and verify my planned zsh/bash error mechanics empirically with inline commands.
Item 2 mechanics confirmed. Redoing the unclosed-quote probe cleanly via std... | All six seeds are complete, self-checked, and in a pristine post-smoke state. Final report:
**Chunk 5S4 — six seeds authored** (all `debug` kind; every `task.json` carries `"catalog": {"wave": 5, "chunk": "5S4", "index": N}`)
1. **bash-poolsheet** — syntax-bash — weekly aquatic-centre sheet renderer; one quote missin... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the required specs and checking the current state of the seed catalog.
Variant B silently swallows to the next terminator — good material. Now probe the missing-terminator EOF report and hashtable brace:
Variant C gives a displaced "missing terminator" report. Now let me probe the item-2 defects... | All six 6P10 seeds are authored, validated, and gated. Final report:
## Wave 6, chunk 6P10 — final report
**ps-kilnsheet** — debug / `syntax-powershell` (item 1). Two-file broken-parse triage: here-string terminator indented inside a function swallows the closing brace and half the program until a later column-1 `'@`... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'm authoring the 6 missing catalog items for wave 5, chunk 5S1. Starting by reading the spec and catalog.
No wave-5 seeds exist yet — all 6 items need authoring. Checking lang-field conventions and example debug seeds for go/ts.
One-line displacement with a helpful message — not good enough. Probing variants for dee... | All six seeds authored, self-checked, and validated. Final report:
**Seed 1 — py-menucost** (debug, syntax-py) — catering order-desk pricing; unclosed paren in a dict-comprehension chain, fix is one char (`)` after `MARGIN`).
- Smoke: `[ok ] py-menucost: pre-fix verify fails as expected (SyntaxError: closing parenthe... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the specs and checking the existing catalog state.
No 6P7 seeds exist yet — all 6 items need authoring. Let me read the wave laws and the chunk section.
Let me examine an existing refactor seed to understand the mechanics.
Pipeline understood. Now let me probe the pwsh behaviors these seeds hin... | Everything checks out. All six chunk-6P7 seeds are authored, self-checked, and gated.
## Final report — wave 6, chunk 6P7 (6/6 authored)
| id | kind-category | description |
|---|---|---|
| ps-depotops | refactor-module | 268-line spare-parts depot ops monolith (report/shortfall/audit CLI) split into `DepotOps.psm1` ... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'm the wave-5 chunk 5W1 author. Let me start by reading the required docs and surveying the existing seed catalog.
Python 3.13.11 — `assertEquals` is fully removed (not just deprecated), and ssl constants don't warn on mere access. Let me probe the actual warning-emission paths.
Now authoring **seed 1: go-bunkboard*... | Wave 5 chunk 5W1 complete — all 6 catalog items authored, self-checked, and validated (`validate_seeds.py --only <ids>`: 6 valid, 0 invalid).
**go-bunkboard** — debug, warnfix-go — hostel occupancy board failing a gofmt+vet+test CI gate: printf verb mismatches ×3, struct tag typos (`jsn:` pinned by tests, malformed `,... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'm the wave-5 chunk 5W1 author. Let me start by reading the required docs and surveying the existing seed catalog.
Python 3.13.11 — `assertEquals` is fully removed (not just deprecated), and ssl constants don't warn on mere access. Let me probe the actual warning-emission paths.
Now authoring **seed 1: go-bunkboard*... | Wave 5 chunk 5W1 complete — all 6 catalog items authored, self-checked, and validated (`validate_seeds.py --only <ids>`: 6 valid, 0 invalid).
**go-bunkboard** — debug, warnfix-go — hostel occupancy board failing a gofmt+vet+test CI gate: printf verb mismatches ×3, struct tag typos (`jsn:` pinned by tests, malformed `,... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'm the wave-5 chunk 5W3 author. Let me start by reading the spec, catalog, and checking existing seeds.
Now I need exact expected strings for the tests. Let me compute them with a scratch run of the corrected format:
All planted warnings fire as errors. Now applying the reference fix in the workspace:
Reference pas... | Wave-5 chunk 5W3 complete — all six catalog items authored, self-checked, and validated. Final report:
**c-snowroute** — debug / warnfix-c — Snowplow shift-report module under `-Wall -Wextra -Werror`: format mismatches ×4 (%d for long/size_t, %u for size_t), sign-compare loops ×2, unused-parameter ×2 with the void-cas... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the required specs and checking the existing seed catalog.
Now let me look at the go and ts example seed files to match conventions, then check the smoke tooling.
Now the ts seed — fuel log with the vanished operand line so `/` after `round(` lexes as a regex start.
Values with float rounding w... | All six 5S2 seeds are authored, self-checked, and verified. Final report:
**py-dyelot** — debug / syntax-py — Dye-lot batching module: decorated `def` missing its trailing colon + stray `\` line-continuation gluing the next statement (second error reports on the line after the guilty one).
- Smoke: `[ok ] py-dyelot: ... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the required docs and surveying the existing seeds.
Now let me look at an example build seed for formatting.
The catalog items are clear. Let me check how `materialize` and smoke work so I can plan the self-check loop.
Reference behavior looks correct (blocks, wins over blocks, center reply). N... | All six seeds complete and every gate is green. Final report:
**py-tictactoe** — build-game — CLI tic-tac-toe with two-player stdin transcripts and a `--bot` perfect minimax opponent (tie-break = lowest cell index), plus an exhaustive never-loses audit over every opponent line on both sides.
- smoke: `[ok ] py-tictac... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll author the three 6P9 capstone seeds. First, let me read the spec and catalog.
Format is clear. Let me check how seeds with shipped fixture files handle `test_files` protection, and glance at the smoke tooling.
Now I have the full picture. Let me check pwsh availability and the ps-depotops domain (to avoid a modu... | Wave 6 chunk 6P9 complete — all 3 capstone seeds authored, self-checked, and validated.
**ps-opskit** — project-module (catalog 6P9 #1)
Greenfield `OpsKit` module: inventory → drift → plan → apply pipeline behind a psd1 manifest; harness drives child-pwsh driver scripts that pin the exported-command set (exactly 4), T... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the required docs and surveying the existing seed catalog.
Odd — grep isn't finding headings that clearly exist. Let me read the file directly.
I have the catalog. Now let me study existing bash/zsh seeds as formatting examples.
Now the self-check for seed 1: smoke (must fail pre-fix), then mat... | PROGRESS.md shows as modified — I never edited it. Checking whether the smoke tool writes it:
PROGRESS.md was modified by another concurrent pipeline session (a trace-worker log entry timestamped 19:36 CDT today) — not by me; my changes are confined to `tasks/seeds/<my-ids>/` and `workspaces/<my-ids>/`. All six 5E3 sl... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the required spec files and checking the existing seed catalog.
Now let me read the wave-4 additional constraints and the wave-5 catalog for chunk 5C3.
The grep found no headings — let me read the file directly.
No 5C3 seeds exist yet — all 6 are mine to author. Let me study the wave-5 java and... | All six chunk-5C3 seeds authored, self-checked, and gated. Final report:
**java-campboard** — debug / compilefix-java — Campground site board with 7 planted compile errors: `? extends` vs `? super` misuse (both directions), unreported IOException in a `Writer` lambda, unimplemented abstract method hidden behind a rena... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the spec, the wave-6 catalog, and surveying existing seeds.
6P1 and 6P2 are complete; no 6P3 seeds exist. Now let me study an existing build seed and the debug seed as formatting references.
All behaviors confirmed. Let me look at a child-process-style harness (ps-hostswap) and check how smoke w... | That hit is just "for**ge**t" inside "People forget…" — a substring false positive; vocabulary is clean. All gates are done: smoke ok → reference passes → non-empty patches touching no test files → validate `[valid]` → re-smoke ok, plus 5× deterministic verify on the debug seed.
## Final report — wave 6, chunk 6P3 (6/... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the required docs: the seed spec, wave-4 constraints, wave-5 catalog, and the existing seed list.
All the models I need are in place. Let me check the smoke/materialize mechanics in `src/` briefly, then start authoring.
Materialize wipes and re-copies the workspace. Let me check npm availability... | All gates are green and the repo contains only my six new seed directories. Final report:
**Chunk 5W5 — all 6 catalog items authored, self-checked, validated**
| id | kind-category | description |
|---|---|---|
| ts-skatehire | debug / warnfix-ts (index 1) | tsc strict retrofit: strictNullChecks + noUnusedLocals/Para... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the required docs and surveying the existing seeds.
Now let me look at an example build seed for formatting.
The catalog items are clear. Let me check how `materialize` and smoke work so I can plan the self-check loop.
Reference behavior looks correct (blocks, wins over blocks, center reply). N... | All six seeds complete and every gate is green. Final report:
**py-tictactoe** — build-game — CLI tic-tac-toe with two-player stdin transcripts and a `--bot` perfect minimax opponent (tie-break = lowest cell index), plus an exhaustive never-loses audit over every opponent line on both sides.
- smoke: `[ok ] py-tictac... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the spec and catalog documents, then check the existing seed state.
The table is derived. Let me verify it solves random boards end-to-end, and also detects unsolvability correctly:
Now I'll author seed 1: **ruby-lightsout**. First, compute the fixture boards and their pinned solutions with the ... | Everything is green and my footprint is exactly the six new seed directories (the other modified/untracked paths were pre-existing in the starting git status). Final report:
## Wave 7 chunk 7G10 — final report
All 6 missing catalog items authored, each with `"catalog": {"wave": 7, "chunk": "7G10", "index": N}`, full ... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the required specs and checking the existing seed catalog.
Now let me look at the go and ts example seed files to match conventions, then check the smoke tooling.
Now the ts seed — fuel log with the vanished operand line so `/` after `round(` lexes as a regex start.
Values with float rounding w... | All six 5S2 seeds are authored, self-checked, and verified. Final report:
**py-dyelot** — debug / syntax-py — Dye-lot batching module: decorated `def` missing its trailing colon + stray `\` line-continuation gluing the next statement (second error reports on the line after the guilty one).
- Smoke: `[ok ] py-dyelot: ... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the required specs and checking the existing seed catalog.
Specs read. Now let me look at recently-authored wave-5 seeds for format conventions (a go one and a rust one), and check toolchains.
Now let me see which wave-5 seeds already landed and their category conventions:
Now let me check the ... | All six seeds are complete and every gate passed. Final report:
**go-sawplan** — debug / compilefix-go — Sawmill cut-list & board-foot costing package with 6 independent compile errors across 3 files (unused import, `:=` shadowing the `strings` package, wrong receiver type on `JobCost` used at two sites, missing retur... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'm authoring wave-5 chunk 5E2 (6 go/rust escape seeds). Starting by reading the spec, the wave-5 catalog laws, and checking existing seeds.
Wave 5 has 30 seeds landed already (chunks 5S1–5E1 presumably), none in 5E2. Let me check a wave-5 seed's task.json format for the catalog provenance shape.
Authoring seed 1 — `... | Wave 5, chunk 5E2 — all 6 seeds authored, self-checked, and gated.
**go-reviewflag** — debug / escape-regex. QA review-term tagger where `"\bterm\b"` in an interpreted string compiles to backspace characters (zero matches) and a quoted-form struct tag missing its closing escaped quote drops `total_matches` from the JS... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the spec, wave-4 constraints, wave-7 catalog chunk, and checking existing seeds.
Hand-computed pins match the reference exactly. Now the real acceptance test suite for rust-nim.
rust-nim reference passes all 17 tests. Now writing the seed files and prompt.
Optimized profile: solve 0.6s, uniquen... | All six seeds for chunk 7G4 are complete and fully self-checked. Final report:
## Wave 7 / Chunk 7G4 — final report
| id | kind-category | description |
|---|---|---|
| **rust-freecell** | build-game | FreeCell engine: text deals, named free cells, foundation/cascade legality, supermove capacity law `(free+1)*2^(empt... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the required docs and checking the existing seed catalog.
All context gathered. Now I'll design and author the six seeds. Quick overlap check on the closest existing seeds first:
**Seed 2: `c-kvunesc`** (C build — escape decoder with strict error positions):
Key catch: `2026-13-01` raises a pla... | All six chunk-5E5 seeds are authored, self-checked, and validator-clean. Final report:
**c-panelframe** — debug / escape-string — Serial operator-panel frame builder where `"\128"` (octal escape overrun: `\12`+`8`) corrupts the 0x80 start-of-frame byte and a missing comma merges `"FAULT" "SERVICE"` in the tag array; t... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the required docs and checking the existing seed catalog.
Probing the exact strings I'll pin in the strkit harness (format-operator output, title-case edges, more case pairs):
All probes confirmed. Now writing the six seeds, starting with `ps-strkit` (item 1, string toolkit).
Now the task.json ... | Wave 6 chunk 6P2 is complete: all six catalog items authored, self-checked, and validated. Final report:
**ps-strkit** — build-strings — trim/pad/truncate-with-ellipsis/title-case toolkit via .NET string methods + `-f` only; regex operators banned by a source scan; invariant-culture non-ASCII edge table pinned from pr... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | You are Fable, the sole seed author responsible for repairing wave 8 chunk 8B.
Do not delegate, do not call another model, and do not generate code-teacher traces.
The four existing seeds were correctly implemented and locally verified, but the
fail-closed wave audit rejected their user prompts because each prompt lac... | I'll start by locating the four seed directories and reading the audit/validation gates so the prompt repairs match exactly what the fail-closed audit expects.
All four seed directories exist. Now let me read the audit gate and the four task.json files.
The pattern is clear: go-k8s-job-watch says "Before you change a... | All four wave 8 chunk 8B seeds are repaired and green through both gates.
**What was wrong:** each prompt used a grammatical variant of the required research instruction — "Before you change any code" (go), "Before touching code" (py), "before writing any code" (ts), "Before any code" (java) — so the fail-closed wave ... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the required spec files and checking the existing seed catalog.
Now let me read the wave-4 constraints and the wave-5 catalog chunk.
The file contains bytes that make grep treat it as binary. Let me read it with `grep -a`.
The plan is set. Let me check `materialize` to understand workspace mech... | All gates pass. Final report:
## Wave 5, chunk 5S3 — 6/6 seeds authored
| # | id | kind-category | description |
|---|----|----|----|
| 1 | `java-lotsheet` | debug / syntax-java | Anonymous-class `Comparator` brace interleave off by one in an auction lot sheet — parse error lands 55 lines later at the class-closing b... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the required specs and checking the existing seed catalog.
Specs read. Now let me look at recently-authored wave-5 seeds for format conventions (a go one and a rust one), and check toolchains.
Now let me see which wave-5 seeds already landed and their category conventions:
Now let me check the ... | All six seeds are complete and every gate passed. Final report:
**go-sawplan** — debug / compilefix-go — Sawmill cut-list & board-foot costing package with 6 independent compile errors across 3 files (unused import, `:=` shadowing the `strings` package, wrong receiver type on `JobCost` used at two sites, missing retur... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll author the three 6P9 capstone seeds. First, let me read the spec and catalog.
Format is clear. Let me check how seeds with shipped fixture files handle `test_files` protection, and glance at the smoke tooling.
Now I have the full picture. Let me check pwsh availability and the ps-depotops domain (to avoid a modu... | Wave 6 chunk 6P9 complete — all 3 capstone seeds authored, self-checked, and validated.
**ps-opskit** — project-module (catalog 6P9 #1)
Greenfield `OpsKit` module: inventory → drift → plan → apply pipeline behind a psd1 manifest; harness drives child-pwsh driver scripts that pin the exported-command set (exactly 4), T... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'm authoring wave 7 chunk 7G9 (6 C# seeds). First, reading the spec and catalog, and surveying existing seeds.
dotnet 10 is present. Let me check the C# build-seed conventions (Part D) and a wave-7 build seed example.
Catching a bug in my own test: scoreboard totals are cumulative, so frame 2's line must read 30, no... | Wave 7 chunk 7G9 complete — all 6 seeds authored, audited, smoked, and round-trip verified.
**cs-uno** — build-game — UNO-lite referee engine: fixture decks, pinned legality law, skip/reverse/draw-two with 2-vs-3+ direction handling, forced-draw-then-play with scripted wild colors, full 3-player transcript diffed whol... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the required docs and surveying existing seeds.
Implicit-decl and int→pointer errors confirmed. The K&R promotion mismatch didn't fire — let me probe it separately.
K&R with matching promotion passes silently. Let me probe the mismatch direction that actually errors.
K&R type-drift mismatch (p6... | All six chunk-5C4 seeds are authored, self-checked, and validated.
**c-tidechart** — debug / compilefix-c — Harbor tide-chart tool that won't build: missing `<ctype.h>`/`<stdlib.h>`/`<string.h>` includes (implicit declarations ×3), `tide_slot_label` never declared in the header (int-returning-pointer truncation in cha... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the spec, the wave-6 catalog, and surveying existing seeds.
6P1 and 6P2 are complete; no 6P3 seeds exist. Now let me study an existing build seed and the debug seed as formatting references.
All behaviors confirmed. Let me look at a child-process-style harness (ps-hostswap) and check how smoke w... | That hit is just "for**ge**t" inside "People forget…" — a substring false positive; vocabulary is clean. All gates are done: smoke ok → reference passes → non-empty patches touching no test files → validate `[valid]` → re-smoke ok, plus 5× deterministic verify on the debug seed.
## Final report — wave 6, chunk 6P3 (6/... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the required docs and surveying the existing seed catalog.
All tttbit pinned values are computed. Now I'll author the c-tttbit seed files.
I caught an error in my test fixtures — the O-win fixture had cell 8 in both masks. Fixing:
Let me verify the turn_bookkeeping frame with the oracle — I sus... | All gates pass. One observation: other parallel authors/orchestrator processes are active in the repo (py-* 7G1 seeds, cpp-refcard trace files, and three task.json modifications appeared mid-session) — none of it touched by me; I stayed inside my six seed dirs and their workspaces.
Final report — wave 7, chunk 7G7 (6/... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the required docs: the seed spec, wave-4 constraints, wave-5 catalog, and the existing seed list.
All the models I need are in place. Let me check the smoke/materialize mechanics in `src/` briefly, then start authoring.
Materialize wipes and re-copies the workspace. Let me check npm availability... | All gates are green and the repo contains only my six new seed directories. Final report:
**Chunk 5W5 — all 6 catalog items authored, self-checked, validated**
| id | kind-category | description |
|---|---|---|
| ts-skatehire | debug / warnfix-ts (index 1) | tsc strict retrofit: strictNullChecks + noUnusedLocals/Para... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the spec, wave-4 constraints, wave-7 catalog chunk, and checking existing seeds.
Hand-computed pins match the reference exactly. Now the real acceptance test suite for rust-nim.
rust-nim reference passes all 17 tests. Now writing the seed files and prompt.
Optimized profile: solve 0.6s, uniquen... | All six seeds for chunk 7G4 are complete and fully self-checked. Final report:
## Wave 7 / Chunk 7G4 — final report
| id | kind-category | description |
|---|---|---|
| **rust-freecell** | build-game | FreeCell engine: text deals, named free cells, foundation/cascade legality, supermove capacity law `(free+1)*2^(empt... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the required docs and checking the existing seed catalog.
All context gathered. Now I'll design and author the six seeds. Quick overlap check on the closest existing seeds first:
**Seed 2: `c-kvunesc`** (C build — escape decoder with strict error positions):
Key catch: `2026-13-01` raises a pla... | All six chunk-5E5 seeds are authored, self-checked, and validator-clean. Final report:
**c-panelframe** — debug / escape-string — Serial operator-panel frame builder where `"\128"` (octal escape overrun: `\12`+`8`) corrupts the 0x80 start-of-frame byte and a missing comma merges `"FAULT" "SERVICE"` in the tag array; t... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'm the wave-5 chunk 5W3 author. Let me start by reading the spec, catalog, and checking existing seeds.
Now I need exact expected strings for the tests. Let me compute them with a scratch run of the corrected format:
All planted warnings fire as errors. Now applying the reference fix in the workspace:
Reference pas... | Wave-5 chunk 5W3 complete — all six catalog items authored, self-checked, and validated. Final report:
**c-snowroute** — debug / warnfix-c — Snowplow shift-report module under `-Wall -Wextra -Werror`: format mismatches ×4 (%d for long/size_t, %u for size_t), sign-compare loops ×2, unused-parameter ×2 with the void-cas... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the spec, wave catalogs, and checking existing seeds.
No 6P5 seeds exist yet — I'll author all six. Let me study an existing PS debug seed as the formatting example.
Default culture is Invariant (empty name) — deterministic box. Now the behavior probes:
All probes confirm the planned bugs. Now ... | All six 6P5 seeds are authored, self-checked, and green on every gate (smoke ×3 runs, reference verify ×5 runs each, patch round-trip on fresh workspaces, patches touch only the script under test).
## Final report — wave 6, chunk 6P5 (6 debug seeds, all symptom-first)
**ps-netroster** — debug-culture — device roster ... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'm authoring wave-5 chunk 5S4 (6 seeds: bash×2, zsh×1, C×2, C++×1). Starting by reading the spec and catalog.
Good pattern. Now let me probe toolchain versions and verify my planned zsh/bash error mechanics empirically with inline commands.
Item 2 mechanics confirmed. Redoing the unclosed-quote probe cleanly via std... | All six seeds are complete, self-checked, and in a pristine post-smoke state. Final report:
**Chunk 5S4 — six seeds authored** (all `debug` kind; every `task.json` carries `"catalog": {"wave": 5, "chunk": "5S4", "index": N}`)
1. **bash-poolsheet** — syntax-bash — weekly aquatic-centre sheet renderer; one quote missin... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the required docs and surveying existing seeds.
Implicit-decl and int→pointer errors confirmed. The K&R promotion mismatch didn't fire — let me probe it separately.
K&R with matching promotion passes silently. Let me probe the mismatch direction that actually errors.
K&R type-drift mismatch (p6... | All six chunk-5C4 seeds are authored, self-checked, and validated.
**c-tidechart** — debug / compilefix-c — Harbor tide-chart tool that won't build: missing `<ctype.h>`/`<stdlib.h>`/`<string.h>` includes (implicit declarations ×3), `tide_slot_label` never declared in the header (int-returning-pointer truncation in cha... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the required specs and checking the existing seed catalog.
The chunk 5C2 grep didn't hit — let me check WAVE5_USECASES.md directly.
Odd — the heading is visible in `head` but grep missed it. Possibly unusual characters. Let me check.
Now the task.json (placeholder prompt until I capture the rea... | Chunk 5C2 complete — all 6 catalog items authored, self-checked, and validated (`validate_seeds.py`: 6 valid, 0 invalid; `audit_seeds.py`: 568 complete, 0 partial; nothing outside my seed dirs touched).
## Final report — wave 5, chunk 5C2
**ts-bikedock** (index 1) — debug / compilefix-ts — Bike-share dock rebalancing... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the required spec documents and surveying existing seeds.
Now let me read the wave-6 catalog and wave-4 constraints, plus an existing PowerShell seed for formatting.
The mandatory-miss probe hung (pwsh prompts without `-NonInteractive`) — confirming why harnesses must pass `-NonInteractive` to c... | All 6 missing chunk-6P4 items authored, self-checked, and gated.
**ps-oddwalk** — build-fs — literal-path walker: recursive cleanup inventory over a root whose own path carries spaces + `[brackets]` (probe-proven: `-Path` binding silently matches nothing), dotfiles via `-Force`, ordinal relpaths, action rules by exten... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the required docs and surveying the existing seeds.
Now let me read the wave-5 catalog laws and the wave-4 authoring constraints.
Now let me check the toolchain prerequisites (shellcheck, rust, ruby) and look at existing wave-5 seeds for the provenance shape.
5W1 is done; 5W2 has zero seeds — a... | All gates pass. Final report:
**Chunk 5W2 — 6/6 seeds authored, all self-checks green**
| # | id | kind-category | description |
|---|----|----|----|
| 1 | `rs-cidermill` | debug / warnfix-rust | `-D warnings` sweep on a cider-mill press-day ledger: 3× unused `#[must_use]` Result (2 behavior-restoring, 1 infallible-w... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the required specs and checking the current state of the seed catalog.
Variant B silently swallows to the next terminator — good material. Now probe the missing-terminator EOF report and hashtable brace:
Variant C gives a displaced "missing terminator" report. Now let me probe the item-2 defects... | All six 6P10 seeds are authored, validated, and gated. Final report:
## Wave 6, chunk 6P10 — final report
**ps-kilnsheet** — debug / `syntax-powershell` (item 1). Two-file broken-parse triage: here-string terminator indented inside a function swallows the closing brace and half the program until a later column-1 `'@`... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the required spec files and checking the existing seed catalog.
Now let me read the wave-4 additional constraints and the wave-5 catalog for chunk 5C3.
The grep found no headings — let me read the file directly.
No 5C3 seeds exist yet — all 6 are mine to author. Let me study the wave-5 java and... | All six chunk-5C3 seeds authored, self-checked, and gated. Final report:
**java-campboard** — debug / compilefix-java — Campground site board with 7 planted compile errors: `? extends` vs `? super` misuse (both directions), unreported IOException in a `Writer` lambda, unimplemented abstract method hidden behind a rena... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'm authoring wave-5 chunk 5E2 (6 go/rust escape seeds). Starting by reading the spec, the wave-5 catalog laws, and checking existing seeds.
Wave 5 has 30 seeds landed already (chunks 5S1–5E1 presumably), none in 5E2. Let me check a wave-5 seed's task.json format for the catalog provenance shape.
Authoring seed 1 — `... | Wave 5, chunk 5E2 — all 6 seeds authored, self-checked, and gated.
**go-reviewflag** — debug / escape-regex. QA review-term tagger where `"\bterm\b"` in an interpreted string compiles to backspace characters (zero matches) and a quoted-form struct tag missing its closing escaped quote drops `total_matches` from the JS... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'm authoring wave 7 chunk 7G5 — six shell-game seeds. Let me start by reading the spec and catalog.
Catalog understood. Now let me look at recent wave-7 seeds for `task.json` shape and a bash/zsh seed for harness style.
Let me look at a couple of harness examples — a bash build seed's test file and one wave-7 game s... | All six catalog items for chunk 7G5 are authored, self-checked, and gated. The `traces/ts-reglob.*` files above are orchestrator artifacts, not mine; no tracked files were modified.
**Final report — wave 7, chunk 7G5 (one seed per catalog item, provenance `{"wave": 7, "chunk": "7G5", "index": N}` in each task.json)**
... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'm authoring the 6 missing TypeScript seeds for wave 7 chunk 7G3. Let me start by reading the spec and catalog.
The six ids (ts-minesweeper, ts-2048, ts-connect4, ts-memory, ts-hangman, ts-pong) are all new. Now let me read the wave-7 catalog laws and an example TS build seed.
Now let me check the pipeline's materia... | All six seeds pass every gate. Final report:
**Wave 7, chunk 7G3 — 6 seeds authored**
| id | kind-category | description | smoke | reference verify |
|---|---|---|---|---|
| ts-minesweeper | build-game | Pure-function minesweeper engine: chording on satisfied numbers, flag/question-mark cycle, flood-fill law (flags b... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the required specs and checking the existing seed catalog.
Let me look at a wave-4 integration seed (python mock server pattern) and a ps build seed (harness pattern).
HttpListener works. Now probe the python-server-under-pwsh-harness pattern, `-SkipHttpErrorCheck`, header access, and byte PUTs.... | All six chunk-6P6 seeds are authored, self-checked, and gated. Final report:
**ps-pagehaul** — build-http — cursor-paginated inventory collector: follows opaque `next` cursors to exhaustion incl. the empty terminal page, pinned per-page-size request log, ordinal (case-preserving) site aggregation, empty-dataset case.
... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | You are Fable, the sole seed author responsible for repairing wave 8 chunk 8C.
Do not delegate, do not call another model, and do not generate code-teacher traces.
The four existing seeds were authored and their reference gates passed, but the
fail-closed wave audit rejected at least one user prompt because it lacks t... | I'll start by looking at the four 8C seed task files and the audit/validation gates so the repairs match exactly what the fail-closed audit checks.
Now let me read the four task.json files and the two gate scripts.
Two more prompts confirmed missing the literal ("Research before you write anything" and "Research firs... | All four wave 8 chunk 8C seeds are repaired and both gates pass: `src/audit_seeds.py` reports all four `[complete]`, and `src/validate_seeds.py --only` reports all four `[valid]` — baselines still fail before the patch, reference patches apply cleanly and pass verify, and protected test files stay byte-identical.
The ... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the required docs and checking the existing seed catalog.
Conventions are clear. Let me check the categories used by other wave-5 escape seeds and confirm the smoke workflow.
All conventions gathered. Now I'll plan the 6 seeds and check the smoke/materialize tooling interface before writing file... | All six chunk-5E4 seeds are authored, self-checked, and gated. Final report:
**java-blockreport** — debug, `escape-textblock` — Library batch renders overdue-export SQL and 32-column printer slips from text blocks; migration kept `\n` literals (doubled lines) and lost trailing-space padding (needs `\s`); output pinned... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'm authoring wave 7 chunk 7G5 — six shell-game seeds. Let me start by reading the spec and catalog.
Catalog understood. Now let me look at recent wave-7 seeds for `task.json` shape and a bash/zsh seed for harness style.
Let me look at a couple of harness examples — a bash build seed's test file and one wave-7 game s... | All six catalog items for chunk 7G5 are authored, self-checked, and gated. The `traces/ts-reglob.*` files above are orchestrator artifacts, not mine; no tracked files were modified.
**Final report — wave 7, chunk 7G5 (one seed per catalog item, provenance `{"wave": 7, "chunk": "7G5", "index": N}` in each task.json)**
... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'm authoring the 6 missing TypeScript seeds for wave 7 chunk 7G3. Let me start by reading the spec and catalog.
The six ids (ts-minesweeper, ts-2048, ts-connect4, ts-memory, ts-hangman, ts-pong) are all new. Now let me read the wave-7 catalog laws and an example TS build seed.
Now let me check the pipeline's materia... | All six seeds pass every gate. Final report:
**Wave 7, chunk 7G3 — 6 seeds authored**
| id | kind-category | description | smoke | reference verify |
|---|---|---|---|---|
| ts-minesweeper | build-game | Pure-function minesweeper engine: chording on satisfied numbers, flag/question-mark cycle, flood-fill law (flags b... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the specs and checking the existing catalog state.
No 6P7 seeds exist yet — all 6 items need authoring. Let me read the wave laws and the chunk section.
Let me examine an existing refactor seed to understand the mechanics.
Pipeline understood. Now let me probe the pwsh behaviors these seeds hin... | Everything checks out. All six chunk-6P7 seeds are authored, self-checked, and gated.
## Final report — wave 6, chunk 6P7 (6/6 authored)
| id | kind-category | description |
|---|---|---|
| ps-depotops | refactor-module | 268-line spare-parts depot ops monolith (report/shortfall/audit CLI) split into `DepotOps.psm1` ... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the required docs and surveying the existing seeds.
Now let me read the wave-5 catalog laws and the wave-4 authoring constraints.
Now let me check the toolchain prerequisites (shellcheck, rust, ruby) and look at existing wave-5 seeds for the provenance shape.
5W1 is done; 5W2 has zero seeds — a... | All gates pass. Final report:
**Chunk 5W2 — 6/6 seeds authored, all self-checks green**
| # | id | kind-category | description |
|---|----|----|----|
| 1 | `rs-cidermill` | debug / warnfix-rust | `-D warnings` sweep on a cider-mill press-day ledger: 3× unused `#[must_use]` Result (2 behavior-restoring, 1 infallible-w... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'm authoring the 6 missing catalog items for wave 5, chunk 5S1. Starting by reading the spec and catalog.
No wave-5 seeds exist yet — all 6 items need authoring. Checking lang-field conventions and example debug seeds for go/ts.
One-line displacement with a helpful message — not good enough. Probing variants for dee... | All six seeds authored, self-checked, and validated. Final report:
**Seed 1 — py-menucost** (debug, syntax-py) — catering order-desk pricing; unclosed paren in a dict-comprehension chain, fix is one char (`)` after `MARGIN`).
- Smoke: `[ok ] py-menucost: pre-fix verify fails as expected (SyntaxError: closing parenthe... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the required docs and surveying the existing seed catalog.
Odd — grep isn't finding headings that clearly exist. Let me read the file directly.
I have the catalog. Now let me study existing bash/zsh seeds as formatting examples.
Now the self-check for seed 1: smoke (must fail pre-fix), then mat... | PROGRESS.md shows as modified — I never edited it. Checking whether the smoke tool writes it:
PROGRESS.md was modified by another concurrent pipeline session (a trace-worker log entry timestamped 19:36 CDT today) — not by me; my changes are confined to `tasks/seeds/<my-ids>/` and `workspaces/<my-ids>/`. All six 5E3 sl... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | router is exact-match only and the gateway team is stuck. Need pattern routing:
- ":name" segments capture one path segment ("/repos/:owner/:repo")
- "*name" as the FINAL segment captures the whole non-empty remainder including slashes; "*" anywhere else, or unnamed "*"/":", is a registration error via a sentinel ErrI... | I'll look at the current router implementation and tests first. | Green — `go test -race -timeout 30s ./...` passes, with both `params_test.go` (the new acceptance contract) and `router_test.go` (existing behavior) intact. All changes are in router.go:
- **`Match(path) (H, Params, bool)`** — `router.go:129`. Normalizes the path the same way `Lookup` does, then scans the parsed route... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the spec, wave catalogs, and checking existing seeds.
No 6P5 seeds exist yet — I'll author all six. Let me study an existing PS debug seed as the formatting example.
Default culture is Invariant (empty name) — deterministic box. Now the behavior probes:
All probes confirm the planned bugs. Now ... | All six 6P5 seeds are authored, self-checked, and green on every gate (smoke ×3 runs, reference verify ×5 runs each, patch round-trip on fresh workspaces, patches touch only the script under test).
## Final report — wave 6, chunk 6P5 (6 debug seeds, all symptom-first)
**ps-netroster** — debug-culture — device roster ... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the required specs and checking the existing seed catalog.
Let me look at a wave-4 integration seed (python mock server pattern) and a ps build seed (harness pattern).
HttpListener works. Now probe the python-server-under-pwsh-harness pattern, `-SkipHttpErrorCheck`, header access, and byte PUTs.... | All six chunk-6P6 seeds are authored, self-checked, and gated. Final report:
**ps-pagehaul** — build-http — cursor-paginated inventory collector: follows opaque `next` cursors to exhaustion incl. the empty terminal page, pinned per-page-size request log, ordinal (case-preserving) site aggregation, empty-dataset case.
... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'm authoring wave 7 chunk 7G9 (6 C# seeds). First, reading the spec and catalog, and surveying existing seeds.
dotnet 10 is present. Let me check the C# build-seed conventions (Part D) and a wave-7 build seed example.
Catching a bug in my own test: scoreboard totals are cumulative, so frame 2's line must read 30, no... | Wave 7 chunk 7G9 complete — all 6 seeds authored, audited, smoked, and round-trip verified.
**cs-uno** — build-game — UNO-lite referee engine: fixture decks, pinned legality law, skip/reverse/draw-two with 2-vs-3+ direction handling, forced-draw-then-play with scripted wild colors, full 3-player transcript diffed whol... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the required docs and checking the existing seed catalog.
Conventions are clear. Let me check the categories used by other wave-5 escape seeds and confirm the smoke workflow.
All conventions gathered. Now I'll plan the 6 seeds and check the smoke/materialize tooling interface before writing file... | All six chunk-5E4 seeds are authored, self-checked, and gated. Final report:
**java-blockreport** — debug, `escape-textblock` — Library batch renders overdue-export SQL and 32-column printer slips from text blocks; migration kept `\n` literals (doubled lines) and lost trailing-space padding (needs `\s`); output pinned... |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the spec and catalog documents, then check the existing seed state.
The table is derived. Let me verify it solves random boards end-to-end, and also detects unsolvability correctly:
Now I'll author seed 1: **ruby-lightsout**. First, compute the fixture boards and their pinned solutions with the ... | All six validate. Final lint of task.json provenance and a scope check: |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the required docs and surveying existing seeds.
Implicit-decl and int→pointer errors confirmed. The K&R promotion mismatch didn't fire — let me probe it separately.
K&R with matching promotion passes silently. Let me probe the mismatch direction that actually errors.
K&R type-drift mismatch (p6... | Baseline shows the circular break from both directions plus the redefinitions. Fix the headers first to confirm the link-stage reveal. |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the required docs and surveying existing seeds.
Implicit-decl and int→pointer errors confirmed. The K&R promotion mismatch didn't fire — let me probe it separately.
K&R with matching promotion passes silently. Let me probe the mismatch direction that actually errors.
K&R type-drift mismatch (p6... | All green (8 checks). Now the task.json. |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the spec and catalog documents, then check the existing seed state.
The table is derived. Let me verify it solves random boards end-to-end, and also detects unsolvability correctly:
Now I'll author seed 1: **ruby-lightsout**. First, compute the fixture boards and their pinned solutions with the ... | All 33 pass. Save patch, re-smoke, then run the full final gate over all six seeds: |
You are an autonomous coding agent running on the user's machine.
You work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are known until you load their schemas wit... | # Wave-4 authoring-agent preamble (read this first, follow exactly)
You are authoring seeds for the fable-code training-data pipeline. Repo root:
this directory. Everything below is mandatory.
## Read before writing anything
1. `tasks/SEED_SPEC.md` — the seed spec, quality bar, and mandatory
self-check. It is the... | I'll start by reading the spec and catalog documents, then check the existing seed state.
The table is derived. Let me verify it solves random boards end-to-end, and also detects unsolvability correctly:
Now I'll author seed 1: **ruby-lightsout**. First, compute the fixture boards and their pinned solutions with the ... | `validate_seeds.py --only` does the full round-trip proof. Running it on all six: |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.