README.md

---
pretty_name: jun_27_1000_g5_200 pure Harbor tasks and rubrics
---

# jun_27_1000_g5_200

Lossless, restorable archives for the `jun_27_1000_g5_200` batch.

## Contents

- `jun_27_1000_g5_200__pure_harbor_tasks.tar.zst`: 183 rollout-ready Harbor tasks and their complete raw skill inputs. Pipeline intermediates, rollout/judge outputs, and `.setup_synthesis` directories are excluded. `dhh-coder` is excluded because its setup requires human classification.
- `jun_27_1000_g5_200__rubrics.tar.zst`: 330 validated final rubric JSON files (166 completion and 164 process rubrics), stored separately from the pure task archive.
- `selection_manifest.json`: exact selection rules, task IDs, exclusions, and rubric coverage.
- `archive_manifest.json`: archive members, restored roots, sizes, checksums, and the original split-transport provenance.
- `verification_report.json`: local extraction and lossless-restoration verification results.
- `pure_tasks_source_snapshot.jsonl.zst` and `rubrics_source_snapshot.jsonl.zst`: compressed source snapshots used for provenance verification.
- `SHA256SUMS`: SHA-256 checksums for all uploaded data and metadata artifacts.

The main archive restores these roots:

```text
raw_skills/jun_27_1000_g5_200/
synthesized_data/jun_27_1000_g5_200/
pipeline_runs/jun_27_1000_g5_200/
```

The `pipeline_runs` location in the main archive is intentionally empty. The separate rubric archive restores rubric files beneath that location.

## Verify and restore

From this repository directory:

```bash
sha256sum -c SHA256SUMS
mkdir -p /path/to/restore
zstd -dc jun_27_1000_g5_200__pure_harbor_tasks.tar.zst | tar -xf - -C /path/to/restore
zstd -dc jun_27_1000_g5_200__rubrics.tar.zst | tar -xf - -C /path/to/restore
```

The archives were verified by full extraction, path-set comparison, file size and SHA-256 comparison, metadata comparison, and a rollout precheck for all 183 selected tasks. See `verification_report.json` for the recorded checks.