Skip to content

moonlight-lupin/agent-skills / plugins/dev-pipeline/skills/review-commit

review-commit

Verified

Use when the user says 'review loop N times', 'build then review', 'fix and re-review', 'review loop until clean', 'have cursor build, codex review', or wants the dev-pipeline review stage: a two-axis single pass (Standards + Spec, parallel, never merged) FIRST, then a budgeted confirm-round loop on the fix diffs until no BLOCKING/MAJOR is open, then the commit/deploy gate — three-reviewer gate before live deploy, release hygiene before tagging. Also triggers: 'two-axis review', 'confirm rounds', 'review diff until SHIP', 'N confirm rounds', ship gate.

FileKindSize
SKILL.mdexecutableskill md6.8 KB
references/background-build-polling.mdexecutablescript3.9 KB
references/confirm-round-protocol.mdexecutablescript12.3 KB
references/dispatch-result-contract.mdexecutablescript3.5 KB
references/feature-build-brief.mdexecutablescript2.9 KB
references/fix-brief-template.mdexecutablescript10.5 KB
references/persistent-loop-plan.mdexecutablescript6.9 KB
references/release-hygiene.mdexecutablescript3.9 KB
references/reviewer-dispatch.mdexecutablescript6.0 KB
references/spec-convergence.mdexecutablescript5.2 KB
references/two-axis-single-pass.mdexecutablescript9.9 KB
references/workflow-steps.mdexecutablescript18.9 KB
scripts/attest_file.pyexecutablescript2.7 KB
scripts/capture_dispatch_result.pyexecutablescript3.6 KB
scripts/loop_plan_check.pyexecutablescript3.8 KB

references/dispatch-result-contract.md

View on GitHub
# Dispatch Result Contract (`delegate-result.v1`)

Ported concept (Sep 2026) from amElnagdy/delegate-skills (MIT): every
implementer dispatch produces one machine-readable result file. Our loop does
not run their relay (it wraps other orchestrators' CLIs); it keeps the same
contract shape. Hermes stays the orchestrator: same model, same CLIs, same
review gates.

## The contract

One JSON file per dispatch, written with
`scripts/capture_dispatch_result.py`, kept OUTSIDE the repo (e.g. the
workspace or `/tmp`), never committed:

```json
{
  "schema": "delegate-result.v1",
  "task": "build R1: batch A",
  "implementer": "cursor",
  "model": "<builder model that ran>",
  "status": "completed",
  "exitCode": 0,
  "reportPath": "/tmp/proj-build-r1-report.md",
  "stderrPath": null,
  "touchedFiles": [" M core_toolset/modules/x/engine.py", "?? tests/test_x.py"],
  "startedAt": null,
  "finishedAt": "2026-09-12T04:11:32+00:00",
  "notes": "delta round for R1"
}
```

- `status` derives from the exit code: 0 → `completed`, 124 → `timeout`,
  127 → `unavailable` (binary missing), anything else → `failed`.
- `touchedFiles` is `git status --porcelain` at capture time — the review
  boundary. `null` means git could not report (non-repo, git error);
  `[]` means the tree is clean. Untracked files are implementer output:
  no `git diff` shows them — open them directly.
- `model` records the model that ACTUALLY ran, after any tier fallback
  (the builder lane's fallback from `orchestrator/references/lanes.md` when the high tier hit resource_exhausted) —
  same discipline as delegate-skills' `codexVersion`.
- `notes` records dispatch deltas: fallback reasons, round lineage.

## Capture commands per implementer

```bash
# After the implementer process exits and BEFORE verification/commit:
python3 <skill-dir>/scripts/capture_dispatch_result.py \
  --repo <project-dir> --task "build R1" --implementer cursor \
  --model "$MODEL_THAT_RAN" --exit-code <N> \
  --report /tmp/proj-build-r1-report.md --out /tmp/proj-dispatch-r1.json
```

For Claude Code print mode (`claude -p ... > review.txt 2>&1`), pass
`--report review.txt` and `--implementer claude-code`. For Hermes-built
fallbacks, use `--implementer hermes` so the result file shows the loop's
cross-model property was degraded.

## What the result file changes in the loop

- Review starts from `touchedFiles`, not from reading the whole diff.
- Review prompts get the touched-file list pre-computed.
- The commit boundary check is mechanical: the commit content must match
  the captured `touchedFiles` — any file present at commit time but absent
  from the capture (or vice versa) is a scope deviation to adjudicate.
- Failed dispatches (`failed`/`timeout`) leave a record the fix round
  resumes from; the fix brief cites the capture, not memory.

## Attestation scope at dispatch time

Attest (`scripts/attest_file.py attest`) every file the builder is
forbidden to modify — the contract-test file, fixture files it must not
generate differently, and BRIEF.md itself — BEFORE dispatching, and add
`*.attest` to `.gitignore` first so the opaque hash sidecars never ride a
`git add -A` into a commit. After the build exits, re-run the attestation
check; a TAMPERED contract file is a BLOCKING finding. Also verify the
red baseline before dispatch: run the orchestrator-authored contract suite
once, confirm every failure is a module-not-found naming the thing the
build introduces (not a harness bug), and commit that baseline — the red
baseline is what makes "it passes after the build" mean anything.
writing-skillsobra/superpowersUse when creating new skills, editing existing skills, or verifying skills work before deploymentStars: 322kUpdated Claude CodeCodexVerifiedInstall
xlsxanthropics/skillsUse this skill any time a spreadsheet file is the primary input or output. This means any task where the user wants to: open, read, edit, or fix an existing .xlsx, .xlsm, .xltx, .csv, or .tsv file (e.g., adding columns, computing formulas, formatting, charting, cleaning messy data); create a new spreadsheet from scratch or from other data sources; or convert between tabular file formats. Trigger especially when the user references a spreadsheet file by name or path — even casually (like "the xlsx in my downloads") — and wants something done to it or produced from it. Also trigger for cleaning or restructuring messy tabular data files (malformed rows, misplaced headers, junk data) into proper spreadsheets. The deliverable must be a spreadsheet file. Do NOT trigger when the primary deliverable is a Word document, HTML report, standalone Python script, database pipeline, or Google Sheets API integration, even if tabular data is involved.Stars: 198kUpdated Claude CodeCodexVerifiedInstall
shipping-and-launchaddyosmani/agent-skillsPrepares production launches. Use when preparing to deploy to production, or when asking what needs to be in place before shipping. Use when you need a pre-launch checklist, when setting up monitoring, when planning a staged rollout, or when you need a rollback strategy.Stars: 119kUpdated Claude CodeCodexVerifiedInstall
childhoodthedotmack/claude-memBabysit a shipped release after merge, publication, or deployment. Verify published packages, production behavior, installed plugins, and late CI or user reports until the observation window is clean; use babysit for PR readiness before release.Stars: 118kUpdated Claude CodeCodexVerifiedInstall