agentleFS
grovegrove/promise-to-proof/skills/productivity/publish-pr-review/SKILL.md

publish-pr-review

Skill4 starsChanged 36 days agoApache-2.0View on GitHub

Publish the existing exact saved implementation review on its matching PR, with a read-only preview, explicit authority, and GitHub readback. Never perform another code review.

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 1 is distinctive, and 5 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 5

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Read the acceptance protocol and /publish-pr's required review/proof, immutable comparison-base, publication commit and candidate-to-commit …

    line 16
  2. Very specific:

    Do not publish a BLOCKED review.

    line 46
  3. Very specific:

    Use the existing controller's authorize-effect against this exact preview SHA-256, with --action pr-review-comment (or …

    line 50
  4. Very specific:

    After authorizing, run with --publish-preview <retained-ignored-preview.json> --authorization <saved-authorization.json> or an exact …

    line 57
  5. Distinctive:

    After any uncertain write, it reads all reviews again.

    line 62
  6. Very specific:

    If GitHub is unavailable or identities differ, preserve the preview and report PARTIAL or BLOCKED.

    line 71

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Publish a saved PR review

This is an optional expert entry point and the internal publication handoff used by /publish-pr after the requested, authorized draft PR is confirmed. It reuses review.md and the controller's original stage receipt. It does not run review-implementation, change a verdict, prove, approve or merge anything.

Preparation (no effects)

  1. Read the acceptance protocol and /publish-pr's required review/proof, immutable comparison-base, publication commit and candidate-to-commit equivalence rules.
  2. Work from the repository that owns the canonical contract. Require a saved, current review with exact scope and contract/source binding, the correct published PR marker and same-repository current PR head. If a changed candidate or head has uncovered product files, or a mismatch cannot be resolved, block rather than pretending that a prior review is current. Confirm the publication commit mapping is retained if the original review covered a snapshot rather than this Git commit.
  3. Prepare one exact review using the deterministic renderer:

    python3 <skill-dir>/scripts/publish_pr_review.py \
      --repo <root> --contract .p2p/work/<slug>/contract.md \
      --pr https://github.com/OWNER/REPO/pull/NUMBER \
      --head-repo <isolated-current-head-checkout> \
      --preview-out <retained-ignored-preview.json>

    It reads the saved independent review and original controller receipt; no model review is run. Show the concrete preview, effect type and digest before asking for any missing authority. Keep this preview under ignored .p2p/work/<slug>/ to retain state across interruptions.

Publication and readback

A clean REVIEWED verdict publishes COMMENT, not GitHub APPROVE. Material CHANGES NEEDED findings require a separately granted pr-review-request-changes effect (never infer it from a PR-create grant). Do not publish a BLOCKED review. Leave inline comments out unless the original saved finding identifies a line that still exists at the exact PR head; a top-level review covers the normal case.

Use the existing controller's authorize-effect against this exact preview SHA-256, with --action pr-review-comment (or pr-review-request-changes), exact repository and PR URL. Alternatively the human must expressly approve the complete exact preview and effect. Approval of PR creation alone is not authority to submit a review. Save the authorization result and reread it.

After authorizing, run with --publish-preview <retained-ignored-preview.json> --authorization <saved-authorization.json> or an exact --approved-sha256 and retrievable --approval-source human approval. The script compares the saved review, contract, source, PR metadata and current head again, reads all existing reviews, and reconciles a matching prior effect first. After any uncertain write, it reads all reviews again. Never repeat an uncertain submission without first reconciling it by exact marker and body. An unchanged retry reuses the same review; a new reviewed candidate has a distinct identity and may receive a distinct separately authorized review.

Return the verified review URL, effect type, the saved review/proof status, and the still-unassessed merge readiness. No publication action changes the source issue, candidate, proof or human GitHub reviews. For read-only reconciliation use --status-preview <retained-ignored-preview.json>. If GitHub is unavailable or identities differ, preserve the preview and report PARTIAL or BLOCKED; never claim that review was published.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.