agentleFS
qrqr2100qrqr2100/ship-gate-kit/.claude/skills/review-checklist/SKILL.md

review-checklist

Skill0 starsChanged 5 months agoMITView on GitHub

Judge a pre-ship artifact (code diff, PR, config, doc) with the Ship Gate Kit gates. Report PASS/WARN/FAIL only, without fixing. Triggers — "run the gates", "check before sending", "ship gate", "ship verdict", "is this OK to ship?", "/review-checklist".

8 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

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

  • 0

    Common

    Most of the 8 edited versions we read keep these.

  • 6

    Distinctive

    This author's own choices. Worth a look.

  • 1

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    If ambiguous, ask the user which gates to apply.

    line 20
  2. Very specific:

    If cross-check is on, fill the cross-check columns/judge names/count from cross-check/judges.yaml (not the example's fixed Claude/Codex — …

    line 35
  3. Distinctive:

    No FAIL, WARN only → WARN (user decides knowingly).

    line 39
  4. Distinctive:

    All pass → PASS (proceed with a standard commit).

    line 40
  5. Distinctive:

    Don't invent findings. If unsure, leave it WARN and write "needs confirmation".

    line 44
  6. Distinctive:

    Don't use self-reported numbers as evidence.

    line 45
  7. Distinctive:

    Don't claim a gate hit rate as fact.

    line 46

“Common” is measured: kept by at least 6 in 10 of the 8 edited versions of this file we could read. “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

review-checklist — pre-ship gate judging skill

This skill does not change your output. It judges and reports by gate criteria only. It does not invent findings (false-positive suppression). It reports only what it found, with severity.

Workflow (5 steps)

1. Input & gate selection

Identify what the user gave and pick the gates. Input is usually a git diff, but it doesn't have to be — a changed file, a pasted code snippet, or live code (running code / hotfix) is also a valid target.

  • code diff / file / snippet → SECRET-SCAN · DEPENDENCY · CODE-DIFF (SCOPE too if it's a git diff)
  • PR → the above + PR-DESCRIPTION
  • config change / config file → CONFIG-CHANGE · SECRET-SCAN
  • doc → DOCUMENT-GATE-TEMPLATE
  • no git / live system → apply SECRET-SCAN · DEPENDENCY · CODE-DIFF · CONFIG-CHANGE directly to the code/file. SCOPE needs the user to declare "the files I changed", and PR-DESCRIPTION only applies when there's a PR. If ambiguous, ask the user which gates to apply.

2. Per-gate check

Apply each selected gate's gates/*.md checklist/procedure as written. For DEPENDENCY·SCOPE, run the procedure (extract → verify → compare) in order.

3. Severity classification

Split findings into FAIL (must stop now) / WARN (can knowingly proceed).

  • FAIL = something whose damage is hard to undo if shipped (secret leak, malicious package, conflict marker, exposure widening, Go _ err discard on a core path).
  • WARN = recommended to clean up but doesn't block shipping itself.

4. Report

Report per-gate verdict + reason (line/file) + overall verdict. Do not modify the code.

  • Quick check in chat → templates/GATE-REPORT-TEMPLATE.md (markdown table).
  • Shareable report → generate gate-report.html in the exact templates/GATE-REPORT.html format (fill overall verdict, gate table, and next actions with the result; keep the design).
    • If cross-check is on, fill the cross-check columns/judge names/count from cross-check/judges.yaml (not the example's fixed Claude/Codex — engine-agnostic). If off, leave the cross-check section as "off".

5. Overall verdict

  • Any FAIL → overall FAIL (don't ship).
  • No FAIL, WARN only → WARN (user decides knowingly).
  • All pass → PASS (proceed with a standard commit).

Constraints (scope)

  • Don't fix the artifact. Judge and report only.
  • Don't invent findings. If unsure, leave it WARN and write "needs confirmation".
  • Don't use self-reported numbers as evidence.
  • Don't claim a gate hit rate as fact.

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.