agentleFS
idavidov13idavidov13/agentic-playwright/.agents/skills/pr-reviewer/SKILL.md

pr-reviewer

Skill294 starsChanged yesterdayMITView on GitHub

Use when the user names a branch with any review intent ("review PR for branch X", "is branch Y good to merge", "check the changes on Z"). Not for the uncommitted working tree, or for authoring new tests or page objects (see `ai-native-workflow`).

23 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

3 rules. 0 are common, 2 are distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 23 edited versions we read keep these.

  • 2

    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:

    Route every changed path to its skills and read them: before judging; the Constitution applies everywhere.

    line 12
  2. Very specific:

    Read the whole diff first: git diff "$BASE"...HEAD --stat, then git diff "$BASE"...HEAD.

    line 22
  3. Distinctive:

    Run the scan script, then walk the checklist against the diff.

    line 43

“Common” is measured: kept by at least 6 in 10 of the 23 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

PR Reviewer

Critical

  • Read-only until the report is delivered. Fixes only when the user asks after reading it; a commit only on explicit approval.
  • Always the three-dot diff against the remote base (git diff "$BASE"...HEAD with BASE=origin/<name>); a two-dot diff reviews commits the author never wrote.
  • Route every changed path to its skills and read them before judging; the Constitution applies everywhere.
  • Run the checks, do not only read: scripts/verify-changed.sh and the affected specs.
  • The Constitution checklist is a gate before the report: scripts/scan-constitution.sh in this skill's directory, then references/constitution-checklist.md with file:line evidence; every ❌ is a finding.
  • Every finding is tied to a rule, a bug or a missing case; a style opinion without a rule is noise.
  • Env failures are not branch failures. A run that dies on a missing env var or token is reported as such; never invent or stub a credential.
  • A schema loosened to swallow a runtime surprise is a must-fix finding (z.strictObject → z.object, .optional(), .nullable(), .catchall(), a widened validator); the sanctioned fix is test.skip + // FIXME. Never --approve or --request-changes on the user's behalf.

Procedure

  1. scripts/checkout-branch.sh <branch> [base] resolves the base, refuses a dirty tree, fetches and switches, prints BASE and ORIGINAL_REF; a STOP line means ask the user.
  2. Read the whole diff first: git diff "$BASE"...HEAD --stat, then git diff "$BASE"...HEAD.
  3. Route with the table and read each listed skill; a path outside the table takes the closest indexed skill.

    Changed pathRead
    tests/**/api/**api-testing, test-standards, type-safety
    tests/**/e2e/**, tests/**/functional/**test-standards, page-objects, fixtures
    tests/**/*.setup.tshelpers, fixtures, test-standards
    pages/**page-objects, selectors, playwright-cli
    fixtures/api/schemas/**type-safety, api-testing
    fixtures/**fixtures, helpers
    test-data/factories/**data-strategy, type-safety
    test-data/static/**data-strategy, refactor-values, type-safety
    enums/**, new membersenums
    enums/**, test-data/static/**, changed existing valuesrefactor-values
    config/**config
    helpers/**helpers
    any **/*.tstype-safety
  4. scripts/verify-changed.sh "$BASE" (eslint, prettier, tsc on the changed files), then the affected specs per references/verification.md.
  5. Review the diff against the routed skills: the Constitution rows for the area, contract fidelity, consistency with the nearest sibling file, real bugs (wrong assertion, .find() on a possibly empty response, copy-paste leftovers), coverage gaps, scope creep.
  6. Run the scan script, then walk the checklist against the diff; ➖ only when the branch does not touch the area, "could not evaluate" recorded as such, never as ✅.
  7. Report in chat with assets/report-template.md and the confidence rubric in references/reporting.md; post to the PR only when asked. Then git switch "$ORIGINAL_REF".
  8. Offer fixes ("Want me to implement any of these findings?"); apply only the chosen ones per the routed skills, re-run step 4, then ask before committing (Conventional Commit subject, co-author trailer).

References

  • references/constitution-checklist.md: step 6, the 31-row gate.
  • references/verification.md: step 4, the tsc detour, running the affected specs, the env limit.
  • references/reporting.md: step 7, the confidence rubric and posting to GitHub.
  • assets/report-template.md: step 7, the report shape.
  • assets/review-prompt.md: a paste-ready invocation for the user.
  • scripts/checkout-branch.sh, scripts/verify-changed.sh, scripts/scan-constitution.sh: steps 1, 4 and 6.

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.