Pr review
Reviews a GitHub pull request with a multi-model panel and posts the result to GitHub. Use when the user says "review this PR", "check this pull request", "review PR #123", "is this PR ready to merge", or provides a GitHub PR URL. Also use before merging any PR. Do NOT use for local code review without a PR, for reviewing uncommitted changes, or on claude.ai — it needs the local gh, codex, and agy CLIs.From its SKILL.md
npx -y skills add zachmayer/skills --skill pr-reviewAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 2 stars2 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
SKILL.md
1.7 KB, as published. Nobody here has run it
Thin wrapper around discussion-partners. That skill handles the parallel
multi-model review; this one stages the PR context and posts the result.
1. Fetch PR context
gh pr view <N> [--repo owner/repo] --json title,body,author,baseRefName,headRefName,files,url
gh pr diff <N> [--repo owner/repo]
Also read CLAUDE.md for repo conventions.
2. Run discussion-partners
Pass PR metadata + full diff + repo conventions as shared context. Invoke
discussion-partners in its default config. Framing: review for correctness,
security, design, and provenance (what the diff invents when something
already owns the job — hand-rolled library stand-ins, unowned shapes,
twice-homed facts, forms foreign to their consumer); skip style nits —
provenance findings are not style nits; LGTM is fine. For data science
code, load the data-science skill.
3. Triage and post
Drop false positives and style nits. Priority: correctness > security > design > provenance > performance. If all reviewers flag the same thing, it's real.
gh pr review <N> --comment --body "..." # issues found
gh pr review <N> --approve --body "..." # clean
Don't invent issues. Don't pad with filler.