talendar/github-pull-request-review

PR Review Workflow

Follow these phases in order. The user provides a PR number (e.g., #123 or 123). The repo must be cloned locally and gh must be authenticated.

Phase 1 — Gather Context

1. Fetch PR metadata:

   gh pr view <PR> --json title,body,baseRefName,headRefName,state,labels,milestone,author,reviewRequests,reviews,additions,deletions,changedFiles
   

2. Fetch PR comments (PR-level discussion):

   gh api --paginate repos/{owner}/{repo}/issues/{pr_number}/comments --jq '.[] | {body, user: .user.login, created_at}'
   

3. Fetch inline review comments (code-level — the issues endpoint above does NOT include these):

   gh api --paginate repos/{owner}/{repo}/pulls/{pr_number}/comments --jq '.[] | {path, line, body, user: .user.login}'
   

   Auto-detect {owner}/{repo} with: gh repo view --json nameWithOwner --jq .nameWithOwner

4. Fetch the diff:

   gh pr diff <PR>
   

   If the diff exceeds 50 files, list changed files first with gh api --paginate repos/{owner}/{repo}/pulls/{pr}/files --jq '.[].filename' and read them individually.

5. Extract linked issues from the PR body. Parse patterns: Fixes #N, Closes #N, Resolves #N, Related to #N, or bare #N. For each:

   gh issue view <N> --json title,body,comments,labels,state
   

6. Check CI status:

   gh pr checks <PR>
   

   If any checks failed, fetch logs: gh run view <RUN_ID> --log-failed

You may also run scripts/fetch-pr-context.sh <PR_NUMBER> to automate steps 1-6.

Phase 2 — Understand Repo Conventions

Read each file if it exists:

• CONTRIBUTING.md, CODE_OF_CONDUCT.md, CODEOWNERS
• .github/PULL_REQUEST_TEMPLATE.md
• README.md (project overview and architecture)
• AGENTS.md or CLAUDE.md (agent-specific instructions)

Detect linter/formatter configs to understand style expectations:

• .eslintrc*, biome.json, .prettierrc*, pyproject.toml, ruff.toml, .rubocop.yml, .golangci.yml, etc.

Read CI workflow files (.github/workflows/*.yml) to understand existing automated checks.

Phase 3 — Analyze the Code Changes

Review the diff against these priorities. See references/review-checklist.md for detailed guidance.

1. Goal alignment — Does the PR achieve what the linked issue(s) describe? Missing requirements? Scope creep?
2. Bugs and logic errors — Off-by-one, null handling, race conditions, missing edge cases, broken error handling.
3. Security risks — Only HIGH-confidence findings. Trace data flow from source to sink. See the checklist for common false positives to avoid.
4. Codebase consistency — Compare against surrounding code, not abstract ideals. Naming, file organization, import patterns, architecture.
5. Test coverage — New paths tested? Existing tests updated? Obvious untested edge cases?
6. Performance — N+1 queries, unbounded loops, missing pagination, large memory allocations.
7. Documentation — Public APIs documented? Complex algorithms explained? Breaking changes noted?

Phase 4 — Run Available Checks

Detect available commands by running scripts/detect-repo-checks.sh or manually inspecting package.json, Makefile, pyproject.toml, and CI workflow files.

Run in this order (stop if a step fails catastrophically):

1. Lint (e.g., npm run lint, ruff check ., golangci-lint run)
2. Type-check (e.g., npx tsc --noEmit, mypy ., pyright)
3. Tests — prefer running only tests relevant to changed files when possible

Guardrails:

• Do NOT run commands that could modify state (deploy, migrate, publish, seed).
• If unsure what a command does, skip it and note it in the report.
• If no checks are available, note this and move on.

Phase 5 — Produce the Report

Generate a structured report. See references/report-format.md for the full template.

## PR Review: #<number> — <title>
### Summary
### Verdict: ✅ APPROVE / ⚠️ APPROVE WITH COMMENTS / ❌ REQUEST CHANGES
### Blocking Issues (must fix before merge)
### Non-blocking Issues (should fix)
### Positive Observations
### Checks (table: Lint, Type-check, Tests, CI)
### Context (linked issues, files changed, additions/deletions)

Rules:

• Rank blocking issues by severity: security > bugs > logic > breaking changes.
• Every issue must reference a specific file and line.
• If nothing significant is found, say so — do not invent issues.
• Include positive feedback for well-written code.

Phase 6 — Post Review on GitHub (only if user requests)

If the user asks to post the review on the PR:

• APPROVE: gh pr review <PR> --approve --body "<summary>"
• REQUEST CHANGES: gh pr review <PR> --request-changes --body "<report>"
• APPROVE WITH COMMENTS: gh pr review <PR> --comment --body "<report>"

You may use scripts/post-review-comment.sh <PR> <verdict> <body_file>.

Always confirm with the user before posting.

References

• references/review-checklist.md — Detailed review checklist by category
• references/report-format.md — Report template with good/bad finding examples
• references/gh-cli-reference.md — gh CLI commands and API endpoints cheat sheet