Stats
Actions
Tags
From pr-review-toolkit
This skill should be used when the user asks to "review my PR", "check my code before merging", "run a PR review", "analyze this pull request", "review my changes", or needs a comprehensive multi-agent pull request review covering code quality, test coverage, error handling, type design, comments, and code simplification.
How this skill is triggered — by the user, by Claude, or both
Slash command
/pr-review-toolkit:reviewing-pull-requestsThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Run multi-agent pull request reviews where each specialized agent focuses on a different aspect of code quality, then aggregate results into an actionable summary.
Run multi-agent pull request reviews where each specialized agent focuses on a different aspect of code quality, then aggregate results into an actionable summary.
Six specialized review agents are available:
| Agent | Focus | When Applicable |
|---|---|---|
| code-reviewer | CLAUDE.md compliance, bugs, general quality | Always |
| pr-test-analyzer | Test coverage quality, critical gaps | Test files changed |
| silent-failure-hunter | Silent failures, catch blocks, error logging | Error handling changed |
| type-design-analyzer | Type encapsulation, invariant expression | New types added/modified |
| comment-analyzer | Comment accuracy, documentation completeness | Comments/docs added |
| code-simplifier | Code clarity, readability, project standards | After passing review |
git diff --name-onlygh pr viewBased on the changes:
Sequential approach (default):
Parallel approach (on user request):
After agents complete, organize findings:
# PR Review Summary
## Critical Issues (X found)
- [agent-name]: Issue description [file:line]
## Important Issues (X found)
- [agent-name]: Issue description [file:line]
## Suggestions (X found)
- [agent-name]: Suggestion [file:line]
## Strengths
- What's well-done in this PR
## Recommended Action
1. Fix critical issues first
2. Address important issues
3. Consider suggestions
4. Re-run review after fixes
Reviews code against project guidelines in CLAUDE.md with confidence-based filtering (only reports issues with confidence >= 80). Checks for CLAUDE.md compliance, bug detection, and code quality. Groups issues by severity (Critical: 90-100, Important: 80-89).
Focuses on behavioral coverage rather than line coverage. Identifies critical code paths, edge cases, and error conditions that must be tested. Rates criticality from 1-10 and maps to severity (9-10: CRITICAL, 7-8: HIGH, 5-6: MEDIUM, 1-4: LOW).
Analyzes for silent failures, empty catch blocks, broad exception catching, inadequate error messages, and unjustified fallback behavior. Checks logging quality, user feedback, catch block specificity, and error propagation.
Evaluates new types on four dimensions (each rated 1-10): encapsulation, invariant expression, invariant usefulness, and invariant enforcement. Flags anti-patterns like anemic domain models, exposed mutable internals, and missing construction validation.
Verifies comment factual accuracy against code, assesses completeness, evaluates long-term value, and identifies misleading elements. Categorizes findings as critical issues, improvement opportunities, and recommended removals.
Simplifies code for clarity and maintainability while preserving all functionality. Applies project standards, reduces unnecessary complexity, eliminates redundant code, and improves naming. Avoids nested ternaries and over-compact solutions.
Full review (default): Run all applicable review agents based on changed files.
Targeted review:
Specify aspects to focus on: tests, errors, comments, types, code, simplify.
Parallel review: Request all agents run simultaneously for faster results.
Before committing:
Before creating PR:
After PR feedback:
npx claudepluginhub jawhnycooke/claude-plugins --plugin pr-review-toolkitReviews PR diffs across code quality, test coverage, error handling, type design, and comment quality with severity-ranked findings.
Systematically evaluates pull requests for correctness, design, readability, and tests, delivering actionable and kind feedback based on Google's engineering practices.