Review Skill
Purpose: Perform exhaustive code reviews using multi-agent analysis across 13+ specialized reviewer agents.
Review Context
Review target: ${ARGUMENTS}
Review Methodology
Phase 1: Setup and Context Gathering
- Identify Review Target
- Determine if reviewing a PR, branch, or commit - Gather basic information about the changes
- Setup Work Environment
- Check if already on review branch - If not, consider using git-worktree for isolated review - Skill: git-worktree (if needed)
- Initial Scan
- Get overview of changed files and lines - Identify major areas of change
Phase 2: Parallel Agent Reviews (13+ Reviewers)
Launch specialized reviewer agents in parallel:
Architecture Reviewers:
- System Architecture Reviewer - Overall system design and boundaries
- Domain Model Reviewer - Business logic and data modeling
- API Design Reviewer - Endpoint design and contracts
Code Quality Reviewers: 4. Patterns & Practices Reviewer - Design patterns and best practices 5. Performance Reviewer - Efficiency and scalability considerations 6. Security Reviewer - Security vulnerabilities and best practices 7. Test Coverage Reviewer - Testing strategy and coverage
Language & Framework Reviewers: 8. Language-Specific Reviewer - Language idioms and conventions 9. Framework Reviewer - Framework-specific patterns and APIs
Integration & Operations Reviewers: 10. Integration Reviewer - External dependencies and APIs 11. Database Reviewer - Data modeling and query optimization 12. UI/UX Reviewer - User interface and experience considerations 13. DevOps & Observability Reviewer - Deployment, monitoring, logging
Phase 3: Analysis and Synthesis
- Collect Agent Findings
- Gather reports from all reviewer agents - Identify common themes and patterns - Prioritize issues by severity and impact
- Create Structured Review
- Categorize findings by type and priority - Provide specific, actionable feedback - Include code examples and suggestions
- Generate Review Summary
- Overall assessment of code quality - Key strengths and areas for improvement - Recommendations for next steps
Phase 4: Report and Follow-up
- Create Review Document
- Write comprehensive review report - Include detailed findings with code references - Suggest fixes and improvements
- Present Findings
- Share review summary with user - Highlight critical issues first - Offer to help with specific fixes
- Next Steps Options
- Fix issues identified in review - Run additional validation tests - Proceed with merge/approval process
Review Output
- Structured Review Document: Detailed analysis with categorized findings
- Priority Classification: P1 (Critical), P2 (Important), P3 (Nice-to-have)
- Actionable Recommendations: Specific fixes and improvements
- File Path References: Exact locations of issues (e.g.,
app/services/example.rb:42)
Severity Classification
P1 (Critical): Must fix before merging
- Security vulnerabilities
- Breaking API changes without migration
- Data loss risks
- Performance regressions > 50%
P2 (Important): Should fix soon
- Code quality issues affecting maintainability
- Missing error handling
- Incomplete test coverage
- Minor performance issues
P3 (Nice-to-have): Optional improvements
- Code style nitpicks
- Documentation improvements
- Test optimization
- Refactoring opportunities