AI Skill Report Card
Reviewing Pull Requests
Pull Request Review
Quick Start14 / 15
Bash# Review checklist for immediate assessment 1. Does it solve the stated problem? 2. Are there any obvious bugs or edge cases? 3. Is the code readable and well-structured? 4. Are tests adequate? 5. Any security concerns?
Recommendation▾
Add concrete input/output examples showing actual code snippets and specific review comments rather than general descriptions
Workflow14 / 15
Progress:
- Read the PR description and linked issues
- Check the diff size (>400 lines = request split)
- Review architecture/design changes first
- Examine code quality and patterns
- Verify test coverage and quality
- Check for security vulnerabilities
- Validate documentation updates
- Test locally if complex changes
Review Priority:
- Correctness (bugs, logic errors)
- Security (injection, auth, data exposure)
- Performance (algorithms, database queries)
- Maintainability (readability, patterns)
- Style (formatting, naming)
Recommendation▾
Include a template or framework section with standardized comment formats and approval criteria
Examples18 / 20
Example 1: Input: API endpoint that processes user input without validation Output: "❌ SECURITY: This endpoint is vulnerable to injection attacks. Add input validation and sanitization before processing user data."
Example 2: Input: Function with 8 parameters and 150 lines Output: "🔄 REFACTOR: This function violates SRP. Consider breaking into smaller functions or using a parameter object."
Example 3: Input: Database query in a loop Output: "⚡ PERFORMANCE: N+1 query detected. Use bulk operations or eager loading instead."
Recommendation▾
Expand the workflow to include specific tools, commands, or IDE features for efficient PR review
Best Practices
- Be constructive: Suggest solutions, not just problems
- Explain the why: Context helps developers learn
- Use prefixes: ❌ (blocking), ⚠️ (concern), 💡 (suggestion), 🔄 (refactor)
- Approve with minor comments: Don't block for style issues
- Test complex logic: Pull and run locally when needed
- Check commit messages: Should be clear and descriptive
Common Pitfalls
- Don't nitpick formatting if auto-formatters exist
- Don't review line-by-line for large PRs (focus on architecture)
- Don't approve without understanding the changes
- Don't request changes without explaining impact
- Don't ignore missing tests for new functionality
- Don't overlook error handling and edge cases