code-review-auto
Automatic code review and fixes before showing to user.Use when: code generation is complete, before preview.Triggers: internal use only.
When & Why to Use This Skill
This Claude skill automates the code review and optimization process by scanning generated code for quality, performance, and accessibility issues before it reaches the user. It functions as an invisible quality gate that automatically fixes common linting errors, refactors suboptimal structures, and ensures adherence to best practices, delivering production-ready code that is clean, efficient, and accessible.
Use Cases
- Case 1: Automated cleanup of code quality issues such as removing unused imports, variables, and debug logs to maintain a professional codebase.
- Case 2: Structural refactoring of generated components to ensure logical folder organization and prevent code duplication in complex projects.
- Case 3: Performance optimization by identifying and flagging N+1 queries or blocking operations before the code is executed in a preview environment.
- Case 4: Enhancing web accessibility by automatically generating missing alt text and verifying heading hierarchies to meet WCAG standards.
| name | code-review-auto |
|---|---|
| description | | |
| Use when | code generation is complete, before preview. |
| Triggers | internal use only. |
Automatic Code Review
Review and fix issues before user sees code.
Review Checklist
Code Quality
- No unused imports
- No unused variables
- No console.log in production
- Consistent naming conventions
- No magic numbers (use constants)
Structure
- Components properly extracted
- No code duplication
- Reasonable file sizes (<300 lines)
- Logical folder organization
Performance
- No N+1 queries
- Images optimized
- Lazy loading where appropriate
- No blocking operations
Accessibility
- Alt text on images
- Proper heading hierarchy
- Focusable interactive elements
- Color contrast sufficient
Auto-Fix Rules
| Issue | Fix |
|---|---|
| Unused import | Remove |
| console.log | Remove |
| Missing alt text | Generate from context |
| Duplicate code | Extract function |
| Long function | Split into smaller |
Process
- Scan generated code
- Identify issues (categorize by severity)
- Auto-fix what's possible
- Re-run tests to verify fixes don't break
- Continue to preview if all good
User Experience
User never sees:
- Linting errors
- Initial messy code
- Intermediate states
User always sees:
- Clean, reviewed code
- Working preview
- ✅ confirmation