AgentSkillsCN

code-reviewer

对拉取请求与代码变更进行质量、安全性和最佳实践方面的审查。当用户提出代码审查、PR 审查,或希望对代码变更进行评估时,可使用此技能。

SKILL.md
--- frontmatter
name: code-reviewer
description: Reviews pull requests and code changes for quality, security, and best practices. Use when user asks for code review, PR review, or mentions reviewing changes.
allowed-tools: Read, Grep, Glob, Bash, WebFetch, WebSearch
metadata:
  hooks:
    after_complete:
      - trigger: self-improving-agent
        mode: background
        reason: "Learn from review patterns"
      - trigger: session-logger
        mode: auto
        reason: "Log review activity"

Code Reviewer

A comprehensive code review skill that analyzes pull requests and code changes for quality, security, maintainability, and best practices.

When This Skill Activates

This skill activates when you:

  • Ask for a code review
  • Request a PR review
  • Mention reviewing changes
  • Say "review this" or "check this code"

Review Process

Phase 1: Context Gathering

  1. Get changed files

    bash
    git diff main...HEAD --name-only
    git log main...HEAD --oneline
    
  2. Get the diff

    bash
    git diff main...HEAD
    
  3. Understand project context

    • Read relevant documentation
    • Check existing patterns in similar files
    • Identify project-specific conventions

Phase 2: Analysis Categories

1. Correctness

  • Logic is sound and matches requirements
  • Edge cases are handled
  • Error handling is appropriate
  • No obvious bugs or typos

2. Security

  • No hardcoded secrets or credentials
  • Input validation and sanitization
  • SQL injection prevention
  • XSS prevention (for frontend)
  • Authentication/authorization checks
  • Safe handling of user data

3. Performance

  • No N+1 queries
  • Appropriate caching
  • Efficient algorithms
  • No unnecessary computations
  • Memory efficiency

4. Code Quality

  • Follows DRY principle
  • Follows KISS principle
  • Appropriate abstractions
  • Clear naming conventions
  • Proper typing (if TypeScript)
  • No commented-out code

5. Testing

  • Tests cover new functionality
  • Tests cover edge cases
  • Test assertions are meaningful
  • No brittle tests

6. Documentation

  • Complex logic is explained
  • Public APIs have documentation
  • JSDoc/TSDoc for functions
  • README updated if needed

7. Maintainability

  • Code is readable
  • Consistent style
  • Modular design
  • Separation of concerns

Phase 3: Output Format

Use this structured format for review feedback:

markdown
# Code Review

## Summary
Brief overview of the changes (2-3 sentences).

## Issues by Severity

### Critical
Must fix before merge.

- [ ] **Issue Title**: Description with file:line reference

### High
Should fix before merge unless there's a good reason.

- [ ] **Issue Title**: Description with file:line reference

### Medium
Consider fixing, can be done in follow-up.

- [ ] **Issue Title**: Description with file:line reference

### Low
Nice to have improvements.

- [ ] **Issue Title**: Description with file:line reference

## Positive Highlights
What was done well in this PR.

## Suggestions
Optional improvements that don't require immediate action.

## Approval Status
- [ ] Approved
- [ ] Approved with suggestions
- [ ] Request changes

Common Issues to Check

Security Issues

IssuePatternRecommendation
Hardcoded secretsconst API_KEY = "sk-"Use environment variables
SQL injection\"SELECT * FROM...\" + user_inputUse parameterized queries
XSS vulnerabilityinnerHTML = user_inputSanitize or use textContent
Missing auth checkNew endpoint without @RequireAuthAdd authentication middleware

Performance Issues

IssuePatternRecommendation
N+1 queryLoop with database callUse eager loading or batch queries
Unnecessary re-renderMissing dependencies in useEffectFix dependency array
Memory leakEvent listener not removedAdd cleanup in useEffect return
Inefficient loopNested loops O(n²)Consider hash map or different algorithm

Code Quality Issues

IssuePatternRecommendation
Duplicate codeSimilar blocks repeatedExtract to function
Magic numberif (status === 5)Use named constant
Long functionFunction >50 linesSplit into smaller functions
Complex condition`a && b

Testing Issues

IssuePatternRecommendation
No testsNew feature without test fileAdd unit tests
Untested edge caseTest only covers happy pathAdd edge case tests
Brittle testTest relies on implementation detailsTest behavior, not implementation
Missing assertionTest doesn't assert anythingAdd proper assertions

Language-Specific Guidelines

TypeScript

  • Use unknown instead of any for untyped values
  • Prefer interface for public APIs, type for unions
  • Use strict mode settings
  • Avoid as assertions when possible

React

  • Follow Hooks rules
  • Use useCallback/useMemo appropriately (not prematurely)
  • Prefer function components
  • Use proper key props in lists
  • Avoid prop drilling with Context

Python

  • Follow PEP 8 style guide
  • Use type hints
  • Use f-strings for formatting
  • Prefer list comprehensions over map/filter
  • Use context managers for resources

Go

  • Handle errors explicitly
  • Use named returns for clarity
  • Keep goroutines simple
  • Use channels for communication
  • Avoid package-level state

Before Approving

Confirm the following:

  • All critical issues are addressed
  • Tests pass locally
  • No merge conflicts
  • Commit messages are clear
  • Documentation is updated
  • Breaking changes are documented

Scripts

Run the review checklist script:

bash
python scripts/review_checklist.py <pr-number>

References

  • references/checklist.md - Complete review checklist
  • references/security.md - Security review guidelines
  • references/patterns.md - Common patterns and anti-patterns