---

name: code-reviewer
description: AI-powered code review for pull requests
model: claude-opus-4-6

---

AI Code Reviewer

Automatically review code changes for quality, security, and best practices.

What It Checks

• Security: SQL injection, XSS, hardcoded secrets, insecure dependencies
• Performance: N+1 queries, unnecessary re-renders, memory leaks, blocking I/O
• Best Practices: SOLID principles, DRY violations, error handling gaps
• Style: Naming conventions, file organization, import ordering
• Tests: Missing test coverage, flaky test patterns, assertion quality

Supported Languages

TypeScript, JavaScript, Python, Go, Rust, Java, C#, Ruby, PHP, Swift, Kotlin, C++, SQL, Shell, Terraform

Usage (illustrative custom-wrapper commands)

These are not built-in Claude Code subcommands. They require a custom wrapper; consult the wrapper's documentation before use.

claude review --diff HEAD~1
# or review a specific PR
claude review --pr 42

Configuration

For a custom reviewer that reads this configuration, create .claude/review.json in your repo; Claude Code does not automatically load it as a standard review configuration:

{
  "severity": "medium",
  "ignore": ["vendor/", "generated/"],
  "focus": ["security", "performance"],
  "max_comments": 15
}

Example Output

⚠ HIGH: SQL injection risk in user_query (line 42)
  Suggestion: Use parameterized queries instead of string concatenation

💡 MEDIUM: Consider memoizing expensive computation in useEffect (line 89)
  The dependency array changes on every render

✓ GOOD: Proper error boundary implementation