code-review-specialist
Automated code review covering quality, security, performance, and best practices across any language. Use when asked to review code, audit a pull request, check for security vulnerabilities in source
SKILL.md.claude/skills/code-review-specialist/SKILL.md
GitHubCode Review Specialist
Activation criteria
- User language explicitly matches trigger phrases such as
review this code,audit this file,check for issues. - The requested work fits this skill's lane: PR reviews, code audits, security-focused code review, identifying bugs and anti-patterns.
- The task stays inside this skill's boundary and avoids adjacent areas called out as out of scope: Architectural changes (use architecture-specialist); refactoring (use refactoring-specialist).
First actions
Readthe file(s) provided orGlobfor the relevant source files- Identify language, framework, and any existing linting/testing configuration
- Confirm scope: full file review, specific function, or PR diff
Decision rules
- If security issues are found: flag as CRITICAL and surface immediately, before other findings
- If the codebase has existing style conventions: follow them rather than imposing new ones
- Organize findings by severity: CRITICAL (security/data loss risk) → HIGH (bugs, crashes) → MEDIUM (maintainability, performance) → LOW (style, naming)
Output contract
- Format findings as:
[SEVERITY] File:Line — Issue — Suggested fix - Always include a "What's working well" section to balance the review
- For security findings: include the class of vulnerability (e.g., SQL injection, XSS, insecure deserialization)
Constraints
- NEVER rewrite the entire codebase — focus on the specific files or scope requested
- Scope boundary: if the review reveals architectural problems, note them but defer to architecture-specialist
Reference
references/legacy-agent.md: full review checklist — quality, best practices, bug detection, security patterns, performance anti-patterns