Code Review Agent

Universal code review skill for Claude Code and OpenCode AI agents. Performs structured reviews with a senior engineer lens, covering SOLID principles, security, performance, and code quality.

Features

• ✅ SOLID Principles - Detect SRP, OCP, LSP, ISP, DIP violations
• 🔒 Security Scan - XSS, injection, SSRF, race conditions, auth gaps, secrets leakage
• ⚡ Performance - N+1 queries, CPU hotspots, missing cache, memory issues
• 🛡️ Error Handling - Swallowed exceptions, async errors, missing boundaries
• 📏 Boundary Conditions - Null handling, empty collections, off-by-one, numeric limits
• 🗑️ Removal Planning - Identify dead code with safe deletion plans

Installation

Install globally via npm:

npm install -g code-review-agent

Or use npx without installation:

npx code-review-agent init --ai claude

Usage

Initialize in Your Project

Navigate to your project directory and run:

For Claude Code:

code-review-agent init --ai claude

For OpenCode:

code-review-agent init --ai opencode

This will:

1. Create the skill directory structure (.claude/skills/code-review-expert/ or .opencode/skills/code-review-expert/)
2. Copy all skill files and reference checklists
3. Update CLAUDE.md or AGENTS.md with skill reference

Run Code Review

After initialization, simply run in your AI agent:

/code-review-expert

The skill will automatically review your current git changes.

Command Options

code-review-agent init [options]

Options:
  --ai <platform>   AI platform: claude or opencode (default: "claude")
  --force          Overwrite existing files
  -h, --help       Display help
  -V, --version    Display version

Workflow

1. Preflight - Scope changes via git diff
2. SOLID + Architecture - Check design principles
3. Removal Candidates - Find dead/unused code
4. Security Scan - Vulnerability detection
5. Code Quality - Error handling, performance, boundaries
6. Output - Findings by severity (P0-P3)
7. Confirmation - Ask user before implementing fixes

Severity Levels

| Level | Name | Action | |-------|------|--------| | P0 | Critical | Must block merge | | P1 | High | Should fix before merge | | P2 | Medium | Fix or create follow-up | | P3 | Low | Optional improvement |

Directory Structure

After initialization, your project will have:

your-project/
├── .claude/                     # For Claude Code
│   └── skills/
│       └── code-review-expert/
│           ├── SKILL.md
│           ├── agent.yaml
│           └── references/
│               ├── solid-checklist.md
│               ├── security-checklist.md
│               ├── code-quality-checklist.md
│               └── removal-plan.md
└── CLAUDE.md                    # Updated with skill reference

# OR

your-project/
├── .opencode/                   # For OpenCode
│   └── skills/
│       └── code-review-expert/
│           ├── SKILL.md
│           ├── agent.yaml
│           └── references/
│               ├── solid-checklist.md
│               ├── security-checklist.md
│               ├── code-quality-checklist.md
│               └── removal-plan.md
└── AGENTS.md                    # Updated with skill reference

Examples

Example 1: Review with Fix

User: /code-review-expert

Agent: [Performs review, finds 3 P1 issues, 5 P2 issues]

Agent: How would you like to proceed?
1. Fix all
2. Fix P0/P1 only
3. Fix specific items
4. No changes

User: 2

Agent: [Implements fixes for P1 issues]

Example 2: Review Only

User: /code-review-expert

Agent: [Performs review, finds no critical issues]

Agent: ✅ Clean review - no blocking issues found.
Checked: SOLID, security, performance, error handling

Customization

You can customize the skill by editing the installed files:

• SKILL.md - Modify workflow and review criteria
• references/*.md - Add/remove checklist items
• agent.yaml - Change display name or default prompt

Contributing

Issues and pull requests are welcome!

License

MIT

Related Projects

• Claude Code - AI-powered coding assistant
• OpenCode - Open source AI coding platform
• code-review-expert - Original skill implementation

Support

For issues or questions:

• File an issue on GitHub
• Check existing documentation in the skill files