Codex Code Review Skill for Claude Code
Five-perspective code review as a Claude Code skill. Security, correctness, compliance, performance, maintainability.
Install
# 1. Install the skill mkdir -p ~/.claude/skills/codex-review curl -sL https://raw.githubusercontent.com/pauhu/claude-codex-review/main/skills/codex-review/SKILL.md \ -o ~/.claude/skills/codex-review/SKILL.md
Restart Claude Code. That's it.
Share with your team
Drop the skill into your repo. Everyone who clones gets it:
your-project/
└── .claude/
└── skills/
└── codex-review/
└── SKILL.md
Use
Reviews uncommitted changes. Also accepts:
/codex-review src/auth.ts– review a specific file/codex-review branch– review all changes vs main/codex-review commit– review the last commit
How it works
The SKILL.md defines a four-step review:
- Scope – determines what code to review
- Second opinion – tries Codex MCP, then Codex CLI, then adversarial self-review
- Five perspectives – security, correctness, compliance, performance, maintainability
- Output – severity table with actionable findings
With Codex (optional)
If you have an OpenAI API key and Codex CLI, the skill uses it as a genuine second opinion – a different model reviewing your code independently.
npm install -g @openai/codex claude mcp add codex -s user -- codex mcp-server
Without Codex, the skill still works. It falls back to adversarial self-review where Claude re-examines the code assuming it's broken.
Experimental: Codex with Claude Max
The proxy/ folder contains an experimental adapter that routes Codex through your Claude Max subscription instead of an OpenAI key. This is work in progress – Codex's agentic tools (exec_shell, apply_patch) don't fully work with Claude as the backend model yet. The proxy infrastructure is ready for when Codex improves its model-agnostic support.
Uninstall
rm -rf ~/.claude/skills/codex-reviewLicense
MIT – Pauhu