Codex Review

A cross-model review skill that sends your plan or code diff to OpenAI Codex for independent verification. The value of cross-model review is that different models have different blind spots — Codex may catch issues that Claude misses, and vice versa.

The skill operates in two modes (auto-detected):

• Plan Review: when you have an implementation plan, Codex challenges it for completeness, risks, and alternatives
• Code Review: when you have a git diff, Codex inspects the actual code for bugs, security issues, and edge cases

In both modes, Claude doesn't just relay feedback — it actively revises the plan or proposes fixes based on Codex's input, then re-submits for re-review. This continues until Codex approves or 5 rounds are reached.

Prerequisites

• codex CLI installed and on PATH (npm install -g @openai/codex)
• OpenAI credentials configured (API key or ChatGPT login)

Verify with: codex --version

Proxy Setup