This hooks Claude up to OpenAI's Codex CLI for code analysis, refactoring, and automated editing. You pick a GPT model (5.3 through 5.5), set reasoning effort, choose a sandbox mode (read-only for reviews, workspace-write for edits, danger-full-access when you need network), and fire off tasks. The skill handles stdin quirks and stderr filtering automatically, lets you resume sessions to iterate on changes, and crucially, treats Codex output as peer review rather than gospel. If Codex says something that contradicts what you know or recent API changes, the skill encourages you to push back and fact-check before accepting its suggestions. Useful when you want another AI's perspective on gnarly code without leaving your Claude workflow.
npx skills add https://github.com/skills-directory/skill-codex --skill codex