Codex skills use the same SKILL.md format as Claude Code, read from .agents/skills. Every skill in this directory installs into Codex with one command, and the proofs show what each one changes before you add it. This directory indexes 1,402 of them from public repositories, refreshed weekly, and shows for each proven one what it changed.
Project (this repo only)
Use This Skill.agents/skills/<name>/SKILL.mdPersonal (every project)
~/.agents/skills/<name>/SKILL.mdRun /skills or type $name in the CLI or IDE extension. In ChatGPT, type @ to pick a skill. Codex also picks skills on its own when the task matches. Official OpenAI Codex skills docs →
Full walkthrough for every agent: how to use a skill.
Wrote one of these? Prove it for $49 and get a badge for your README.
Codex skills use the same SKILL.md format as Claude Code, read from .agents/skills. Every skill in this directory installs into Codex with one command, and the proofs show what each one changes before you add it.
Copy the skill folder to .agents/skills/<name>/SKILL.md in your project, or ~/.agents/skills/<name>/SKILL.md for every project. Each skill page here has a one-line install command that does this for you.
Run /skills or type $name in the CLI or IDE extension. In ChatGPT, type @ to pick a skill. Codex also picks skills on its own when the task matches.
A proven skill was run through Claude twice on a fixed brief, once without the skill and once with it, on the same model. Both results are live pages, and the brief, model and skill commit are recorded so anyone can repeat the comparison.
Yes. Every skill belongs to its author and links to its source repository with its license. Browsing, installing and the first proof are free; a $99 one-time Pro unlock opens every proof.
