Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/openai/skills/llms.txt

Use this file to discover all available pages before exploring further.

Agent Skills extend Codex with specialized capabilities that activate automatically when relevant. This guide walks you through discovering what skills are available, installing one, and using it in a real workflow — all from inside Codex itself, using the built-in $skill-installer system skill.

​
Prerequisites

Before you begin:
  • You need a Codex account with an active session.
  • System skills (skill-installer, skill-creator) are pre-installed in every Codex environment. You do not need to install them.
  • Curated and experimental skills require an explicit install step, followed by a Codex restart.

​
Installing and Using a Skill

1

Open Codex and list available skills

Start a new Codex session and type the following to invoke the skill-installer:
$skill-installer
Codex will call the list-skills.py helper script and display the available curated skills from skills/.curated, annotating any that are already installed:
Skills from openai/skills:
1. gh-address-comments
2. vercel-deploy (already installed)
3. ...
Which ones would you like installed?
2

Install a skill by name

Install the gh-address-comments skill by referencing it directly:
$skill-installer gh-address-comments
The installer downloads the skill from the curated catalog and places it at $CODEX_HOME/skills/gh-address-comments (which defaults to ~/.codex/skills/gh-address-comments).
3

Restart Codex

After installing any new skill, restart Codex so it picks up the new skill’s metadata.
Skills are discovered at startup. A restart is required before a newly installed skill becomes active. Codex will confirm the skill is available on your next turn.
4

Use the skill

In your next session, navigate to a repository with an open pull request and ask Codex to address the PR comments:
Address the review comments on my current PR
Codex matches the request to gh-address-comments, loads the skill, fetches the PR comments using the bundled scripts/fetch_comments.py, and asks you which comment threads to address before applying changes.

​
Installing Experimental Skills

Experimental skills live in skills/.experimental and are not included in the default curated listing. You can install them in two ways. By folder name: 
$skill-installer install the create-plan skill from the .experimental folder
By full GitHub URL: 
$skill-installer install https://github.com/openai/skills/tree/main/skills/.experimental/create-plan
Both approaches call install-skill-from-github.py under the hood and install the skill to $CODEX_HOME/skills/<skill-name>. Restart Codex after installing.

​
What Happens Under the Hood

When you run $skill-installer, Codex activates the skill-installer system skill, which uses two helper scripts:
  • list-skills.py — Queries the GitHub API for the contents of skills/.curated (or another path you specify) and cross-references with your locally installed skills.
  • install-skill-from-github.py — Downloads the skill directory from GitHub (using direct download for public repos, falling back to git sparse checkout for private repos or auth failures) and copies it into $CODEX_HOME/skills/.
Private GitHub repositories are supported via GITHUB_TOKEN or GH_TOKEN environment variables.

​
Next Steps

All Installation Methods

Learn every install option: by name, by folder, by URL, with private repo support and CLI flags.

Create a Custom Skill

Build your own skill for team-specific workflows using the skill-creator and init_skill.py.

Build docs developers (and LLMs) love

Get started for freeTalk to us