Skill library · updated Nov 6, 2026

Codex Skills: install and use skills in OpenAI Codex

A Codex skill is a folder with a SKILL.md file that tells Codex how to do one job, and you install one by typing $skill-installer followed by its name or a GitHub folder URL. Below you get open-source skills with copy-ready commands, a SKILL.md template and fixes for skills that Codex does not pick up.

A Codex skill: a folder with SKILL.md, installed with $skill-installer

Codex skill library

Every OpenAI and Anthropic skill here is open source under Apache-2.0. Click Copy and paste the command into Codex. Skills from Answer the Web are tested in Claude Code and not yet tested in Codex.

Offpage AI Skill

Book backlinks, brand mentions and listicles on 170+ niche sites straight from your Claude chat. German-language publishing network.

Answer the WebTested in Claude CodeNot yet tested in Codex

Skill details (German) →

Google-Traffic-Box

The site-building and traffic method from 2,000+ projects as a skill: strategies, templates and quality checks for your AI.

Answer the WebTested in Claude CodeNot yet tested in Codex

Skill details (German) →

Pinterest-Traffic-Box

Pinterest as a traffic channel: pin templates, board structure and a posting plan, automated with your AI.

Answer the WebTested in Claude CodeNot yet tested in Codex

Skill details (German) →

Expired Domain Skill

Find, check and catch expired domains in the chat, with a drop-catching playbook from pending delete to catch.

Answer the WebTested in Claude CodeNot yet tested in Codex

Skill details (German) →

gh-address-comments

Reads the review comments on your open GitHub pull request with the gh CLI and works through them.

OpenAI curatedApache-2.0GitHub
$skill-installer gh-address-comments

gh-fix-ci

Inspects failing GitHub Actions checks on a pull request, summarizes the logs and proposes a fix.

OpenAI curatedApache-2.0GitHub
$skill-installer gh-fix-ci

playwright

Drives a real browser from the terminal: navigate, fill forms, take screenshots, extract data.

OpenAI curatedApache-2.0Testing
$skill-installer playwright

security-best-practices

Language-specific security review with secure-by-default guidance for the code you are writing.

OpenAI curatedApache-2.0Security
$skill-installer security-best-practices

jupyter-notebook

Creates and edits Jupyter notebooks for experiments, analyses and tutorials.

OpenAI curatedApache-2.0Data
$skill-installer jupyter-notebook

screenshot

Captures desktop or window screenshots when the task needs a visual check.

OpenAI curatedApache-2.0Testing
$skill-installer screenshot

linear

Reads and updates Linear issues and projects. Needs the Linear MCP server.

OpenAI curatedApache-2.0Workflow
$skill-installer linear

mcp-builder

Guides you through building a Model Context Protocol server, from tool design to evaluation.

AnthropicApache-2.0Build skills & MCPInstall via GitHub URL, not documented by Anthropic
$skill-installer https://github.com/anthropics/skills/tree/main/skills/mcp-builder

webapp-testing

Tests local web apps with Playwright: start the server, click through, capture logs and screenshots.

AnthropicApache-2.0TestingInstall via GitHub URL, not documented by Anthropic
$skill-installer https://github.com/anthropics/skills/tree/main/skills/webapp-testing

frontend-design

Builds distinctive, production-grade frontend interfaces instead of template-looking UI.

AnthropicApache-2.0DesignInstall via GitHub URL, not documented by Anthropic
$skill-installer https://github.com/anthropics/skills/tree/main/skills/frontend-design

skill-creator

Creates new skills and improves existing ones, including descriptions that trigger reliably.

AnthropicApache-2.0Build skills & MCPInstall via GitHub URL, not documented by Anthropic
$skill-installer https://github.com/anthropics/skills/tree/main/skills/skill-creator

algorithmic-art

Generative art with p5.js: seeded randomness, flow fields, particle systems.

AnthropicApache-2.0DesignInstall via GitHub URL, not documented by Anthropic
$skill-installer https://github.com/anthropics/skills/tree/main/skills/algorithmic-art

Sources: github.com/openai/skills (the repository now points to openai/plugins for new work) and github.com/anthropics/skills, checked 06.11.2026. The Anthropic document skills (docx, pdf, pptx, xlsx) are source-available, not open source, so they are not listed.

How to install a skill in Codex

Codex ships a built-in installer skill. Type the installer with the name of a curated skill, or with the URL of any GitHub folder that contains a SKILL.md. Codex downloads the folder and detects the new skill automatically; if it does not appear, restart Codex.

$skill-installer playwright

To install by hand, copy the skill folder into one of the folders Codex scans: .agents/skills inside your repository for a project skill, or ~/.agents/skills for every project on your machine.

mkdir -p ~/.agents/skills && cp -r ./my-skill ~/.agents/skills/

Use a skill by typing /skills and picking it, by mentioning it with $ and its name, or just by describing the task. Codex compares your request with each description and loads the matching skill.

How to write your own SKILL.md

A skill needs one file. The header between the two lines of three dashes holds the name and a description that says what the skill does and when to use it. Codex requires both fields. The body holds the steps. Keep it short and concrete; put long reference material into a references folder next to the file.

---
name: release-notes
description: Write release notes from merged pull requests. Use when the user asks for release notes, a changelog or a summary of recent merges.
license: Apache-2.0
---

# Release notes

1. List merged pull requests since the last tag with `gh pr list --state merged`.
2. Group them into features, fixes and maintenance.
3. Write one plain-language line per change and link the pull request.
4. Put breaking changes at the top.

Save it as .agents/skills/release-notes/SKILL.md. Optional folders: scripts for code the skill runs, references for documents it reads on demand, assets for templates, and agents/openai.yaml for Codex-specific settings.

Codex skills vs Claude Code skills

Both tools follow the same open format, so the differences are small but they matter when a skill does not load.

TopicCodexClaude Code
Project folder.agents/skills.claude/skills
User folder~/.agents/skills (installer: ~/.codex/skills)~/.claude/skills
Required headername and descriptionnone, all fields optional
Run a skill/skills, $name or by description/skill-name or by description
Turn off[[skills.config]] enabled = false in config.tomlremove or move the folder

For one skill that works in both, use only the standard fields name, description, license, compatibility, metadata and allowed-tools, and keep the folder name equal to the name field. Comparing the tools themselves? Read our Codex vs Claude Code comparison (German) or browse the Claude Skills hub.

Troubleshooting: skills not discovered

  1. Header fields missing. Codex skips a SKILL.md without name or description. Check that the header starts and ends with a line of three dashes.
  2. Wrong folder. The installer writes to ~/.codex/skills, while the docs list ~/.agents/skills for user skills. OpenAI support suggests moving skills to ~/.agents/skills when they do not show up.
  3. Broken config.toml. A syntax error in ~/.codex/config.toml can leave the skills picker empty. Fix the file, then restart Codex.
  4. Skill turned off. Look for a [[skills.config]] entry with enabled = false for that path.
  5. Too many skills. Codex reserves a small share of the context, about 2 percent, for skill descriptions. With many skills, shorten descriptions or turn off the ones you do not need.
  6. Old session. Start a new session. Users reported skills that only appeared after a restart.

Sources: OpenAI: build skills, OpenAI: Codex skills, Anthropic: Claude Code skills, openai/codex issue 15136, OpenAI community threads 1379522, 1381921 and 1374491, all checked 06.11.2026.

Codex skills: FAQ

What are Codex skills?

A Codex skill is a folder with a SKILL.md file. The file starts with a short header that holds a name and a description, followed by instructions. Codex reads the description, decides when the skill fits a task and loads the full instructions only then. Optional folders hold scripts, references and assets.

Where does Codex look for skills?

Codex scans .agents/skills from your current folder up to the repository root, then $HOME/.agents/skills for your user, /etc/codex/skills for admins, plus the built-in system skills. Symlinked folders are followed. The skill installer writes to $CODEX_HOME/skills, which is ~/.codex/skills by default.

How do I use a skill once it is installed?

Type /skills to pick one, mention it with $ followed by its name, or simply describe the task. Codex matches the task against each skill description and loads the right skill on its own. New skills are picked up automatically; if one does not show up, restart Codex.

Can I use Claude Code skills in Codex?

Usually yes. Both tools read the same SKILL.md format. Codex requires the name and description fields, while Claude Code treats every header field as optional. If you stick to the open standard fields (name, description, license, compatibility, metadata, allowed-tools), one skill folder works in both tools.

How do I turn a skill off without deleting it?

Add a [[skills.config]] entry with the path of the skill and enabled = false to ~/.codex/config.toml. The folder stays where it is, Codex just stops loading it. Remove the entry to turn it back on.

More from Answer the Web: home · Claude Code statusline builder · Offpage AI Skill · Google Traffic Box

Answer the Web is not affiliated with OpenAI or Anthropic.