A skill is a folder with a SKILL.md file that teaches a coding agent how to do one kind of job well. This page covers what goes in that folder, where it goes for each agent, and how to tell whether a skill is worth installing.
1 / What a skill is
One folder, one job.
Inside the folder is a Markdown file named SKILL.md. Its header has a name and a description of when the skill applies. The body is the instructions the agent follows once the skill is active: house style, a checklist, commands to run, files to read. The folder can also hold scripts and reference material the instructions point to.
Agents only read the name and description at startup. The full instructions load when a task matches, so you can keep many skills installed without filling the context window.
frontend-design/
├── SKILL.md ← name, description, instructions
├── reference/ ← optional notes the instructions cite
└── scripts/ ← optional helpers the agent can run
2 / Install one
Copy the folder into your project.
Every listing on Use This Skill has a one-line command that copies the skill folder into the right place. This one installs frontend-design for Claude Code:
Commit the folder so the skill travels with the repository and your team gets it too. For a skill you want in every project, copy it to the personal folder in your home directory instead.
3 / Where it goes
The folder for each agent.
Swap the directory in the install command to match your agent. These are the paths each one reads, from its own documentation.
Claude Code
Project
.claude/skills/<name>/SKILL.md
Personal
~/.claude/skills/<name>/SKILL.md
Type /name at the start of a message, or just describe the task and Claude loads the skill when its description matches.
Type / in Agent chat and search for the skill. Cursor also reads .agents/skills, .claude/skills and .codex/skills, so a skill installed for another agent works here too.
GitHub Copilot and VS Code read .agents/skills; most other agents that support the format are listed on agentskills.io.
4 / Use it
Describe the task, or call it by name.
Once the folder is in place, start a new session and ask for the work the way you normally would. If the request matches the skill's description, the agent loads it. To force it, call the skill by name: /frontend-design in Claude Code or Cursor, $frontend-design in Codex.
If a skill never seems to load, read its description. Agents decide from that one line, so a vague description means a skill that rarely triggers.
5 / Before you install
See what it changes.
Star counts tell you a skill is popular. They don't show the output. A proof runs one brief through Claude twice on the same model, once without the skill and once with it, and publishes both results as live pages next to each other.
The proof for high-end-visual-design is free to try. Drag the divider and judge the difference yourself.
Copy the skill's folder into the skills directory your agent reads (for Claude Code that is .claude/skills in your project), then describe the task as you normally would. The agent reads the skill's description, decides it applies, and loads the full instructions. Most agents also let you call a skill by name with a slash command.
What is a SKILL.md file?+
A Markdown file with a short header (a name and a description of when to use the skill) followed by instructions the agent follows when the skill is active. The folder around it can hold scripts, templates and reference files the instructions point to.
Where do skills go?+
Claude Code reads .claude/skills, Codex reads .agents/skills, Cursor reads .cursor/skills and Gemini CLI reads .gemini/skills, each inside your project. Each one also has a personal folder in your home directory for skills you want in every project.
Do skills work across agents?+
Yes. SKILL.md is an open format, so the same folder works in Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot and others. Cursor and Gemini CLI even read the other agents' folders, so you often only need to install a skill once.
How do I know a skill is any good?+
Run the same brief with and without it and compare. That is what a proof on Use This Skill is: one brief, two runs on the same model, both results published as live pages with the skill's exact version recorded.
Use This Skill