Skills / Guides
What are Claude skills?
A skill is a folder with a SKILL.md file that teaches Claude how to do one job. Claude reads a one-line description of every skill you install and loads the full instructions only when a task calls for them.
What is inside a skill
The only required file is SKILL.md. It starts with a short block of settings (the frontmatter) and continues with plain Markdown instructions. A skill can also carry scripts, reference documents and templates in the same folder.
weekly-status/
├── SKILL.md required: settings + instructions
├── scripts/ optional: code the skill can run
├── references/ optional: longer docs, read when needed
└── assets/ optional: templates and other files---
name: weekly-status
description: Turns meeting notes into a weekly status email. Use when the user pastes notes and asks for an update or summary.
---
1. Group the notes by project.
2. For each project write: done, next, blocked.
3. Keep the email under 200 words.name becomes the command (/weekly-status) and must be lowercase letters, numbers and hyphens. description is the most important line in the file: it is what Claude reads to decide whether the skill fits your request. Try the SKILL.md generator to write one and check it against the spec.
How Claude decides to use a skill
Skills load in layers, which the open standard calls progressive disclosure. At the start of a session Claude only sees each skill's name and description, roughly a hundred tokens each. When your request matches a description, Claude loads that skill's full instructions, and it opens reference files or runs scripts only if the instructions ask for them. That is why you can install many skills without filling the context window, and why a vague description is the usual reason a skill never triggers.
You can also call a skill directly by typing a slash and its name. Two settings control who can start it:
disable-model-invocation: true: only you can start it. Good for anything with side effects, like deploys or commits. Example: grill-me.user-invocable: false: only Claude can use it; it is hidden from the slash menu. Good for background knowledge.
Where skills live in Claude Code
| Scope | Folder | Available in |
|---|---|---|
| Personal | ~/.claude/skills/<name>/SKILL.md | All your projects on this machine |
| Project | .claude/skills/<name>/SKILL.md | This repository; commit it to share with your team |
| Nested | <subdir>/.claude/skills/<name>/SKILL.md | Sessions working in that subfolder (useful in monorepos) |
| Enterprise | Managed settings directory | Every user on managed machines |
When two skills share a name, enterprise beats personal and personal beats project. Claude Code watches these folders, so a new or edited skill shows up without restarting. Skills you enable on Claude.ai are synced to ~/.claude/skills/synced/, so do not name one of your own skills synced.
How to install a skill
From a plugin marketplace (Claude Code)
Most popular skills ship as plugins. Add the marketplace once, then install the plugin:
/plugin marketplace add anthropics/skills
/plugin install example-skills@anthropic-agent-skillsWith the npx skills installer
Many repositories also support the cross-agent installer, which copies skill folders into your project:
npx skills add https://github.com/Leonxlnx/taste-skillBy hand
Copy the skill's folder into ~/.claude/skills/ (all projects) or .claude/skills/ (one project). The folder name should match the name in its SKILL.md.
On Claude.ai
Paid plans include Anthropic's own skills and let you upload custom ones from settings. See Anthropic's help article on using skills in Claude for the current steps.
Skills vs MCP servers vs plugins vs CLAUDE.md
| Thing | What it is | Loaded | Use it for |
|---|---|---|---|
| Skill | Instructions, optionally with scripts and references | Only when a task matches | Teaching Claude a repeatable way to do a job |
| MCP server | A connection to an outside system that exposes tools | Tools are always available once connected | Reading and writing data in other apps: databases, issue trackers, APIs |
| Plugin | A package of skills, agents, hooks and MCP settings | Whatever it contains, on its own schedule | Installing and updating a set of things in one step |
| CLAUDE.md | Project memory: facts and rules for one repository | Every session, always | Things Claude should always know, kept short |
A good rule: if Claude needs to know it every time, put it in CLAUDE.md. If it is a procedure Claude needs now and then, make it a skill. If Claude needs to reach another system, connect an MCP server. If you want to share any of these, wrap them in a plugin.
How to write your own skill
- Fastest: ask Claude to use Anthropic's skill-creator. It interviews you, writes the SKILL.md, and tests it with and without the skill.
- Quick and manual: fill in the SKILL.md generator, then save the file to
~/.claude/skills/<name>/SKILL.md. - Keep SKILL.md under about 500 lines and move long material into reference files that load on demand.
Skills worth starting with
- Superpowers: brainstorm, plan, test first and review before shipping.
- frontend-design or Impeccable: less generic UI.
- grill-me: get interviewed until your plan is clear.
- claude-mem: memory that carries over between sessions.
FAQ
Are Claude skills free?
Skills themselves are just files, and most public skills are open source. You need a Claude plan or API access to run Claude; on Claude.ai, skills are available on paid plans.
Do Claude skills work in Codex or Cursor?
Usually. Claude Code skills follow the open Agent Skills standard, which Codex and many other agents also read. Codex looks in .agents/skills rather than .claude/skills, and it ignores frontmatter fields that only Claude Code understands.
What is the difference between a skill and an MCP server?
A skill is instructions (plus optional scripts) that teach Claude how to do a task, loaded only when relevant. An MCP server connects Claude to an outside system, such as a database or an app's API, and exposes tools Claude can call.
How many skills can I install?
There is no hard limit, but every skill's name and description sits in context so Claude can choose between them. Claude Code caps each skill's listing text at 1,536 characters; install the skills you actually use.
Based on Anthropic's Claude Code skills documentation and the open Agent Skills specification (agentskills.io), checked on 2026-10-16. Also see the Codex skills guide.