Claude Code Skills: Install, Write and Use Them (2026)

If you keep pasting the same instructions into Claude Code (how to write tests in this repo, how to cut a release, how to review a diff), make them a skill: write it once, and Claude Code (and anyone who clones the repo) does it the same way every time.

Where Claude Code looks for skills

Scope	Path	Who gets it
Personal	~/.claude/skills/<skill-name>/SKILL.md	You, in every project
Project	.claude/skills/<skill-name>/SKILL.md	Everyone who clones the repo
Nested	<subdir>/.claude/skills/<skill-name>/SKILL.md	Sessions started in that folder
Plugin	<plugin>/skills/<skill-name>/SKILL.md	Run as /plugin-name:skill-name
Enterprise	.claude/skills/ in the managed settings directory	The whole organisation

When two skills share a name, enterprise beats personal and personal beats project; plugin skills are namespaced so they never clash (Claude Code docs). Skills you’ve enabled in your claude.ai account also sync into ~/.claude/skills/synced/.

Install Anthropic’s official skills

Anthropic’s skills repo doubles as a plugin marketplace. In Claude Code:

/plugin marketplace add anthropics/skills
/plugin install document-skills@anthropic-agent-skills
/plugin install example-skills@anthropic-agent-skills

document-skills handles Word, Excel, PowerPoint and PDF files (these aren’t built into Claude Code, unlike the Claude apps). example-skills covers design, web-app testing, MCP server building and more. For a skill that writes skills for you, add Anthropic’s official plugin marketplace:

/plugin marketplace add anthropics/claude-plugins-official
/plugin install skill-creator@claude-plugins-official

Use /skills to see what’s installed and /reload-skills after editing one.

For community skills, skills.sh installs into Claude Code and other agents with npx skills add <owner/repo>.

Write your own

Create ~/.claude/skills/review-diff/SKILL.md. This example is from Anthropic’s docs:

---
description: Summarizes uncommitted changes and flags anything risky. Use when the user asks what changed, wants a commit message, or asks to review their diff.
---

## Current changes

!`git diff HEAD`

## Instructions

Summarize the changes above in two or three bullet points, then list any risks you notice such as missing error handling, hardcoded values, or tests that need updating. If the diff is empty, say there are no uncommitted changes.

Two Claude Code extras show up there. Every frontmatter field is optional (the name defaults to the folder name, though a description is strongly recommended), and a line starting with !`command` runs before Claude reads the skill, so the skill arrives with live context such as the current diff.

Other useful frontmatter fields in Claude Code:

Field	What it does
disable-model-invocation: true	Only runs when you type /name
user-invocable: false	Hidden from the / menu; only Claude can use it
argument-hint, arguments	Describe the arguments for /name arg
allowed-tools / disallowed-tools	Limit which tools the skill can use
model, effort	Run the skill on a specific model or effort level
context: fork	Run it in a separate sub-agent context

These extras only work in Claude Code. Uploads to the Claude app and the API accept just name, description, license, compatibility, metadata and allowed-tools, so keep shared skills to those if you want them portable.

How skills get used

• Automatically: Claude reads every skill’s description at startup and loads the full skill when your request matches.
• By command: /review-diff, or with arguments, /fix-issue 123.

If a skill never fires, its description is too vague. Name the situations and phrases that should trigger it.

Skills in other coding agents

The format is the open Agent Skills standard, so a good skill travels:

• Cursor reads .cursor/skills/ and .agents/skills/, and also .claude/skills/ for compatibility (Cursor docs).
• Codex reads .agents/skills/.
• GitHub Copilot reads .github/skills, .claude/skills or .agents/skills (GitHub docs).

Safety

Skills in Claude Code run with the same access as any program on your machine, including the network. Read every file in a skill, especially scripts and anything that fetches URLs, before installing it. Open registries have had real problems: a February 2026 scan by Snyk found at least one security flaw in 36.82% of 3,984 skills from ClawHub and skills.sh, and 76 confirmed malicious payloads (Snyk).