What Are Claude Code Skills? Complete Guide & Directory
Claude Code skills are reusable prompt templates that extend what Claude can do. Learn how to install, create, and use skills — plus browse 55+ free templates.
Claude Code skills are reusable, task-focused prompt templates that teach Claude how to perform specific workflows — like writing a PRD, running a code review, or creating an incident postmortem. You install a skill into your project, then invoke it with a slash command (/skill-name) or let Claude load it automatically when relevant. Skills follow the open Agent Skills specification, which works across 25+ AI tools including Cursor, Windsurf, and VS Code.
AI-Directory hosts 55+ free skill definitions you can install into any project.
How Do Claude Code Skills Work?
A skill is a directory containing a SKILL.md file with two parts: YAML frontmatter that tells Claude when to use the skill, and markdown content with the instructions Claude follows.
---
name: code-review
description: Reviews code for bugs, security issues, and style violations
---
When reviewing code:
1. Check for correctness first
2. Look for security vulnerabilities
3. Evaluate readability and maintainability
4. Suggest specific improvements with examples
When you invoke /code-review or ask Claude something that matches the description, Claude loads the full skill content and follows those instructions. The skill acts as a specialized playbook — turning Claude from a general-purpose assistant into a focused expert for that specific task.
Skills can also include supporting files: templates for Claude to fill in, example outputs showing the expected format, scripts Claude can execute, and reference documentation.
my-skill/
SKILL.md # Main instructions (required)
template.md # Template for Claude to fill in
examples/
sample.md # Example output
scripts/
validate.sh # Script Claude can execute
How Are Skills Different from Agents?
Skills and agents serve different purposes in Claude Code:
| Agents | Skills | |
|---|---|---|
| Shape | Role-based persona (who Claude becomes) | Task-based capability (what Claude does) |
| Example | "You are a senior code reviewer" | "Run this code review checklist" |
| Scope | Persistent across the session | Active only when invoked |
| Invocation | Loaded as system prompt / CLAUDE.md | Triggered via /skill-name or auto-detected |
| Composability | An agent can use multiple skills | A skill works independently |
An agent template defines who Claude is — a VP of Product, a Security Engineer, a Technical Writer — with decision-making frameworks and behavioral patterns. A skill defines what Claude does — write a PRD, run a security audit, create a test plan — with step-by-step instructions for a specific task.
You can use skills without agents, or combine them: load a "Senior Engineer" agent and invoke the /code-review skill for a focused review that carries the agent's judgment and standards.
How to Install Claude Code Skills
There are three ways to add skills to your Claude Code environment:
1. Install from AI-Directory
Browse the skill directory and use the install command shown on each skill's detail page:
npx skills add code-review
This downloads the skill files into your project's .claude/skills/ directory.
2. Create a Skill Manually
Create a directory and add a SKILL.md file:
mkdir -p .claude/skills/my-skill
Then write your SKILL.md with frontmatter and instructions.
3. Install from a Plugin
Plugins can bundle skills alongside other extensions. When you enable a plugin, its skills become available automatically with a plugin-name:skill-name namespace.
Where Skills Live
Where you store a skill determines who can use it:
| Location | Path | Applies to |
|---|---|---|
| Personal | ~/.claude/skills/<name>/SKILL.md | All your projects |
| Project | .claude/skills/<name>/SKILL.md | This project only |
| Plugin | <plugin>/skills/<name>/SKILL.md | Where plugin is enabled |
| Enterprise | Managed settings | All users in your organization |
Personal skills (in ~/.claude/skills/) are available across every project you work on. Project skills (in .claude/skills/) are committed to version control and shared with your team.
What Skill Frontmatter Options Are Available?
The YAML frontmatter at the top of SKILL.md controls how the skill behaves:
| Field | What It Does |
|---|---|
name | Display name and slash command. Defaults to directory name. |
description | Tells Claude when to use the skill. Front-load the key use case — descriptions over 250 chars are truncated. |
disable-model-invocation | Set true to prevent Claude from auto-loading. Use for skills with side effects like /deploy. |
user-invocable | Set false to hide from the / menu. Use for background knowledge skills. |
allowed-tools | Tools Claude can use without permission prompts when the skill is active. |
model | Override the model for this skill (e.g., force Opus 4.6 for complex tasks). |
context | Set to fork to run in an isolated subagent context. |
agent | Which subagent type to use with context: fork (e.g., Explore, Plan). |
paths | Glob patterns that limit auto-activation to specific file paths. |
What Skills Are Available on AI-Directory?
AI-Directory maintains a curated library of 55+ free skill definitions spanning engineering, product, design, and DevOps workflows. Here are some of the most popular:
Engineering:
- Code Review Checklist — structured review with security, performance, and readability checks
- API Design Guide — RESTful API design patterns and conventions
- Test Plan Writing — comprehensive test plans with coverage analysis
- Architecture Decision Record — document architectural decisions with context and consequences
Product:
- PRD Writing — product requirements documents with user stories and acceptance criteria
- Technical Spec Writing — technical specifications with system design and dependencies
- User Story Mapping — map user journeys into development tasks
Operations:
- Incident Postmortem — structured post-incident reviews with timeline and action items
- Data Migration Plan — migration planning with rollback strategies
- System Design Document — architecture documentation with diagrams and trade-offs
Browse the full skill directory to find templates for your workflow.
How to Create Your Own Claude Code Skill
Building a custom skill takes under 5 minutes. Here is a complete example — a skill that generates release notes from git history:
Step 1: Create the skill directory:
mkdir -p .claude/skills/release-notes
Step 2: Write SKILL.md:
---
name: release-notes
description: Generate release notes from recent git commits. Use when preparing a release, writing a changelog, or summarizing recent work.
allowed-tools: Bash(git *)
---
Generate release notes for the latest release:
1. Run `git log` to get commits since the last tag
2. Group commits by type (features, fixes, refactors, docs)
3. Write user-facing release notes in markdown
4. Include breaking changes prominently at the top
5. Link to relevant PRs or issues where available
Format as a changelog entry with the version number and date.
Step 3: Test it:
claude
/release-notes
Claude reads the git history, categorizes the commits, and outputs formatted release notes.
Can Claude Auto-Detect When to Use a Skill?
Yes. By default, Claude reads all skill descriptions at the start of each session. When your request matches a skill's description, Claude loads and follows that skill's instructions automatically — you do not need to type the slash command.
For example, if you have a code-review skill with description "Reviews code for bugs, security issues, and style violations" and you ask Claude "review this pull request," Claude recognizes the match and loads the skill.
To prevent auto-detection for skills that have side effects (like deploying or sending messages), add disable-model-invocation: true to the frontmatter. You can still invoke those skills manually with /skill-name.
Frequently Asked Questions
Do skills work in Cursor and other AI tools?
Skills follow the Agent Skills specification, an open standard adopted by 25+ AI tools. The core SKILL.md format works across Claude Code, Cursor, Windsurf, and other compatible tools. Some Claude Code-specific features like context: fork and allowed-tools are extensions that may not work in all tools.
Can I share skills with my team?
Yes. Commit the .claude/skills/ directory to your repository. Anyone who clones the project gets the skills automatically. For organization-wide distribution, use managed settings or plugins.
How many skills can I have?
There is no hard limit on the number of skills. However, skill descriptions are loaded into context at session start, which uses tokens. The description budget scales at 1% of the context window. If you have many skills, keep descriptions concise (under 250 characters) and use disable-model-invocation: true for skills that do not need auto-detection.
Last updated: March 2026. Browse the full AI skill directory on AI-Directory — 55+ free templates for Claude Code, Cursor, and other AI tools. See also: AI agent templates for role-based personas that complement skills.