Claude Code Skills: Complete Guide [2026]
Everything you need to know about Claude Code skills — what they are, how to install them, the top 10 examples, and how to create your own custom skills for any workflow.
Quick Answer
Claude Code skills are reusable task instructions stored as SKILL.md files. Install them in ~/.claude/skills/ or your project's .claude/skills/ folder. Trigger any skill by typing /skill-name in Claude Code. Browse 50+ skills at mcpdirectory.app/skills.
What Are Claude Code Skills?
Claude Code skills are user-defined, reusable task instructions that extend Claude's behavior for specific workflows. Stored as SKILL.md files, they act like custom slash commands — type /commit and Claude follows the exact steps you defined for writing commit messages.
Unlike system prompts that reset every session, skills persist across all your projects. Unlike MCP servers that connect to external tools, skills define behaviors — the exact logic, format, and steps Claude should follow for a given task. Learn more about the difference in our complete AI skills explainer.
# Skill file structure
~/.claude/
skills/
commit/
SKILL.md ← Global skill (available in all projects)
review-pr/
SKILL.md
your-project/
.claude/
skills/
deploy/
SKILL.md ← Project-specific skillSkills use a simple frontmatter + markdown format. The name is the trigger, the description tells Claude when to suggest it, and the body contains the full instructions:
---
name: commit
description: >
Generate a semantic commit message for staged changes.
Use when the user wants to commit code or asks for a commit message.
---
# Commit Skill
Analyze the staged git diff and write a commit message following
conventional commits format: type(scope): description
Types: feat, fix, docs, style, refactor, test, chore
Rules:
- Subject line ≤72 chars
- Use imperative mood ("Add feature" not "Added feature")
- Include body if changes are non-trivialSkills vs System Prompts: Why Skills Win
| Aspect | Claude Code Skills | System Prompts |
|---|---|---|
| Reusability | ✅ Invoke with /skill-name anytime | ⚠️ Must re-type every session |
| Consistency | ✅ Same output every time | ⚠️ Varies based on phrasing |
| Shareability | ✅ Commit to git, share across team | ⚠️ Stored in chat history only |
| Maintenance | ✅ Single SKILL.md to update | ⚠️ Update every prompt manually |
| Discovery | ✅ Browsable in skills catalog | ⚠️ Private to each user |
For a deeper dive, read our AI Skills vs System Prompts comparison.
Top 10 Claude Code Skills in 2026
The most popular and useful Claude Code skills, ranked by developer adoption and productivity impact. Find all of these and more at the AI Skills marketplace.
Git Commit Writer
/commitAutomatically generates semantic commit messages from your staged changes using conventional commit format.
Use case: Save time on every commit with consistent, descriptive messages that follow team standards.
PR Reviewer
/review-prPerforms a comprehensive code review covering logic, security, performance, and style — before you merge.
Use case: Catch bugs and issues before human reviewers, speeding up the review cycle.
Test Runner
/test-runnerWrites and runs unit tests for selected functions, integrating with Jest, Vitest, or pytest.
Use case: Instantly generate test suites for untested code with full edge-case coverage.
Deploy Workflow
/deployGuides you through staging and production deployments with pre-flight checks and rollback plans.
Use case: Reduce deployment anxiety with automated checklists and environment verification.
Code Refactorer
/refactorAnalyzes and rewrites messy code for clarity, performance, and modern patterns — with explanations.
Use case: Clean up legacy code systematically without breaking existing functionality.
Debug Assistant
/debugWalks through bugs step-by-step: hypotheses, root cause analysis, and targeted fix suggestions.
Use case: Resolve tricky bugs faster by leveraging structured debugging methodology.
Doc Generator
/documentCreates JSDoc, TSDoc, or Python docstrings for functions, classes, and modules automatically.
Use case: Keep documentation up-to-date without the manual effort.
Security Auditor
/security-auditScans code for OWASP Top 10 vulnerabilities: SQL injection, XSS, insecure dependencies, and more.
Use case: Add a security review layer to your workflow before shipping to production.
DB Migrator
/migrateGenerates database migration scripts and data transformation logic for schema changes.
Use case: Handle complex migrations safely with auto-generated, reviewed SQL scripts.
Blog SEO Writer
/blog-seoCreates fully SEO-optimized blog posts with structured data, metadata, and internal linking.
Use case: Publish high-quality blog content at scale without sacrificing SEO quality.
How to Install Claude Code Skills
Installing skills takes less than 60 seconds. Choose global (available in all projects) or project-specific (committed to your repo).
Create the skills directory
mkdir -p ~/.claude/skills/commit
Create your SKILL.md file
echo "--- name: commit description: Generate commit messages --- # Commit Skill ..." > ~/.claude/skills/commit/SKILL.md
Verify Claude Code finds it
# Type /commit in Claude Code — it should auto-suggest the skill
Browse skills from the marketplace
# Visit mcpdirectory.app/skills to find pre-built skills to download
For more installation options including GitHub-hosted skills, see our Claude Code skills GitHub guide.
How to Create Your Own Claude Code Skill
Creating a skill is straightforward. Here's a complete example of a custom skill that runs a pre-commit security check:
--- name: security-check description: > Run a security audit before committing code. Use when the user wants to check for vulnerabilities or asks about security before pushing to production. --- # Security Check Skill ## What This Skill Does Performs a pre-commit security review covering: 1. Dependency vulnerabilities (npm audit / pip check) 2. Hardcoded secrets and API keys 3. SQL injection patterns 4. XSS vulnerabilities in frontend code 5. Insecure direct object references ## Steps 1. Scan staged files for hardcoded secrets using regex patterns 2. Check package.json/requirements.txt for known vulnerabilities 3. Review API endpoints for authentication gaps 4. Report findings with severity levels (Critical/High/Medium/Low) 5. Suggest specific fixes for each finding ## Output Format Provide a security report with: - Summary: X issues found (N critical, N high, N medium) - Per-issue: file:line, description, severity, fix suggestion
Once created, trigger it with /security-check in Claude Code. For best practices on writing skills, see our SKILL.md guide and how to write perfect SKILL.md files.
Frequently Asked Questions
What are Claude Code skills?
Claude Code skills are reusable task instructions stored as SKILL.md files that tell Claude how to perform specific tasks. When you type /skill-name in Claude Code, it loads the skill's instructions and executes the defined workflow. Skills are stored in ~/.claude/skills/ (global) or .claude/skills/ (project-specific).
How do I install skills in Claude Code?
To install a skill globally, create a directory in ~/.claude/skills/[skill-name]/ and add a SKILL.md file inside it. For project-specific skills, use .claude/skills/[skill-name]/SKILL.md in your repository root. Claude Code automatically discovers and loads skills from both locations.
Where can I find Claude Code skills examples?
The best places to find Claude Code skills examples are: mcpdirectory.app/skills (curated catalog), GitHub (search 'SKILL.md claude code'), and the OpenClaw community ecosystem. Many developers publish their skill collections publicly on GitHub.
How do I create a custom Claude Code skill?
Create a folder at .claude/skills/my-skill/ and write a SKILL.md file with a name, description, and instructions. The description tells Claude when to suggest the skill. The instructions define the exact steps Claude should follow. Use the /skill-name trigger in your chat to invoke it.
What is the difference between Claude Code skills and MCP?
Claude Code skills (SKILL.md) are reusable task instructions that shape how Claude responds to specific prompts — they define workflows and behaviors. MCP (Model Context Protocol) servers provide external tool access: file systems, databases, APIs. Skills are behavioral; MCP is connectivity. They work best together.
Find Your Next Favorite Claude Code Skill
Browse 50+ verified Claude Code skills in our directory, or create your own custom skill in minutes with the skill creator tool.