Awesome Claude Skills - Curated collection of Claude AI customizations

Repository: GitHub (travisvn/awesome-claude-skills)
Stars: ~12,800
License: MIT
Owner: travisvn
Repo: https://github.com/travisvn/awesome-claude-skills

Core Components:
- SKILL.md files with YAML frontmatter - Skill definitions and instructions
- Markdown - Documentation and instruction format
- Git - Version control and distribution
- Various scripting languages - Python, JavaScript, Bash for skill logic
- Claude AI API - Integration endpoint

Supported Platforms:
- Claude.ai (web interface)
- Claude Code (desktop/CLI)
- Claude API (programmatic access)

Architecture:
- Progressive disclosure loading (~100 tokens metadata → <5k tokens instructions → on-demand resources)
- Composable skills that automatically stack
- Git-based distribution and version control

Required:
- Claude Pro, Max, Team, or Enterprise subscription
- Access to Settings > Capabilities (for Claude.ai)
- For Team/Enterprise: Admin authorization

Optional (for advanced usage):
- Git (for cloning and managing custom skills)
- Claude Code CLI (for programmatic skill management)
- Development environment (Python, Node.js, etc.) for creating custom skills

skill-name/
├── SKILL.md          # Main skill file with YAML frontmatter
├── scripts/          # Optional: Executable code (Python, JS, etc.)
├── resources/        # Optional: Supporting files, templates, data
└── README.md         # Optional: Human-readable documentation

SKILL.md structure:
---
name: skill-name
description: Brief description of what the skill does
---

# Detailed Instructions

[Markdown content with instructions for Claude]

## Usage Examples

[Examples showing how to use the skill]

1. Navigate to https://claude.ai
2. Click your profile icon → Settings
3. Go to Capabilities section
4. Enable the "Skills" toggle
5. Click "Browse Skills" to see official skills
6. Click on a skill to view details and add it to your workspace
7. Alternatively, click "Upload Skill" to add a custom skill from your computer

# Install from marketplace (official skills)
/plugin marketplace add anthropics/skills

# Add a local skill directory
/plugin add /path/to/skill-directory

# Add a skill from GitHub URL
/plugin add https://github.com/username/skill-repo

# List installed skills
/plugin list

# Remove a skill
/plugin remove skill-name

# Update all skills
/plugin update

# List available skills
curl https://api.anthropic.com/v1/skills \
  -H "x-api-key: $ANTHROPIC_API_KEY" \
  -H "anthropic-version: 2023-06-01"

# Enable a skill for your API requests by including it in the request
curl https://api.anthropic.com/v1/messages \
  -H "x-api-key: $ANTHROPIC_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "content-type: application/json" \
  -d '{
    "model": "claude-opus-4-7",
    "skills": ["skill-name"],
    "messages": [
      {"role": "user", "content": "Your prompt here"}
    ]
  }'

Document Skills:
- docx: Create and edit Word documents
- pdf: Generate and analyze PDFs
- pptx: Build PowerPoint presentations
- xlsx: Work with Excel spreadsheets

Design & Creative:
- algorithmic-art: Generate algorithmic and generative art
- canvas-design: Create visual designs and layouts
- slack-gif-creator: Generate custom GIFs for Slack

Development:
- frontend-design: Build frontend interfaces
- web-artifacts-builder: Create web components and artifacts
- mcp-builder: Build Model Context Protocol servers
- webapp-testing: Automated web application testing

Communication:
- brand-guidelines: Maintain brand consistency
- internal-comms: Write internal communications

Skill Creation:
- skill-creator: Interactive tool for building new skills

Popular Collections:
- obra/superpowers: 20+ battle-tested productivity skills
- Trail of Bits: Security-focused skills for penetration testing
- iOS Simulator: Automation for iOS development
- Web Fuzzing: ffuf integration for web security testing
- Playwright Browser Automation: End-to-end testing
- D3.js Visualizations: Data visualization skills
- Scientific Computing: NumPy, SciPy, pandas integrations
- Web Asset Generation: Automated asset creation

Explore more at: https://github.com/travisvn/awesome-claude-skills

# Automatic activation (Claude detects relevance)
"Can you create a PDF report of this data?"
→ Claude loads the 'pdf' skill automatically

# Explicit invocation (mention the skill)
"Using the algorithmic-art skill, generate a fractal pattern."

# Multiple skills can compose automatically
"Analyze this Excel file and create a PowerPoint presentation."
→ Claude loads both 'xlsx' and 'pptx' skills

# In Claude Code, skills work with your codebase
"Using the webapp-testing skill, write tests for this component."

# 1. Create the skill directory structure
mkdir -p my-skill/scripts my-skill/resources

# 2. Create SKILL.md with frontmatter
cat > my-skill/SKILL.md << 'EOF'
---
name: my-custom-skill
description: Brief description of what this skill does
---

# My Custom Skill

Detailed instructions for Claude on how to use this skill.

## When to Use

Describe scenarios where this skill should activate.

## How to Use

Step-by-step instructions for Claude to follow.

## Examples

Show example usage patterns.
EOF

# 3. Add any scripts or resources
touch my-skill/scripts/helper.py
touch my-skill/resources/template.txt

# 4. Test locally before sharing
# Load the skill in Claude Code or Claude.ai

1. Enable the 'skill-creator' skill in Claude
2. Describe what you want your skill to do:
   "I want to create a skill that helps me write SQL queries"
3. Claude will guide you through:
   - Defining the skill name and description
   - Writing instructions and usage patterns
   - Creating example prompts
   - Structuring the SKILL.md file
4. Review and refine the generated skill
5. Export and install the new skill

Discovery:
- GitHub: https://github.com/travisvn/awesome-claude-skills
- Official site: https://awesomeclaude.ai/
- Claude.ai marketplace: Settings > Capabilities > Browse Skills

Management:
- View active skills: Check your workspace settings
- Disable temporarily: Toggle individual skills off
- Update skills: Pull latest from Git or re-download
- Remove skills: Delete from your skills directory or use /plugin remove

Version Control:
- Track custom skills in Git
- Fork community skills to customize
- Submit PRs to improve existing skills
- Star repositories to track updates

Submission Guidelines:
1. Skills must be community-tested (have GitHub stars)
2. No purely commercial/marketing submissions
3. Skills should be generalizable, not narrow-focused
4. Include clear documentation and examples
5. Follow the standard skill structure

Contribution Process:
1. Fork travisvn/awesome-claude-skills
2. Add your skill to the appropriate category
3. Update README.md with a link and description
4. Submit a Pull Request
5. Respond to reviewer feedback

Quality Standards:
- Clear SKILL.md with comprehensive instructions
- Working examples demonstrating usage
- Minimal dependencies when possible
- Security considerations documented
- MIT or compatible open-source license

Three-Tier Loading System:

1. Metadata Scan (~100 tokens per skill)
   - Name and description only
   - Claude scans all available skills
   - Identifies potentially relevant matches
   - Minimal context window usage

2. Full Instructions (<5k tokens)
   - Loaded only when skill is deemed relevant
   - Contains complete usage instructions
   - Examples and detailed guidance
   - Still efficient for context window

3. Bundled Resources (on-demand)
   - Scripts and files loaded only when executed
   - Large templates or data files
   - External dependencies
   - Maximum efficiency

Benefit: 100+ skills available without overwhelming the context window

Before Installing:
✓ Review the SKILL.md and all scripts
✓ Check the source repository's reputation
✓ Look for GitHub stars and community feedback
✓ Verify the license and terms
✓ Examine any network requests or file access

Trusted Sources:
✓ Official Anthropic skills
✓ Well-known community contributors (obra, Trail of Bits, etc.)
✓ Skills with high star counts and active maintenance
✓ Open-source skills with visible code

Red Flags:
✗ Obfuscated code
✗ Requests for API keys or credentials without clear need
✗ Excessive file system access
✗ Network requests to unknown domains
✗ New repositories with no community validation

Sandboxing:
- Use Claude Code's isolation features when available
- Test skills in isolated environments first
- Monitor skill behavior during initial uses
- Report suspicious skills to the repository maintainers

Skill Not Loading:
- Verify SKILL.md has valid YAML frontmatter
- Check that 'name' and 'description' fields exist
- Ensure the skill is enabled in Settings > Capabilities
- Try restarting Claude or reloading the workspace

Skill Not Activating:
- Make the skill name explicit in your prompt
- Check if the skill's activation criteria match your task
- Review the skill's "When to Use" documentation
- Verify the skill doesn't conflict with other active skills

Script Execution Errors:
- Check script file permissions (chmod +x for Unix scripts)
- Verify required dependencies are installed
- Review script paths in SKILL.md (use relative paths)
- Check Claude Code's execution policy settings

API Integration Issues:
- Verify your API key has skills access
- Check the 'skills' parameter syntax in API requests
- Ensure you're using a compatible Claude model
- Review API version compatibility

Performance Problems:
- Disable unused skills to reduce metadata scanning
- Keep skill instructions concise (<5k tokens)
- Move large files to on-demand resources
- Use progressive disclosure patterns properly

Skill Composition:
- Multiple skills automatically compose when relevant
- Explicitly chain skills: "Using X and Y skills, do Z"
- Create meta-skills that orchestrate other skills

Workspace-Specific Skills:
- Create project-specific skill directories
- Version control skills alongside your project
- Share team skills via Git repositories
- Override global skills with local versions

Parameterized Skills:
- Use placeholders in SKILL.md for customization
- Create skill variants for different contexts
- Environment-based skill configuration

Skill Development Workflow:
1. Draft skill instructions in natural language
2. Test with Claude interactively
3. Refine based on Claude's interpretation
4. Formalize into SKILL.md structure
5. Add scripts and resources as needed
6. Version control and iterate
7. Share with team or community

Integration Patterns:
- API-based skills that call external services
- MCP (Model Context Protocol) server integration
- File format converters and processors
- Testing and validation frameworks
- Documentation generators

Official Resources:
- Main Repository: https://github.com/travisvn/awesome-claude-skills
- Contributing Guide: https://github.com/travisvn/awesome-claude-skills/blob/main/CONTRIBUTING.md
- Awesome Claude Site: https://awesomeclaude.ai/
- Anthropic Documentation: https://docs.anthropic.com/
- Claude API Docs: https://docs.anthropic.com/en/api/

Community Collections:
- obra/superpowers: https://github.com/obra/superpowers
- Trail of Bits Security Skills: https://github.com/trailofbits/

Support:
- GitHub Issues: https://github.com/travisvn/awesome-claude-skills/issues
- Discord: Check repository README for invite link
- Claude Help Center: https://support.anthropic.com/

Related Tools:
- Model Context Protocol: https://modelcontextprotocol.io/
- Claude Code Documentation: https://docs.anthropic.com/claude/code
- Anthropic Console: https://console.anthropic.com/