Skill Creator

Skill Creator

This skill provides guidance for creating effective skills that extend Claude's capabilities.

About Skills

Skills are self-contained packages providing specialized knowledge, workflows, and tools. They transform Claude from general-purpose to domain-specific agent.

Core Principles

1. Concise is Key: Context window is shared resource. Only add what Claude doesn't already know.

2. Set Appropriate Freedom:

• High freedom: Multiple valid approaches, context-dependent decisions
• Medium freedom: Preferred patterns with some variation
• Low freedom: Fragile operations requiring specific sequences

Skill Structure

skill-name/
├── SKILL.md (required)
│   ├── YAML frontmatter (name, description)
│   └── Markdown instructions
└── Optional Resources
    ├── scripts/      - Executable code
    ├── references/   - Documentation for context
    └── assets/       - Output templates/files
        └── templates/ - Template files (for skill-creator only)

Creation Workflow

1. Initialize Skill

# Interactive mode (recommended)
uv run scripts/init_skill.py my-skill-name

# Non-interactive mode
uv run scripts/init_skill.py my-skill-name --non-interactive

# Custom path
uv run scripts/init_skill.py my-skill-name --path ./custom/location

2. Edit SKILL.md

Frontmatter Requirements:

• name: skill-name (hyphen-case, lowercase)
• description: What it does + when to use it (this triggers skill selection)

Body Guidelines:

• Keep under 500 lines
• Use imperative form
• Include essential workflows only
• Reference bundled resources clearly

3. Add Resources (Optional)

Scripts (scripts/): Executable code for deterministic tasks

Each generated skill includes dependency management:

• pyproject.toml - Defines required packages
• scripts/check_dependencies.py - Checks and installs dependencies

#!/usr/bin/env python3
def process_file(input_path):
    # Implementation here
    pass

Dependency Management Workflow:

# First time or when dependencies change
uv run scripts/check_dependencies.py --install

# Before using any scripts
uv run scripts/check_dependencies.py

# Run scripts
uv run scripts/helper.py

References (references/): Detailed documentation loaded on-demand

# API Reference
## Endpoints
- GET /users - List users
- POST /users - Create user

Assets (assets/): Templates, images, boilerplate for output

• PowerPoint templates
• HTML boilerplate directories
• Configuration files
• templates/ (skill-creator only) - Template files for skill generation:
  • skill.md.template - Main SKILL.md template
  • helper_script.py.template - Python script template
  • reference.md.template - Reference documentation template
  • asset_readme.md.template - Asset directory template
  • pyproject.toml.template - Python project configuration
  • check_dependencies.py.template - Dependency management script

4. Package Skill

# Package to current directory
uv run scripts/package_skill.py path/to/my-skill

# Package to specific output directory
uv run scripts/package_skill.py path/to/my-skill ./dist

Best Practices

Progressive Disclosure

1. Metadata (name/description) - Always loaded (~100 words)
2. SKILL.md body - When skill triggers (<5k words)
3. Bundled resources - As needed by Claude

Content Organization

• Keep core workflow in SKILL.md
• Move detailed examples to references/
• Split by domain for multi-domain skills:

  bigquery-skill/
  ├── SKILL.md (overview + navigation)
  └── references/
      ├── finance.md
      ├── sales.md
      └── product.md
  

What NOT to Include

• README.md files
• Installation guides
• Auxiliary documentation
• Only include files Claude needs to do the job

Note:

• Python skills: Dependency management (pyproject.toml + check_dependencies.py) is automatically included for seamless setup
• Other language skills: Include appropriate config files (package.json, Gemfile, go.mod, etc.) and setup instructions in SKILL.md to guide Claude on dependency management

Using Skills with Dependencies

When Claude encounters a skill with scripts in any programming language:

General Dependency Management Principles

Always check for dependency files first:

• pyproject.toml (Python with uv/pip)
• package.json (Node.js/JavaScript)
• Gemfile (Ruby)
• go.mod (Go)
• Cargo.toml (Rust)
• requirements.txt (Python legacy)
• composer.json (PHP)
• pom.xml/build.gradle (Java/JVM)

Common dependency setup patterns:

# Python (modern)
uv run scripts/check_dependencies.py --install
uv run scripts/script.py

# Node.js
npm install  # or yarn install, pnpm install
npm run script

# Ruby
bundle install
bundle exec ruby script.rb

# Go
go mod tidy
go run script.go

# Rust
cargo build
cargo run --bin script

1. Python Skills (Automated)

# Always run this before using Python skill scripts
uv run scripts/check_dependencies.py --install

2. Other Language Skills (Manual Check)

For non-Python skills, check the skill directory for:

1. Look for dependency files (package.json, Gemfile, go.mod, etc.)
2. Check SKILL.md for setup instructions
3. Run standard dependency commands for that language
4. Follow any custom setup scripts provided

3. Standard Skill Usage Pattern

# Step 1: Navigate to skill directory
cd path/to/skill

# Step 2: Check for and install dependencies
# (Language-specific command based on what files are present)

# Step 3: Use skill scripts/tools
# (Follow instructions in SKILL.md)

3. Troubleshooting Dependencies

If dependency issues occur:

# Clean and reinstall
rm -rf .venv
uv sync --reinstall

# Check specific package versions
uv tree

Validation

# Validate skill structure
uv run scripts/quick_validate.py path/to/skill

Validates:

• YAML frontmatter format
• Required fields (name, description)
• Naming conventions
• File structure

Common Patterns

Task-Based: Different operations

## Quick Start
## Merge PDFs  
## Split PDFs
## Extract Text

Workflow-Based: Sequential processes

## Overview
## Step 1: Preparation
## Step 2: Processing
## Step 3: Output

Reference-Based: Standards/specifications

## Guidelines
## Color Specifications
## Typography Rules
## Usage Examples

Examples

See existing skills for patterns:

• python: Development workflow with uv
• slidev: Presentation creation
• code-styleguide: Coding standards

Start simple, iterate based on actual usage.