Skill Seekers Makeup

This skill automates the creation of Claude skills from various content sources using the Skill_Seekers tool.

When to Use

Use this skill when:

•User wants to create a new skill from documentation websites
•User wants to create a new skill from GitHub repositories
•User wants to create a new skill from PDF documents
•User wants to create a new skill combining multiple sources (docs + GitHub + PDF)
•User wants to update an existing skill with new content

Quick Start

Installation Location

By default, generated skills are installed to the project root's .iflow/skills/ directory. To install skills to the global ~/.iflow/skills/ directory, use the --global flag.

Basic Usage

Use the scripts/generate_skill.py script to generate skills:

bash

# Generate skill from documentation website
python scripts/generate_skill.py \
  --name react \
  --description "React framework documentation" \
  --urls https://react.dev

# Generate skill from GitHub repository
python scripts/generate_skill.py \
  --name typescript \
  --description "TypeScript programming language" \
  --repos microsoft/TypeScript

# Generate skill from PDF documents
python scripts/generate_skill.py \
  --name manual \
  --description "Product manual" \
  --pdfs docs/manual.pdf

# Generate skill from multiple sources
python scripts/generate_skill.py \
  --name my-framework \
  --description "Complete framework knowledge" \
  --urls https://docs.myframework.com \
  --repos owner/myframework \
  --pdfs docs/advanced.pdf

# Generate skill with smart path-aware scraping
python scripts/generate_skill.py \
  --name iflow-cli \
  --description "iFlow CLI documentation" \
  --urls https://platform.iflow.cn/cli/quickstart \
  --smart-scrape

Updating an Existing Skill

To update a skill with new content:

bash

python scripts/generate_skill.py \
  --name existing-skill \
  --update \
  --urls https://new-docs.com \
  --repos owner/new-repo

Parameters

Required Parameters

•--name <name>: Name of the skill to generate (used for installation path)
•--description <description>: Description of what the skill does and when to trigger it

Content Sources (at least one required)

•--urls <url1,url2,...>: Comma-separated list of documentation website URLs
•--repos <repo1,repo2,...>: Comma-separated list of GitHub repositories (format: owner/repo)
•--pdfs <pdf1,pdf2,...>: Comma-separated list of PDF file paths

Optional Parameters

•--update: Update an existing skill instead of creating a new one
•--async: Enable async scraping mode for faster processing (2-3x speedup)
•--workers <n>: Number of async workers (default: 4)
•--output-dir <path>: Output directory for generated skill (default: output/)
•--skip-enhance: Skip AI enhancement step (not recommended)
•--skip-package: Skip packaging step (keeps raw output)
•--skip-install: Skip installation to .iflow/skills/
•--global: Install skill to global ~/.iflow/skills/ directory (default: install to project root's .iflow/skills/)
•--verbose: Enable verbose logging
•--smart-scrape: Use smart path-aware crawler (limits to same path level and max 3 hops)

Workflow

The skill executes the following steps automatically:

Phase 1: Scrape Content

For each content source:

Documentation Websites:

•Uses skill-seekers scrape command (default) or smart scraper (with --smart-scrape)
•
Smart scraper mode: Limits crawling to same path level and max 3 hops from entry URL
- •Example: Entry https://www.example.com/docs/api
- •✅ https://www.example.com/docs/changelog (same level)
- •✅ https://www.example.com/docs/api/abc (deeper level)
- •✅ https://www.example.com/docs/protocol/def (same level's subpath)
- •❌ https://www.anotherexample.com (different host)
- •❌ https://www.example.com/ (higher level)
- •❌ https://www.example.com/blogs (different path)
•Detects and uses llms.txt if available (10x faster)
•Scrapes all pages from the documentation
•Categorizes content by topic
•Supports async mode with --async flag

GitHub Repositories:

•Uses skill-seekers github command
•Extracts README, code structure, and signatures
•Fetches GitHub Issues and PRs
•Extracts CHANGELOG and Releases
•Performs deep code analysis with AST parsing

PDF Documents:

•Uses skill-seekers pdf command
•Extracts text, code, and images
•Supports OCR for scanned PDFs
•Handles password-protected PDFs
•Extracts tables from complex layouts

Phase 2: AI Enhancement

•Uses skill-seekers enhance command
•Enhances SKILL.md with comprehensive examples
•Extracts best practices and key concepts
•Improves documentation quality
•Note: Enhancement is mandatory for production-ready skills

Phase 3: Package Skill

•Uses skill-seekers package command
•Creates Claude-ready .skill package
•Validates skill structure
•Ensures quality standards
•Creates distributable skill file

Phase 4: Install Skill

•Installs to project root's .iflow/skills/<skill-name>/ by default
•Installs to global ~/.iflow/skills/<skill-name>/ when --global flag is used
•Preserves existing skill content when updating
•Maintains proper directory structure
•Ready for immediate use

Examples

Example 1: Create React Skill

bash

python scripts/generate_skill.py \
  --name react \
  --description "React framework for building user interfaces. Use when working with React components, hooks, state management, or React-specific patterns." \
  --urls https://react.dev \
  --async \
  --workers 8

Example 2: Create Django Skill with GitHub

bash

python scripts/generate_skill.py \
  --name django \
  --description "Django web framework. Use when building Django applications, working with models, views, templates, or Django ORM." \
  --urls https://docs.djangoproject.com \
  --repos django/django \
  --async

Example 3: Create PDF-Based Skill

bash

python scripts/generate_skill.py \
  --name api-reference \
  --description "REST API reference documentation. Use when working with API endpoints, authentication, or data models." \
  --pdfs docs/api-reference.pdf

Example 4: Multi-Source Skill

bash

python scripts/generate_skill.py \
  --name godot \
  --description "Godot game engine. Use when developing games with Godot, working with GDScript, scenes, nodes, or Godot-specific features." \
  --urls https://docs.godotengine.org \
  --repos godotengine/godot \
  --pdfs docs/godot-manual.pdf \
  --async \
  --workers 8

Example 5: Update Existing Skill

bash

python scripts/generate_skill.py \
  --name react \
  --update \
  --urls https://react.dev,https://react-redux.js.org \
  --repos facebook/react

Example 6: Install Skill to Global Directory

bash

python scripts/generate_skill.py \
  --name react \
  --description "React framework for building user interfaces" \
  --urls https://react.dev \
  --global

Advanced Usage

Using Custom Config Files

For advanced configurations, create a custom Skill_Seekers config file:

bash

# Create custom config
cat > my-config.json << EOF
{
  "name": "my-framework",
  "description": "My custom framework",
  "base_url": "https://docs.myframework.com",
  "max_pages": 500,
  "async": true,
  "workers": 8
}
EOF

# Use custom config
python scripts/generate_skill.py \
  --name my-framework \
  --description "My custom framework" \
  --config my-config.json

Performance Optimization

For large documentation sets (10K+ pages):

bash

python scripts/generate_skill.py \
  --name large-docs \
  --description "Large documentation set" \
  --urls https://docs.example.com \
  --async \
  --workers 16 \
  --verbose

Batch Processing

Generate multiple skills at once:

bash

# Create a batch script
cat > batch-generate.sh << EOF
#!/bin/bash

python scripts/generate_skill.py \
  --name react \
  --description "React framework" \
  --urls https://react.dev

python scripts/generate_skill.py \
  --name vue \
  --description "Vue.js framework" \
  --urls https://vuejs.org

python scripts/generate_skill.py \
  --name angular \
  --description "Angular framework" \
  --urls https://angular.io
EOF

# Run batch
chmod +x batch-generate.sh
./batch-generate.sh

Troubleshooting

Skill_Seekers Not Found

If you get "command not found: skill-seekers":

bash

# Install Skill_Seekers
pip install skill-seekers

# Or install from source
cd /path/to/Skill_Seekers
pip install -e .

GitHub Rate Limiting

If you hit GitHub rate limits:

bash

# Set GitHub token
export GITHUB_TOKEN=ghp_your_token_here

# Retry the command
python scripts/generate_skill.py --name my-skill --repos owner/repo

Async Mode Issues

If async mode fails:

bash

# Try without async mode
python scripts/generate_skill.py --name my-skill --urls https://docs.com --no-async

Enhancement Fails

If AI enhancement fails:

bash

# Check Claude Code Max is available
claude --version

# Or skip enhancement (not recommended)
python scripts/generate_skill.py --name my-skill --urls https://docs.com --skip-enhance

Best Practices

•Always provide a clear description that includes when to trigger the skill
•Use async mode for large documentation to speed up processing
•Combine multiple sources when possible for comprehensive coverage
•Test generated skills before deployment
•Keep skill names descriptive and unique
•Update skills regularly when documentation changes
•Use specific URLs rather than root URLs when possible

Output Structure

Generated skills follow this structure:

Default (project root):

code

<project-root>/.iflow/skills/<skill-name>/
├── SKILL.md              # Main skill file with metadata and instructions
├── api/                  # API documentation (if applicable)
├── guides/               # User guides and tutorials
├── examples/             # Code examples
├── reference/            # Reference documentation
└── assets/               # Images, diagrams, etc.

Global (with --global flag):

code

~/.iflow/skills/<skill-name>/
├── SKILL.md              # Main skill file with metadata and instructions
├── api/                  # API documentation (if applicable)
├── guides/               # User guides and tutorials
├── examples/             # Code examples
├── reference/            # Reference documentation
└── assets/               # Images, diagrams, etc.

References

For detailed Skill_Seekers documentation, see: