Agent Skills Generator

Agent Skills Generator is a high-performance CLI tool that crawls documentation websites and generates structured SKILL.md files optimized for AI agents. These skill files enable LLMs to access up-to-date documentation directly, enhancing their capabilities with domain-specific knowledge.

• Blazing Fast — Built in Rust for maximum performance with async crawling
• Smart Content Extraction — Automatically removes navigation, ads, and noise
• Token Optimized — Clean markdown output minimizes token usage
• Configurable — Fine-grained control over crawling behavior with YAML config
• Polite Crawling — Respects robots.txt and implements request delays
• Cross-Platform — Works on Linux, macOS, and Windows

Linux / macOS:

curl -fsSL https://raw.githubusercontent.com/AmanSikarwar/agent-skills-generator/master/install.sh | bash

Windows (PowerShell):

iwr -useb https://raw.githubusercontent.com/AmanSikarwar/agent-skills-generator/master/install.ps1 | iex

If you have Rust installed:

cargo install --git https://github.com/AmanSikarwar/agent-skills-generator

git clone https://github.com/AmanSikarwar/agent-skills-generator
cd agent-skills-generator
cargo build --release

The binary will be at target/release/agent-skills-generator.

Download pre-built binaries from the Releases page.

agent-skills-generator init

This launches an interactive wizard that guides you through setting up:

• Target IDE/agent (Cursor, Claude Code, GitHub Copilot, etc.)
• Installation scope (project or user level)
• Crawl settings (delay, depth, concurrency)

The wizard creates a skills.yaml configuration file.

Tip: Use --no-interactive to skip prompts and create a default config:

agent-skills-generator init --no-interactive

agent-skills-generator crawl https://docs.example.com

Generated skills are saved to .agent/skills/ by default:

.agent/skills/
├── getting-started/
│   └── SKILL.md
├── api-reference/
│   └── SKILL.md
└── tutorials-basics/
    └── SKILL.md

Each SKILL.md contains:

---
name: getting-started
description: Learn how to get started with our platform
metadata:
  url: https://docs.example.com/getting-started
---

# Getting Started

[Full documentation content converted to clean markdown...]

Create a skills.yaml file to customize crawling behavior:

# Output directory for generated skills
output: .agent/skills

# Crawl settings
delay_ms: 100           # Delay between requests
max_depth: 25           # Maximum crawl depth
request_timeout_secs: 30
respect_robots_txt: true
subdomains: false
concurrency: 4          # Parallel page processing

# URL filtering rules
rules:
  # Only crawl documentation pages
  - url: "*/docs/*"
    action: allow

  # Ignore authentication pages
  - url: "*/login*"
    action: ignore
  - url: "*/auth/*"
    action: ignore

  # Ignore API internals
  - url: "*/api/internal/*"
    action: ignore

# CSS selectors for elements to remove
remove_selectors:
  - ".advertisement"
  - "#cookie-banner"
  - ".feedback-widget"

agent-skills-generator validate --show

# Verbose output
agent-skills-generator -v crawl https://docs.example.com

# Very verbose (debug)
agent-skills-generator -vv crawl https://docs.example.com

# Custom config file
agent-skills-generator -c my-config.yaml crawl https://docs.example.com

# Custom output directory
agent-skills-generator -o ./my-skills crawl https://docs.example.com

# Limit pages crawled
agent-skills-generator crawl https://docs.example.com --max-pages 50

# Dry run (don't write files)
agent-skills-generator crawl https://docs.example.com --dry-run

Generate skills for specific AI coding assistants:

# Generate for Cursor (outputs to .cursor/skills/)
agent-skills-generator --target cursor crawl https://docs.example.com

# Generate for Claude Code (outputs to .claude/skills/)
agent-skills-generator --target claude-code crawl https://docs.example.com

# Generate for VS Code Copilot (outputs to .github/skills/)
agent-skills-generator --target github-copilot crawl https://docs.example.com

# Install at user level (~/.cursor/skills/)
agent-skills-generator --target cursor --user crawl https://docs.example.com

Supported Targets:

You can also set the target in skills.yaml:

# Target IDE/agent
target: cursor

# Scope: "project" or "user"
scope: project

agent-skills-generator crawl https://docs.flutter.dev/ui

agent-skills-generator crawl \
  https://docs.example.com/getting-started \
  https://docs.example.com/tutorials \
  https://docs.example.com/api

agent-skills-generator single https://docs.example.com/quick-start --stdout

agent-skills-generator crawl https://docs.example.com --resume

┌─────────────────┐     ┌─────────────────┐     ┌─────────────────┐
│   Web Crawler   │────▶│ Content Cleaner │────▶│ Skill Generator │
│                 │     │                 │     │                 │
│ • Async crawl   │     │ • Remove nav    │     │ • YAML metadata │
│ • Robots.txt    │     │ • Remove ads    │     │ • Clean markdown│
│ • URL filtering │     │ • Remove noise  │     │ • Organized dirs│
└─────────────────┘     └─────────────────┘     └─────────────────┘

1. Crawl — Discovers and fetches pages respecting robots.txt and rate limits
2. Clean — Removes navigation, scripts, styles, ads, and other noise
3. Convert — Transforms HTML to clean, token-efficient markdown
4. Generate — Creates structured SKILL.md files with metadata

Give your AI agents access to up-to-date documentation:

# Load skills into your agent's context
skills_dir = ".agent/skills"
for skill in load_skills(skills_dir):
    agent.add_context(skill)

Create searchable documentation archives:

agent-skills-generator crawl https://docs.company.com -o ./docs-archive

Build knowledge bases for RAG systems:

agent-skills-generator crawl https://wiki.example.com --max-depth 10

Agent Skills Generator is built for speed:

Contributions are welcome! Please feel free to submit a Pull Request.

1. Fork the repository
2. Create your feature branch (git checkout -b feature/amazing-feature)
3. Commit your changes (git commit -m 'Add amazing feature')
4. Push to the branch (git push origin feature/amazing-feature)
5. Open a Pull Request

# Run tests
cargo test

# Run with debug logging
RUST_LOG=debug cargo run -- crawl https://example.com

# Check formatting
cargo fmt --check

# Run clippy
cargo clippy --all-targets --all-features -- -D warnings

This project is licensed under the MIT License - see the LICENSE file for details.