A universal skill ecosystem for AI agents.

Build once. Reuse everywhere. Run skills across Claude Code, Cursor, OpenClaw, and more.

🎨 Works with any AI agent that supports skill/instruction files

Claude Code

Cursor

Codex CLI

Continue

OpenClaw

+ Custom Agents

🇨🇳 中文 • 🇯🇵 日本語 • 🇰🇷 한국어 • 🇪🇸 Español • 🇫🇷 Français • 🇩🇪 Deutsch • 🇧🇷 Português
🇷🇺 Русский • 🇸🇦 العربية • 🇮🇹 Italiano • 🇻🇳 Tiếng Việt • 🇹🇷 Türkçe

📱 Scan to join WeChat Group

Quick Start • Features • Documentation

📖 What is a Skill?

A Skill is a portable instruction file for AI agents. It defines triggers, workflows, and parameters:

---
name: git-commit-helper
description: Activate when user asks to "commit changes" or "create a commit"
triggers:
  - commit changes
  - create commit
---

## Workflow

1. Stage changes with `git add`
2. Generate commit message based on diff
3. Create commit with `git commit`

Skills are just markdown. Easy to write, easy to share, version-controlled with your code.

🤔 What Problem Does Skillbolt Solve?

Every time you use an AI coding assistant, you repeat the same instructions — "follow conventional commits", "use our code style", "check for security issues"...

Your AI forgets. Every. Single. Time.

Skillbolt fixes this by turning your instructions into reusable Skills — and then goes further with LLM-powered discovery and intelligent execution:

🎯 Create Author skills once as portable markdown files. Works with Claude Code, Cursor, Codex, and more.

🧠 Discover LLM-powered capability tree + intelligent search finds the right skills for any task automatically.

🚀 Execute Auto-generated DAG plans orchestrate multi-skill workflows with parallel execution and cost tracking.

✨ Features

Skill lifecycle

• ⚡ CLI-first workflow — create, validate, package, and publish Skills
• 🧬 Skill distillation — extract reusable Skills from agent conversations
• 📦 Modular packages — use the full stack or only what you need

Discovery & routing

• 🌳 Capability taxonomy — organize hundreds of Skills into a navigable tree
• 🔍 Intelligent retrieval — fast matching with pruning + ranking across tree levels

Orchestration & visibility

• 📊 DAG planning — compose Skills into parallelizable execution plans
• 🖥️ Web dashboard — run status, logs, and DAG viewer in real time
• 🐾 OpenClaw sync — share Skills across multi-channel agent surfaces

🚀 Quick Start

1. Install

npm install -g @skillbolt/cli

Other package managers

pnpm add -g @skillbolt/cli    # pnpm
npx @skillbolt/cli --help      # npx (no install)

2. Configure

Set your LLM provider so Skillbolt can power intelligent features:

export LLM_API_KEY=sk-...                  # Your OpenAI or Anthropic key
export LLM_PROVIDER=openai                 # "openai" | "anthropic"

Optional: project-level config

Create .skillboltrc.json in your project root:

{
  "llm": { "provider": "openai", "model": "gpt-4o-mini" }
}

See Configuration Guide for all options.

3. Create Your First Skill

skill init my-skill --template standard --platform claude-code

4. Discover & Install Skills

skill search git                           # Browse marketplace
skill install github:skillbolt/git-master  # Install from GitHub

5. Validate

skill lint ./my-skill --fix

📚 Documentation

📖 Browse Full Documentation — All guides, references, and architecture docs on GitHub

Getting Started

• Installation Guide — Prerequisites, setup & your first Skillbolt project
• SKILL.md Format — Skill file specification — triggers, workflows, parameters & metadata
• Command Reference — Complete documentation for all 20+ CLI commands
• Configuration — Environment variables, .skillboltrc.json & LLM provider settings

Core Concepts

• Runtime Intelligence — Capability trees, intelligent search, DAG orchestration & visual dashboard
• Capability Trees — LLM-powered classification organizes hundreds of skills into navigable hierarchies
• Intelligent Search — Multi-level tree traversal with automatic pruning finds the best skills for any task
• DAG Orchestration — Compose skills into parallelizable execution plans from natural language descriptions
• Skill Distillation — Extract reusable skills from real AI conversation history

Architecture & Integration

• API Reference — Programmatic usage for building custom integrations and pipelines
• Package Architecture — 17 modular packages — use the full CLI or pick individual components
• OpenClaw Integration — Sync and share skills across 13+ messaging channels via the OpenClaw gateway
• Cross-Platform Support — Write once, run on Claude Code, Cursor, Codex CLI, Continue & more

Development

• Contributing — Development setup, build instructions, testing & PR workflow
• GitHub Issues — Bug reports, feature requests & roadmap discussions
• Discord Community — Get help, share skills & connect with other developers