@langchain/skills

A deepagents-powered agent that scrapes LangChain documentation and generates skill.md files following the Agent Skills protocol.

Overview

This agent automatically explores the LangChain documentation and creates comprehensive skill files that help AI agents use LangChain effectively. Each skill file includes:

• Decision tables for choosing the right components
• Code examples with essential patterns
• Boundaries defining what can/cannot be configured
• Gotchas highlighting common mistakes
• Links to full documentation

Installation

bun install

Environment Variables

You'll need to set up your API keys:

export ANTHROPIC_API_KEY=your-anthropic-api-key

Usage

Generate All Skills

Run the agent to automatically explore LangChain docs and generate skill files:

bun run start

Or:

bun run src/agent.ts

Output

Skill files are generated in the ./skills/ directory with the structure:

skills/
├── langchain-chat-models/
│   └── SKILL.md
├── langchain-prompts/
│   └── SKILL.md
├── langchain-chains/
│   └── SKILL.md
├── langchain-agents/
│   └── SKILL.md
├── langchain-memory/
│   └── SKILL.md
├── langchain-rag/
│   └── SKILL.md
├── langchain-document-loaders/
│   └── SKILL.md
├── langchain-vector-stores/
│   └── SKILL.md
├── langchain-output-parsers/
│   └── SKILL.md
└── langchain-callbacks/
    └── SKILL.md

Skill File Format

Each skill file follows the Agent Skills protocol:

---
name: skill-name
description: Brief description of the skill
---

# Skill Title

## Overview
Brief summary of the topic.

## Decision Table
| Need | Use |
| --- | --- |
| Simple Q&A | ChatOpenAI |
| Complex reasoning | Claude |

## Code Examples
Essential usage patterns.

## Boundaries
### Can Configure
- List of configurable items

### Requires External Setup
- List of items needing external setup

## Gotchas
- Common mistakes to avoid

## Links
- [Full Documentation](https://docs.langchain.com/...)

How It Works

1. MCP Integration: Uses the LangChain docs MCP server to search documentation
2. Deep Agent: Powered by deepagents with filesystem middleware
3. Systematic Exploration: Explores all major LangChain topics
4. Skill Generation: Creates structured skill files following the protocol

Development

This project uses:

• Bun - JavaScript runtime
• deepagents - Agentic AI framework
• @modelcontextprotocol/sdk - MCP client