langchain-skills

内容来源：README.md（说明文档） · 原始地址 · 查看安装指南

原始内容

# LangChain Skills

> **⚠️** — This project is in early development. APIs and skill content may change.

Agent skills for building agents with LangChain, LangGraph, and Deep Agents.

> For LangSmith-specific trace and dataset workflows, use [langsmith-skills](https://github.com/langchain-ai/langsmith-skills).

## Supported Coding Agents

These skills can be installed via [`npx skills`](https://github.com/vercel-labs/skills) for any agent that supports the [Agent Skills specification](https://skills.sh), including Claude Code, Cursor, Windsurf, and more.

## Installation

### Quick Install

Using [`npx skills`](https://github.com/vercel-labs/skills):

**Local** (current project):

```bash
npx skills add langchain-ai/langchain-skills --skill '*' --yes
```
**Global** (all projects):

```bash
npx skills add langchain-ai/langchain-skills --skill '*' --yes --global
```
To link skills to a specific agent (e.g. Claude Code):

```bash
npx skills add langchain-ai/langchain-skills --agent claude-code --skill '*' --yes --global
```

---

### Claude Code Plugin

Install directly as a [Claude Code plugin](https://code.claude.com/docs/en/plugins):

```bash
/plugin marketplace add langchain-ai/langchain-skills
/plugin install langchain-skills@langchain-skills
```

---

### Install Script (Claude Code & Deep Agents CLI only)

Alternatively, clone the repo and use the install script:

```bash
# Install for Claude Code in current directory (default)
./install.sh

# Install for Claude Code in a specific project directory
./install.sh ~/my-project

# Install for Claude Code globally
./install.sh --global

# Install for Deep Agents CLI in a specific project directory
./install.sh --deepagents ~/my-project

# Install for Deep Agents CLI globally (includes agent persona)
./install.sh --deepagents --global
```

| Flag / Argument | Description |
|------|-------------|
| `DIRECTORY` | Target project directory (default: current directory, ignored with `--global`) |
| `--claude` | Install for Claude Code (default) |
| `--deepagents` | Install for Deep Agents CLI |
| `--global`, `-g` | Install globally instead of current directory |
| `--force`, `-f` | Overwrite skills with same names as this package |
| `--yes`, `-y` | Skip confirmation prompts |

## Usage

After installation, set your API keys:

```bash
export OPENAI_API_KEY=<your-key>      # For OpenAI models
export ANTHROPIC_API_KEY=<your-key>   # For Anthropic models
```

Then run your coding agent from the directory where you installed (for local installs) or from anywhere (for global installs).

### Eval Engineering

To install only the eval-engineering skill:

```bash
npx skills add langchain-ai/langchain-skills --skill eval-engineering --yes
```

Or ask Codex: “Install the `eval-engineering` skill from `langchain-ai/langchain-skills`.”

Eval tasks require [Harbor](https://www.harborframework.com/docs). Run Harbor locally with Docker or use a supported cloud environment.

Then ask your coding agent:

```text
Use eval-engineering to inspect this agent and propose two or three Harbor evals. Traces for this agent can be found here [Tracing Project/Location].
```

## Available Skills (21)

### Getting Started
- **ecosystem-primer** - Start-here primer: framework selection (LangChain vs LangGraph vs Deep Agents), env setup, and which skill to load next
- **langchain-dependencies** - Full package version and dependency management reference (Python + TypeScript)

### Quickstarts (local)
Thin wrappers around the official Mintlify quickstarts — ask for provider/model (default `anthropic:claude-sonnet-5`), new directory, provider API key only:
- **langchain-python-quickstart** / **langchain-typescript-quickstart** → [Python](https://docs.langchain.com/oss/python/langchain/quickstart) / [JS](https://docs.langchain.com/oss/javascript/langchain/quickstart) (weather)
- **langgraph-python-quickstart** / **langgraph-typescript-quickstart** → [Python](https://docs.langchain.com/oss/python/langgraph/quickstart) / [JS](https://docs.langchain.com/oss/javascript/langgraph/quickstart) (math)
- **deepagents-python-quickstart** / **deepagents-typescript-quickstart** → [Python](https://docs.langchain.com/oss/python/deepagents/quickstart) / [JS](https://docs.langchain.com/oss/javascript/deepagents/quickstart) (research; provider web search instead of Tavily)

### Deep Agents
- **deep-agents-core** - Agent architecture, harness setup, and SKILL.md format
- **deep-agents-memory** - Memory, persistence, filesystem middleware
- **deep-agents-orchestration** - Subagents, task planning, human-in-the-loop
- **managed-deep-agents** - Managed Deep Agents: deploy with the CLI, use the SDKs, stream runs, connect MCP tools, and build React `useStream` UIs

### LangChain
- **langchain-fundamentals** - Agents with create_agent, tools, structured output, middleware basics
- **langchain-middleware** - Human-in-the-loop approval, custom middleware, Command resume patterns
- **langchain-rag** - RAG pipeline (document loaders, embeddings, vector stores)

### LangGraph
- **langgraph-fundamentals** - StateGraph, nodes, edges, state reducers
- **langgraph-persistence** - Checkpointers, thread_id, cross-thread memory
- **langgraph-cli** - CLI lifecycle: scaffold, dev, build, deploy, langgraph.json config
- **langgraph-human-in-the-loop** - Interrupts, human review, approval workflows

### Evaluation
- **eval-engineering** - Build, run, and audit Harbor evals for an existing agent with user approval

### Utilities
- **swarm** - Dispatch independent work items in parallel and aggregate the results

LangChain Skills

⚠️ — This project is in early development. APIs and skill content may change.

Agent skills for building agents with LangChain, LangGraph, and Deep Agents.

For LangSmith-specific trace and dataset workflows, use langsmith-skills.

Supported Coding Agents

These skills can be installed via npx skills for any agent that supports the Agent Skills specification, including Claude Code, Cursor, Windsurf, and more.

Installation

Quick Install

Using npx skills:

Local (current project):

npx skills add langchain-ai/langchain-skills --skill '*' --yes

Global (all projects):

npx skills add langchain-ai/langchain-skills --skill '*' --yes --global

To link skills to a specific agent (e.g. Claude Code):

npx skills add langchain-ai/langchain-skills --agent claude-code --skill '*' --yes --global

Claude Code Plugin

Install directly as a Claude Code plugin:

/plugin marketplace add langchain-ai/langchain-skills
/plugin install langchain-skills@langchain-skills

Install Script (Claude Code & Deep Agents CLI only)

Alternatively, clone the repo and use the install script:

# Install for Claude Code in current directory (default)
./install.sh

# Install for Claude Code in a specific project directory
./install.sh ~/my-project

# Install for Claude Code globally
./install.sh --global

# Install for Deep Agents CLI in a specific project directory
./install.sh --deepagents ~/my-project

# Install for Deep Agents CLI globally (includes agent persona)
./install.sh --deepagents --global

Flag / Argument	Description
`DIRECTORY`	Target project directory (default: current directory, ignored with `--global`)
`--claude`	Install for Claude Code (default)
`--deepagents`	Install for Deep Agents CLI
`--global`, `-g`	Install globally instead of current directory
`--force`, `-f`	Overwrite skills with same names as this package
`--yes`, `-y`	Skip confirmation prompts

Usage

After installation, set your API keys:

export OPENAI_API_KEY=<your-key>      # For OpenAI models
export ANTHROPIC_API_KEY=<your-key>   # For Anthropic models

Then run your coding agent from the directory where you installed (for local installs) or from anywhere (for global installs).

Eval Engineering

To install only the eval-engineering skill:

npx skills add langchain-ai/langchain-skills --skill eval-engineering --yes

Or ask Codex: “Install the eval-engineering skill from langchain-ai/langchain-skills.”

Eval tasks require Harbor. Run Harbor locally with Docker or use a supported cloud environment.

Then ask your coding agent:

Use eval-engineering to inspect this agent and propose two or three Harbor evals. Traces for this agent can be found here [Tracing Project/Location].

Available Skills (21)

Getting Started

ecosystem-primer - Start-here primer: framework selection (LangChain vs LangGraph vs Deep Agents), env setup, and which skill to load next
langchain-dependencies - Full package version and dependency management reference (Python + TypeScript)

Quickstarts (local)

Thin wrappers around the official Mintlify quickstarts — ask for provider/model (default anthropic:claude-sonnet-5), new directory, provider API key only:

langchain-python-quickstart / langchain-typescript-quickstart → Python / JS (weather)
langgraph-python-quickstart / langgraph-typescript-quickstart → Python / JS (math)
deepagents-python-quickstart / deepagents-typescript-quickstart → Python / JS (research; provider web search instead of Tavily)

Deep Agents

deep-agents-core - Agent architecture, harness setup, and SKILL.md format
deep-agents-memory - Memory, persistence, filesystem middleware
deep-agents-orchestration - Subagents, task planning, human-in-the-loop
managed-deep-agents - Managed Deep Agents: deploy with the CLI, use the SDKs, stream runs, connect MCP tools, and build React useStream UIs

LangChain

langchain-fundamentals - Agents with create_agent, tools, structured output, middleware basics
langchain-middleware - Human-in-the-loop approval, custom middleware, Command resume patterns
langchain-rag - RAG pipeline (document loaders, embeddings, vector stores)

LangGraph

langgraph-fundamentals - StateGraph, nodes, edges, state reducers
langgraph-persistence - Checkpointers, thread_id, cross-thread memory
langgraph-cli - CLI lifecycle: scaffold, dev, build, deploy, langgraph.json config
langgraph-human-in-the-loop - Interrupts, human review, approval workflows

Evaluation

eval-engineering - Build, run, and audit Harbor evals for an existing agent with user approval

Utilities

swarm - Dispatch independent work items in parallel and aggregate the results