LangChainWhat Are Skills?
Skills are curated instructions, scripts, and resources that enhance AI coding agent performance in specialized domains. Unlike traditional tool-calling approaches where providing too many tools degrades agent performance, skills employ progressive disclosure—agents only load a skill when it's relevant to the current task. This design significantly improves agent effectiveness while keeping cognitive load minimal.
LangChain Skills Package
LangChain is now maintaining 11 open-source skills across three primary categories:
- LangChain: Guidance on
create_agent(), middleware patterns, and tool calling fundamentals - LangGraph: Instructions on working with LangGraph's primitives, Human-in-the-Loop support, and durable execution
- Deep Agents: Documentation on the open-source Deep Agents package, including prebuilt middleware and FileSystem utilities
The skills are portable and shareable as markdown files and scripts, making them compatible with any coding agent framework that supports skill functionality.
Significant Performance Gains
Testing with Claude Code (Sonnet 4.6) showed dramatic improvements:
| Scenario | Pass Rate |
|---|---|
| Without Skills | 25% |
| With Skills | 95% |
This 70-percentage-point improvement was measured using LangSmith evaluations. LangChain plans to open-source the full testing benchmark used for these measurements.
Getting Started
Install skills using npx skills:
# Local installation
npx skills add langchain-ai/langchain-skills --skill '*' --yes
# Global installation
npx skills add langchain-ai/langchain-skills --skill '*' --yes --global
# Link to Claude Code specifically
npx skills add langchain-ai/langchain-skills --agent claude-code --skill '*' --yes --global
Skills are available in the langchain-skills GitHub repository. LangChain plans to continue expanding the skills library as new capabilities are added to their open-source projects and LangSmith platform.