LangGraph - Skill

Purpose

This skill explains core patterns for building directed graph based AI workflows with LangGraph in Python. Use it for agent orchestration, stateful pipelines, subgraph composition, RAG plus node processing, and checkpointed experiments.

When an AI assistant should apply this skill

• When editing files that import langgraph or create StateGraph objects
• When authoring node functions that use LLMs or tools
• When adding checkpoints, persistence, or streaming to agent runs
• When composing subgraphs or multi-agent flows

Quick start

1. Model the workflow state with a TypedDict. 2. Write small pure node functions that accept and return partial state. 3. Use START as entry. 4. Compile the builder to get a runnable graph. 5. Add a checkpointer for persistence and safe experimentation.

Core concepts and cheat sheet