lg-design

Guide eight structured questions to craft a LangGraph agent design doc.

4|Updated Apr 15, 2026

Search Tags:#design #planning #architecture #langgraph #design-doc #agent-design #pattern-selection

Ready to run

npx skills add https://github.com/markhazlett/agent-harness --skill lg-design

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: lg-design
Source: https://github.com/markhazlett/agent-harness/tree/main/.claude/skills/lg-design
Command: npx skills add https://github.com/markhazlett/agent-harness --skill lg-design

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This skill helps teams design LangGraph agents before implementation by guiding a structured design process and producing a formal design doc.

Core Features & Use Cases

Eight-phase conversation that asks one question at a time: purpose, trigger, external surfaces, multi-agent setup, run length, HITL, streaming, and memory.
Pattern recommendation and design artifact generation to feed scaffolding tools.
Works in sprint or standalone mode to generate a reusable design doc for future code.

Quick Start

Begin by defining the agent's core purpose and expected workflow, then let the skill generate a complete LangGraph design doc.

Frequently Asked Questions about lg-design

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I design a LangGraph agent before writing code?▼

Guide LangGraph agent design by answering eight structured questions covering purpose, triggers, external surfaces, and memory. This produces a formal design doc and a slug suitable for code scaffolding.

What is the best way to select an agent design pattern for LangGraph?▼

Selecting an agent design pattern for LangGraph involves evaluating multi-agent setup, run length, and human-in-the-loop needs. A guided design process captures these inputs to recommend the appropriate pattern for your workflow.

How do I create a design doc for LangGraph multi-agent workflows?▼

Generate a LangGraph design doc by defining the agent's core purpose and expected workflow, then completing an eight-phase conversation. This produces a formal design artifact capturing memory, streaming, and persistence.

Can I use a structured process to plan LangGraph human-in-the-loop and streaming requirements?▼

Yes, you can plan LangGraph human-in-the-loop and streaming requirements through a structured design process. It explicitly asks about HITL, streaming, and persistence to ensure these operational needs are formalized before implementation.

Does LangGraph agent design require defining external surfaces and triggers?▼

LangGraph agent design requires defining external surfaces and triggers to formalize the workflow. Capturing these inputs ensures the generated design doc accurately reflects how the agent interacts with external systems and initiates actions.