Repolode
Submit
walkinglabs

walkinglabs/learn-harness-engineering

A project-based course teaching how to build the environment, state, verification, and control mechanisms that make AI coding agents reliable.

View on GitHubWebsite
Stars
23,454
Forks
2,700
Open issues
18
Last commit
16 days ago

What is it?

Learn Harness Engineering is a beginner-friendly course on the practice of building the surrounding system around an AI coding agent. Its premise is that a strong model still fails on real engineering tasks without instructions, persisted state, verification, scope limits, and session lifecycle rules. The course organizes harness design into five subsystems: instructions, state, verification, scope, and session lifecycle.

The course is structured as 14 lectures and 8 hands-on projects that all build on the same Electron-based personal knowledge base desktop app. Each project compares weak and strong harness setups on the same task, so learners can see the measured difference. Later lectures cover loop engineering (goal loops, timer loops, maker-checker separation) and graph engineering (nodes, edges, shared state, routing).

The repository also includes a resource library of copy-ready templates such as AGENTS.md, feature_list.json, and init.sh, a reusable harness-creator skill, and content translated into 15 languages. The course text is hosted on a VitePress documentation site.

Key features

  • 01Five-subsystem framework covering instructions, state, verification, scope, and session lifecycle.
  • 0214 lectures, each answering one core question about agent reliability.
  • 038 hands-on projects that evolve a single Electron knowledge base app.
  • 04Copy-ready templates including AGENTS.md, feature_list.json, and init.sh.
  • 05Coverage of loop engineering and graph engineering for multi-step agent workflows.
  • 06Content available in 15 languages, including Simplified Chinese.

Use cases

  • Engineers who already use Claude Code, Codex, or similar agents on real repositories and want steadier results.
  • Tech leads evaluating how repository structure and environment design affect agent output.
  • Developers who want to scaffold AGENTS.md, progress logs, and verification steps for an existing project.

Quick start

  1. 1

    Clone the repository

  2. 2

    Use the harness-creator skill to scaffold a harness for your project (see README quick start)

Things to know

  • The project is a course of written lectures and projects, not an installable software tool.

In collections

Open-source AI agent frameworks & appsAI coding assistants & agents

Topics

#ai-agent#llm#agentic-ai#harness-engineering#prompt-engineering#tutorial#coding-agent#agent#agentic#ai

FAQ

What is learn-harness-engineering?
A project-based course teaching how to build the environment, state, verification, and control mechanisms that make AI coding agents reliable. Learn Harness Engineering is a beginner-friendly course on the practice of building the surrounding system around an AI coding agent. Its premise is that a strong model still fails on real engineering tasks without instructions, persisted state, verification, scope limits, and session lifecycle rules. The course organizes harness design into five subsystems: instructions, state, verification, scope, and session lifecycle.
Is learn-harness-engineering free to use? What license does it have?
learn-harness-engineering is open source under the MIT license. Check the license terms before using it in commercial products.
Is learn-harness-engineering still maintained?
Actively maintained. The last commit to learn-harness-engineering was on Oct 22, 2026.
What can I use learn-harness-engineering for?
Engineers who already use Claude Code, Codex, or similar agents on real repositories and want steadier results. Tech leads evaluating how repository structure and environment design affect agent output. Developers who want to scaffold AGENTS.md, progress logs, and verification steps for an existing project.
How do I install learn-harness-engineering?
Clone the repository Use the harness-creator skill to scaffold a harness for your project (see README quick start)
Where can I get learn-harness-engineering?
The source code and installation instructions are on GitHub: https://github.com/walkinglabs/learn-harness-engineering

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects