Welcome to Learn Harness Engineering 欢迎学习 Harness Engineering（工程约束设计）

Learn Harness Engineering is a course dedicated to the engineering of AI coding agents. We have deeply studied and synthesized the most advanced Harness Engineering theories and practices in the industry. Our core references include:Learn Harness Engineering 是一门专注于 AI 编程智能体工程设计的课程。我们深入研究并综合了业内最先进的 Harness Engineering 理论与实践。我们的核心参考资料包括：

• OpenAI: Harness engineering: leveraging Codex in an agent-first worldOpenAI：Harness engineering: leveraging Codex in an agent-first world（工程约束设计：在智能体优先的世界中利用 Codex）
• Anthropic: Effective harnesses for long-running agentsAnthropic：Effective harnesses for long-running agents（针对长时运行智能体的有效约束设计）
• Anthropic: Harness design for long-running application developmentAnthropic：Harness design for long-running application development（长时运行应用程序开发的约束设计）
• Awesome Harness EngineeringAwesome Harness Engineering（Harness Engineering 精选资源）

Through systematic environment design, state management, verification, and control systems, this course teaches you how to make agentic coding tools like Codex and Claude Code truly reliable. It helps you build features, fix bugs, and automate development tasks by constraining your AI coding assistant with explicit rules and boundaries.通过系统化的环境设计、状态管理、验证和控制系统，本课程将教你如何使 Codex 和 Claude Code 等智能体编程工具真正变得可靠。它能帮助你通过明确的规则和边界来约束 AI 编程助手，从而构建功能、修复漏洞并实现开发任务自动化。

Get started 开始学习

Choose your learning path to get started. The course is divided into theoretical lectures, hands-on projects, and a copy-ready resource library.选择你的学习路径以开始。本课程分为理论讲座、实践项目和可直接使用的资源库。

Lectures讲座

Understand why strong models still fail and learn the theory behind effective harnesses.了解为什么强大的模型仍然会失败，并学习有效约束设计背后的理论。

Projects项目

Hands-on practice building a reliable agentic environment from scratch.从零开始构建可靠智能体环境的动手实践。

Resource Library资源库

Copy-ready templates (AGENTS.md, feature_list.json) to use in your own repositories.可直接复制的模板（AGENTS.md, feature_list.json），供你在自己的代码仓库中使用。

Frontier Harness Design Breakdowns前沿约束设计案例解析

How Pi, Claude Code, Codex, and DeepSeek actually design their harnesses—mapped to the course framework.Pi、Claude Code、Codex 和 DeepSeek 是如何实际设计其约束系统的——并将其映射到本课程框架中。

The Core Mechanism of a Harness 约束设计的核心机制

A harness doesn't "make the model smarter"; rather, it establishes a closed-loop working system for the model. You can understand its core workflow through this simple diagram:约束设计并不会“让模型变得更聪明”；相反，它为模型建立了一个闭环的工作系统。你可以通过这张简单的图表理解其核心工作流程：

74x6m8

What you will learn 你将学到什么

Here are some of the key concepts you will master:以下是你将掌握的一些关键概念：

• Constrain agent behavior with explicit rules and boundaries.通过明确的规则和边界来约束智能体行为。
• Maintain context across long-running, multi-session tasks.在长时运行的多会话任务中保持上下文。
• Stop agents from declaring victory too early.防止智能体过早宣布任务完成。
• Verify work using full-pipeline tests and self-reflection.使用全流程测试和自我反思来验证工作成果。
• Make runtime observable and debuggable.使运行时具备可观测性和可调试性。

Next steps 后续步骤

Once you understand the core concepts, these guides help you go deeper:一旦理解了核心概念，这些指南将帮助你深入学习：

• Lecture 01: Why Capable Agents Still Fail: Start with the theory behind harness engineering.讲座 01：为什么强大的智能体仍会失败：从约束设计背后的理论开始。
• Project 01: Baseline vs Minimal Harness: Walk through your first real task.项目 01：基准与最小化约束设计：完成你的第一个实际任务。
• Templates: Grab the minimal harness pack (AGENTS.md, feature_list.json, claude-progress.md) for your own projects.模板：获取适用于你项目的最小化约束设计包（AGENTS.md, feature_list.json, claude-progress.md）。