简体中文

Harness Engineering: From Claude Code Internals to AI Coding Best Practices

This repository publishes a technical book about Harness Engineering through the lens of Claude Code. The material is derived from the publicly distributed Claude Code v2.1.88 package and source map reconstruction, then turned into a book focused on reusable engineering patterns rather than product marketing or unofficial feature rumors.

The English edition is currently a scaffolded preview with the full table of contents, cover, and placeholder chapter pages. The Chinese edition remains the canonical version while the English translation is added chapter by chapter.

Read Online

• English edition (preview): https://zhanghandong.github.io/harness-engineering-from-cc-to-ai-coding/en/
• Chinese edition: https://zhanghandong.github.io/harness-engineering-from-cc-to-ai-coding/

What the Book Covers

• Claude Code architecture, the Agent Loop, and tool execution orchestration
• System prompts, tool prompts, and model-specific tuning
• Automatic compaction, token budgeting, and prompt caching
• Permission modes, safety rules, hooks, and user instruction overlays
• Multi-agent orchestration, the skills system, and feature-flag pipelines
• Production-oriented lessons for building AI coding systems

Audience

• Engineers building AI coding products or agent infrastructure
• Developers who want to understand how Claude Code is put together
• Teams looking for transferable implementation patterns from a real-world agent system

Local Preview

mdbook build book
mdbook serve book

Default local URL:

• http://localhost:3000

Notes

• This book is based on analysis of publicly distributed artifacts and is intended for research and engineering discussion
• It does not represent official Anthropic documentation or statements
• This repository tracks only the files required to publish the book to GitHub Pages