A prompt-based skill that makes AI coding agents write only the code a task needs, favoring reuse and minimal solutions over large implementations.
wquguru/harness-books
Two free online books explaining how agent harnesses for code-writing models, such as Claude Code and Codex, are designed and controlled.
- Stars
- 3,194
- Forks
- 387
- Open issues
- 6
- Last commit
- 6 months ago
What is it?
Harness Books is a pair of books on harness engineering: the surrounding system that controls a code-writing model once it runs inside terminals, repositories, permission systems and team workflows. The books cover prompt layering, query loops, tool permissions, context governance, error recovery, multi-agent verification and team rules, with the goal of keeping such a system bounded and accountable.
The first book, A Design Guide to Claude Code, uses Claude Code as its subject and focuses on runtime structure, explaining why a system grows components such as a control plane, query loop and recovery paths. The second book, The Harness Design Philosophies of Claude Code and Codex, compares the two tools side by side and examines where each one places control over execution, including sandboxes, policy languages, skills and hooks.
Both books are available to read online in English and Chinese, and each can be downloaded as a PDF. The repository also includes a Python build script for generating the sites locally.
Key features
- 01Covers query loops, tool permissions, context governance and error recovery in agent systems.
- 02Compares Claude Code and Codex on control-plane design, sandboxes and policy languages.
- 03Explains how skills, hooks and local rules encode team habits into an agent system.
- 04Available to read online in English and Chinese, with PDF downloads.
- 05Includes a Python script to build the bilingual book sites locally.
Use cases
- Developers building their own coding agent who want to understand its control layers.
- Engineering leads deciding how to govern AI coding tools in a team.
- Engineers comparing Claude Code and Codex architectures before choosing a tool.
Quick start
- 1
Read the books online
- 2
Build the books locally (Book 1, English)
python3 tools/book-kit/build_honkit.py book1-claude-code --locale en - 3
Assemble the unified Pages site
python3 tools/book-kit/build_pages_site.py
Things to know
- The project is a set of books, not software, and it has no license listed.
- The local build requires running Python build scripts and writes output to dist/.
Topics
FAQ
- What is harness-books?
- Two free online books explaining how agent harnesses for code-writing models, such as Claude Code and Codex, are designed and controlled. Harness Books is a pair of books on harness engineering: the surrounding system that controls a code-writing model once it runs inside terminals, repositories, permission systems and team workflows. The books cover prompt layering, query loops, tool permissions, context governance, error recovery, multi-agent verification and team rules, with the goal of keeping such a system bounded and accountable.
- Is harness-books free to use? What license does it have?
- harness-books has no declared license. Its code is public, but without a license you have no explicit permission to reuse it; ask the author before relying on it.
- Is harness-books still maintained?
- No commits for 6+ months. The last commit to harness-books was on Apr 19, 2026.
- What can I use harness-books for?
- Developers building their own coding agent who want to understand its control layers. Engineering leads deciding how to govern AI coding tools in a team. Engineers comparing Claude Code and Codex architectures before choosing a tool.
- How do I install harness-books?
- Read the books online Build the books locally (Book 1, English): python3 tools/book-kit/build_honkit.py book1-claude-code --locale en Assemble the unified Pages site: python3 tools/book-kit/build_pages_site.py
- Where can I get harness-books?
- The source code and installation instructions are on GitHub: https://github.com/wquguru/harness-books
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
A portable Agent Skill that removes AI-writing tells from prose, repairing narrative structure in fiction and applying venue-specific rules to professional documents.
A VS Code extension that puts Claude Code, Codex and other coding agents in one conversation, so they can give separate opinions, hand off work and debate solutions.
A curated directory of 1000+ Claude Skills and plugins for Claude Code, Claude.ai and coding agents like Codex, Cursor and Gemini CLI.
A curated, continuously updated best-practices guide and reference collection for using Anthropic's Claude Code, covering subagents, commands, skills, hooks, memory and workflows.
A curated GitHub list of practical Codex skills for automating coding, productivity and data workflows in the Codex CLI and API.