Agent Harness 101

《从零构建 AI Agent 与 Harness Engineering 实战课程》静态课程站点。

English: This is the static course site for "Build AI Agents and Harness Engineering from Scratch" (Chinese).

本课程以 Harness 12-Factor 为主线组织：前置 → F1-F12 → 进阶阅读 → 结课项目。每章提供骨架代码（starter/）与参考实现（solution/），把“能跑的 Agent”打磨成“可上线 Harness”。

📚 Course Contents

A hands-on course covering:

• AI Agent fundamentals and design patterns
• Harness Engineering for reliable, observable agent systems
• The Harness 12-Factor methodology:
  1. Single Agent Loop
  2. Explicit Tool Contract
  3. Context Budgeting
  4. Failure-First Design
  5. Graceful Degradation
  6. Least-Privilege Tooling
  7. Human-in-the-Loop Gates
  8. Observable by Default
  9. Reproducible Runs
  10. Composable Agents
  11. Config-Driven Behavior
  12. Continuous Evaluation
• Tool use, planning, memory, sandboxing, and evaluation
• Building production-ready agent workflows

🚀 Quick Start

No build step required. Open the course directly:

# macOS / Linux
open index.html

# Windows
start index.html

Or serve it locally:

python -m http.server 8000
# Visit http://localhost:8000

🌐 Deploy

Cloudflare Pages（默认）

本项目通过 wrangler.jsonc 部署到 Cloudflare Pages：

npx wrangler pages deploy .

访问：https://harness-101.pages.dev/

GitHub Pages（备选）

1. Push the main branch to GitHub.
2. Go to Settings → Pages.
3. Select Deploy from a branch.
4. Choose main and root /.
5. Visit https://<username>.github.io/harness-101/.

🏗️ Repository Structure

├── index.html                 # Single-file course website (CSS/JS inlined)
├── agent-harness-course.md    # Course Markdown source (single source of truth)
├── GLOSSARY.md                # Consistent terminology for contributors
├── examples/                  # Per-Factor starter code and reference solutions
│   ├── prep/                  # 前置：环境与第一次 API 调用
│   ├── factor-1/              # Single Agent Loop
│   ├── factor-2/              # Explicit Tool Contract
│   ├── factor-3/              # Context Budgeting
│   ├── factor-4/              # Failure-First Design
│   ├── factor-5/              # Graceful Degradation
│   ├── factor-6/              # Least-Privilege Tooling
│   ├── factor-7/              # Human-in-the-Loop Gates
│   ├── factor-8/              # Observable by Default
│   ├── factor-9/              # Reproducible Runs
│   ├── factor-10/             # Composable Agents
│   ├── factor-11/             # Config-Driven Behavior
│   └── factor-12/             # Continuous Evaluation
├── wrangler.jsonc             # Cloudflare Pages deployment config
├── sitemap.xml                # Sitemap for search engines
├── robots.txt                 # Robots directives
├── _headers                   # Cloudflare Pages response headers
├── README.md                  # This file
└── CONTRIBUTING.md            # Contribution guidelines

✏️ Contributing

Content changes should be made in agent-harness-course.md first, then synchronized to index.html.

For typo fixes, content corrections, or translation proposals, please open an Issue or Pull Request.

See CONTRIBUTING.md for details.

📄 License

This project is licensed under the MIT License.

🔗 Keywords

ai-agent agent-engineering harness-engineering llm prompt-engineering tool-use rag course education chinese github-pages