# Agent 工程实践

## 从工具循环到可靠系统

- [导读](https://levon.gitbook.io/agent-engineering/readme.md)
- [第 0 课——Agent 工程史](https://levon.gitbook.io/agent-engineering/00agent-gong-cheng-shi.md)
- [第 1 课：Agent 基础——从语言模型到行动系统](https://levon.gitbook.io/agent-engineering/01agent-ji-chu.md)
- [第 2 课：Agent Runtime——Model、Harness、Tool 与 Environment](https://levon.gitbook.io/agent-engineering/02agent-yun-xing-shi.md)
- [第 3 课：Tool Calling Loop——从调用请求到最终回答](https://levon.gitbook.io/agent-engineering/03-gong-ju-diao-yong-xun-huan.md)
- [第 4 课：Session 持久化——Transcript、Checkpoint 与 Memory](https://levon.gitbook.io/agent-engineering/04-hui-hua-chi-jiu-hua.md)
- [第 5 课：Context Engineering——Compaction 与 Prompt View](https://levon.gitbook.io/agent-engineering/05-shang-xia-wen-gong-cheng.md)
- [第 6 课：Tool Reliability——幂等、Ledger 与故障恢复](https://levon.gitbook.io/agent-engineering/06-gong-ju-ke-kao-xing.md)
- [第 7 课：Agent 执行安全——审批、权限与 Sandbox](https://levon.gitbook.io/agent-engineering/07agent-zhi-xing-an-quan.md)
- [第 8 课：Agent Tracing——用 Trace 与 Span 还原一次运行](https://levon.gitbook.io/agent-engineering/08agent-ke-guan-ce-xing.md)
- [第 9 课：Agent Evaluation——任务验收与回归门禁](https://levon.gitbook.io/agent-engineering/09agent-ping-gu.md)
- [第 10 课：Agent Orchestration——Workflow、Routing 与长任务](https://levon.gitbook.io/agent-engineering/10agent-bian-pai.md)
- [第 11 课：持久状态与云端执行——关掉网页后，任务还能接上吗](https://levon.gitbook.io/agent-engineering/11agent-chang-qi-gong-zuo-huan-jing-yu-yun-duan-zhi-xing.md)
- [第 12 课：RAG 检索增强生成——从资料入库到有据可查的回答](https://levon.gitbook.io/agent-engineering/12rag-jian-suo-zeng-qiang-sheng-cheng.md)
- [第 13 课：MCP 与 Skills——工具接入、任务方法与权限边界](https://levon.gitbook.io/agent-engineering/13mcp-yu-skills.md)
- [第 14 课：Agent 运行控制——并发、预算与资源释放](https://levon.gitbook.io/agent-engineering/14agent-yun-xing-kong-zhi.md)
- [第 15 课：Computer Use——从页面信息到真实操作](https://levon.gitbook.io/agent-engineering/15-computeruse.md)


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
This site has an agentic ask interface you may use to query the documentation dynamically by asking a question.

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://levon.gitbook.io/agent-engineering/index.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.