AI Agent: Design Principles and Engineering Practice is an open-source book that explains how AI agents combine language models, context, and tools, with accompanying experiments and code. It is intended for readers studying the principles and engineering of AI agents, from fundamentals through production use. The catalogue skills support coding-agent work related to the book's subject matter.
Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add bojieli/ai-agent-book --skill multi-agent-designgit clone --depth 1 https://github.com/bojieli/ai-agent-bookWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/bojieli/ai-agent-book/multi-agent-design)<a href="https://agentmods.dev/skills/bojieli/ai-agent-book/multi-agent-design"><img src="https://agentmods.dev/badge/skills/bojieli/ai-agent-book/multi-agent-design/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/bojieli/ai-agent-book/multi-agent-design"><img src="https://agentmods.dev/badge/skills/bojieli/ai-agent-book/multi-agent-design.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00126 | $0.03337 |
| Opus 5.5 | $0.00050 | $0.01335 |
| Sonnet 5.5 | $0.00025 | $0.00667 |
| Haiku 4.5 | $0.00013 | $0.00334 |
Grade A, and why
multi-agent-design scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 22d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
Copies of this mod
7 near-identical copies found in the catalogue:
- agent-orchestra — 92% identical, 444 lines differ
- multi_agent_collaboration — 89% identical, 167 lines differ
- chat_with_agent — 88% identical, 150 lines differ
- exp-orchestration-design — 88% identical, 178 lines differ
- summarize-article — 86% identical, 223 lines differ
- agent — 86% identical, 152 lines differ
- agent-flow — 83% identical, 401 lines differ
How it starts
The opening of the file, as written. The whole thing — 108 lines — stays where its author put it; the contents beside it link to each section on GitHub.
多 Agent 协作设计
何时使用
- 要搭建多 Agent 系统，或评审一个已有架构（"该用几个 Agent？谁指挥谁？"）
- 纠结"这个任务拆成多个 Agent 是否真的更好"，需要可判定的依据
- 决定 Agent 之间是共享上下文、还是各自独立上下文
- 设计 Agent 间通信机制（工具参数 / 共享文件系统 / 消息总线）与文件系统布局
- 决定角色切换应替换 system prompt，还是加载 Skill
- 多 Agent 系统出问题：文件互相覆盖、结论越传越歪、多个 Agent 不约而同做同一件事、互相扯皮、子 Agent 数量爆炸
核心原则
- 新信息判据（唯一实质判据）。 多 Agent 优于单 Agent，当且仅当协作过程引入了单个 Agent 在生成时无法获得的新信息。同一模型重新读自己的输出、多个 Agent 辩论同一段文本，都不引入新信息——等计算量下单 Agent 持平甚至更好；而用测试执行结果、渲染截图、外部工具输出来审查，则显著提升。学术研究说"多 Agent 无效"与工程实践"多 Agent 更好"并不矛盾：前者比较的是"看着同一段上下文讨论"，后者包含外部反馈环路。
- 两个维度先定，再谈实现。 维度一：上下文是否共享（继承式 vs 进程隔离式）；维度二：协作拓扑（对等 / 管理者 / 去中心化）。二者共同决定架构。
- 成本必须被收益覆盖。 并行探索与反复迭代要烧数倍乃至一个数量级的 token。收益不够大时，一个调校得当的单 Agent 更划算。
- 步骤预算不是越多越好。 单纯给 Agent 更多步骤/工具调用次数不保证性能提升——Agent 缺乏预算意识会很快"饱和"。Manager 应按子任务复杂度动态分配预算，并引导子 Agent"先规划、再实现、再测试、再改进"。
- 最强的模型给 Manager。 Plan-and-Act 的实证结论：弱规划者是系统最关键的瓶颈；规划错了，后续所有执行都建立在错误前提上。不要把资源平均分给所有 Agent。
- Agent 的故障是拜占庭式的。 它很少径直停止，而是继续给出看似可信的错误结论，且错误不会声明自己是错误。因此交叉验证与多数表决是必需手段，而非可选优化。
- 上下文隔离优先。 Manager 上下文中只放任务描述、计划、调用记录与文件索引；完整产物写入文件系统，Agent 间传递轻量路径字符串而非把内容载入上下文。
- handoff 不暴露私有轨迹。 对等移交只应传递明确的任务包和产物引用。
实践模式
1. 拓扑选择决策树
| 任务特征 | 选择 |
|---|---|
| 2-3 个角色，需要互相反馈、多轮迭代提质 | 对等协作（起草者/评论者、提议者/审核者） |
| 子任务多、需要动态调度、子任务间有复杂依赖 | 管理者模式（Manager 规划 + 子 Agent 执行） |
| 需职责对等的角色自主决定与谁沟通，或管理者不能成为单点故障 | 去中心化模式（handoff / 消息池订阅） |
| 开放搜索空间，需要广覆盖与多样化发现路径 | 多 Agent 并行（用更高 token 预算换覆盖） |
对等协作实现复杂度最低：定义好两个 Agent 的角色、通信机制和迭代终止条件即可跑起来。
2. 共享上下文 vs 不共享上下文
- 共享上下文：每阶段是独立 Agent（自己的 system prompt 与工具集），但继承前序完整轨迹。优势是信息不丢失；挑战是上下文快速膨胀，当前 Agent 容易被历史干扰。
- 不共享上下文：每个 Agent 有独立上下文与轨迹，彼此看不到对方的"思考过程"。模块化、隔离性、可扩展性更好，可真正并发；代价是信息同步与调试困难，接口规范与数据格式变得至关重要。
- 通信机制就三种，且都落在经典 IPC 范式内：工具调用参数（同步消息传递）、共享文件系统（共享内存）、消息总线（异步消息传递）。Agent 数量少且拓扑固定用点对点；数量多且需异步并行改消息总线——点对点连接数随 Agent 数平方增长且要求双方同时在线。
3. 角色切换：替换 system prompt 还是加载 Skill？
transfer_to_agent / 替换 system prompt |
Skill | |
|---|---|---|
| 工具可见性 | 只暴露当前角色工具 | 通常固定暴露全集 |
| 前缀缓存 | 每次切换改变请求前缀，缓存从变化点起失效 | 静态前缀不变，Skill 内容追加到末尾轨迹，前缀可复用 |
| 约束能力 | 强：越界工具在 schema 层不可见 | 弱：Skill 是行为指令，硬权限仍需 Harness 门 |
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 22d ago First seen · 108 lines · 126 tokens per session scan A c23551dec7b5
multi-agent-design is a skill published in the GitHub repository bojieli/ai-agent-book (54,406 stars, last pushed 7d ago), licensed Apache-2.0. It adds 126 tokens to every session and 3,337 once invoked, about $0.0005 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-24.
Other skills, from other repositories
remem-mcp
Long-term memory for coding agents. Auto-applies at the start of any coding task — recall past context before answering, capture decisions/learnings/fixes after work, use CodeGraph instead of grep for symbol lookup. Invoke when you see [remem-mcp] in your context or when starting any non-trivial coding work.
tura
Work in the Tura agent-runtime repository. Use for Tura architecture, Rust backend, GUI/TUI, prompts, commands, providers, sessions, documentation, tests, packaging, and release work in this directory.
acontext-installer
Install Acontext, Login & Init Acontext Project, Add Skill Memory to Agent.
ccc
This skill should be used when code search is needed (whether explicitly requested or as part of completing a task), when indexing the codebase after changes, or when the user asks about ccc, cocoindex-code, or the codebase index. Trigger phrases include 'search the codebase', 'find code related to', 'update the…
openlore
Query and publish to an OpenLore knowledge repository over SSH using ordinary shell commands. Use when a task needs project documentation, runbooks, shared team knowledge, or a place to publish findings.
system-monitor
Monitor system health: CPU, memory, disk, processes, and network. Alert on thresholds via any channel.