book-analysis

Book Analysis — One-click deep learning of a book

Enter a book title, automatically collect information across six dimensions in parallel, and after cross-analysis output deep learning notes.

When to Use

Triggered when the user requests the following operations:

"Analyze The Lychee of Chang'an" / "Learn Sapiens"
"Help me read Poor Charlie's Almanack" / "Reading notes for The Three-Body Problem"
"book analysis: Thinking, Fast and Slow"
"What is the book XX about" / "Is XX worth reading"

Phase 0: Parse Input

Based on the user input, identify the following information:

Book Title
- Extract the book title from the user message (remove book title marks 《》 or quotation marks)
- If the user gives a vague description, first execute the "Book Locating Process" below to confirm the full book title

Book Locating Process (must be executed, cannot be skipped)

Step 1: Douban two-step method (preferred for Chinese books, highest success rate)
```
Step 1: WebFetch https://m.douban.com/search/?query={书名}&type=book
        → 从返回页面中找 /book/subject/{数字ID}/ 格式的链接，提取 Subject ID

Step 2: WebFetch https://book.douban.com/subject/{Subject ID}/
        → 从详情页获取：准确书名、作者、出版社、出版日期、ISBN、内容简介、书籍分类
```
After success, go directly to Step 3,no other searches are needed。

Step 2: Dangdang verification (used when Douban fails or cross-confirmation is needed)
```
WebFetch https://search.dangdang.com/?key={书名}&act=input
→ 获取 ISBN 和出版信息，与豆瓣数据交叉核对
```
Step 3: Vague book title handling (used when the book title is uncertain)
- Use quotation marks to exactly match the core term:WebSearch ""{书名中最独特的2-3字}" 豆瓣图书作者"
- Or search:WebSearch {书名} 书名作者出版社
- After confirming the book title, execute the Douban two-step method from Step 1
Not recommended for Chinese books: Google Books API, Open Library (very little coverage of Chinese books, basically ineffective)
Determine Key Variables(needed by all subsequent Agents)
- {book_name}: full book title
- {author}: author name
- {book_type}: book type
- {type_hint}: type adaptation prompt (see below)
- {date}: current date, format YYYY-MM-DD
- {book_name_en}: English or pinyin form of the book title (lowercase, hyphen-separated), used for file naming
Type Adaptation Prompts(passed to each subagent)
- Novel/Literature → "This is a literary work; focus on analyzing narrative techniques, character development, and thematic metaphors. The 'solutions' dimension should focus on thematic insights and life reflections."
- Non-fiction/Business → "This is a non-fiction work; focus on analyzing arguments and evidence, methodology, and actionable recommendations."
- Technology → "This is a technical book; focus on analyzing technical concepts, practical methods, and applicable scenarios for code examples."
- History → "This is a historical work; focus on analyzing historical background, causal chains, and insights for the present."
- Philosophy → "This is a philosophical work; focus on analyzing core propositions, argumentative logic, and relationships between schools of thought."
- Popular Science → "This is a popular science work; focus on analyzing scientific principles, research methods, and impact on public understanding."
- Other → "Please determine the analysis focus yourself based on the book's content."

If the book type or author cannot be determined, first use one WebSearch for a quick query,do not skip this step。

Phase 1: Parallel Data Collection (6 Subagents)

Key Requirements: The following 6 Tasks must be placed inthe same messageto ensure parallel execution.

Each Agent uses subagent_type: "general-purpose"。

Agent 1: Chapter Structure and Logical Relationships

description: "Analyze the chapter structure of {book_name}"

prompt template:

你是书籍结构分析专家。请分析《{book_name}》（{author}）的章节结构和内在逻辑关系。

类型提示：{type_hint}

任务:
1. 使用 WebSearch 搜索以下关键词（至少搜 2 次不同关键词）:
   - "{book_name} 目录 章节"
   - "{book_name} 结构分析" 或 "{book_name} table of contents"
2. 使用 WebFetch 访问至少 1 个包含详细目录的页面，验证章节信息
3. 整理以下信息:

   **章节概览:**
   - 列出所有章节（或主要部分），每章附一句话概要
   - 如果章节过多（>20章），可按"部/篇"分组概括

   **逻辑关系:**
   - 章节之间的逻辑关系类型（递进、并列、因果、转折、总分）
   - 全书叙事/论证主线（一段话描述）
   - 关键转折点或高潮章节

输出要求:
- 章节列表用「章节号 | 标题 | 概要」格式
- 逻辑关系用文字描述，标明关系类型
- 控制在 600 字以内
- 全部使用中文输出
- **标注信息来源**: 注明来源网站名称及原始链接

Agent 2: Background and Key Points

description: "Research the background and core viewpoints of {book_name}"

prompt template:

你是书籍研究专家。请研究《{book_name}》（{author}）的创作背景和核心要点。

类型提示：{type_hint}

任务:
1. 使用 WebSearch 搜索以下关键词（至少搜 2 次不同关键词）:
   - "{author} {book_name} 创作背景"
   - "{book_name} 核心观点" 或 "{book_name} 书评 要点"
2. 使用 WebFetch 访问至少 1 条高质量书评或访谈原文，验证内容
3. 整理以下信息:

   **作者与创作背景:**
   - 作者简介（生平、代表作、专业领域）
   - 写作动机与创作契机
   - 时代/社会背景（出版年代的相关背景）

   **核心要点:**
   - 提炼 3-5 个核心要点/论点
   - 每个要点附 1-2 句展开说明
   - 标注该要点主要体现在哪些章节

输出要求:
- 作者简介控制在 100 字以内
- 核心要点用编号列表
- 控制在 600 字以内
- 全部使用中文输出
- **标注信息来源**: 注明来源网站名称及原始链接

Agent 3: Issues and Impact

description: "Analyze the issues and impact raised by {book_name}"

prompt template:

你是文化评论专家。请分析《{book_name}》（{author}）提出或揭示的关键问题及其影响。

类型提示：{type_hint}

任务:
1. 使用 WebSearch 搜索以下关键词（至少搜 2 次不同关键词）:
   - "{book_name} 问题 思考"
   - "{book_name} 影响 意义" 或 "{book_name} 书评 评价"
2. 使用 WebFetch 访问至少 1 条深度书评或学术评论，验证观点
3. 整理以下信息:

   **关键问题:**
   - 书中提出或揭示的 3-5 个核心问题
   - 每个问题的影响范围（个人/行业/社会）
   - 这些问题在当下是否仍然有现实意义

   **社会反响与影响:**
   - 出版后的社会讨论和反响
   - 获奖情况、销量、翻译语种等影响力指标
   - 争议观点（如有）

输出要求:
- 问题按重要性排序
- 影响分析要具体，避免空泛
- 控制在 600 字以内
- 全部使用中文输出
- **标注信息来源**: 注明来源网站名称及原始链接

Agent 4: Solutions and Implementation Recommendations

description: "Extract the insights and practical recommendations of {book_name}"

prompt template:

你是知识应用专家。请提取《{book_name}》（{author}）中的解决方案、启示和可实践的建议。

类型提示：{type_hint}

任务:
1. 使用 WebSearch 搜索以下关键词（至少搜 2 次不同关键词）:
   - "{book_name} 启示 收获"
   - "{book_name} 读后感 实践" 或 "{book_name} takeaways"
2. 使用 WebFetch 访问至少 1 篇高质量读后感或实践分享，验证内容
3. 整理以下信息:

   **书中的思路/方案:**
   - 书中针对核心问题提出的思路或方案
   - 如果是文学作品，提取主题层面的启示和人生思考
   - 方法论或思维框架（如有）

   **读者行动建议:**
   - 3-5 条具体可操作的实践建议
   - 每条建议注明适用场景和局限性
   - 按难易程度排序（从易到难）

输出要求:
- 方案和建议要具体可操作，避免泛泛而谈
- 文学作品侧重思想启示，非虚构侧重方法论
- 控制在 600 字以内
- 全部使用中文输出
- **标注信息来源**: 注明来源网站名称及原始链接

Agent 5: Terms and Index

description: "Compile a key term index for {book_name}"

prompt template:

你是知识整理专家。请整理《{book_name}》（{author}）中的关键术语和重要概念索引。

类型提示：{type_hint}

任务:
1. 使用 WebSearch 搜索以下关键词（至少搜 2 次不同关键词）:
   - "{book_name} 关键概念 术语"
   - "{book_name} 名词解释" 或 "{book_name} glossary key concepts"
2. 使用 WebFetch 访问至少 1 个包含术语解释的页面，验证定义准确性
3. 整理以下信息:

   **术语索引表:**
   - 列出书中 10-20 个关键术语/概念/人名/地名
   - 每个条目包含: 术语名称、简要定义（1-2句）、所在章节
   - 按首次出现的章节顺序排列

   **核心概念关联:**
   - 标注哪些术语之间存在关联（如: A是B的前提，C是D的应用）

输出要求:
- 术语表格式: 「术语 | 定义 | 出现章节」
- 定义简洁精确
- 控制在 600 字以内
- 全部使用中文输出
- **标注信息来源**: 注明来源网站名称及原始链接

Agent 6: Further Reading

description: "Recommend further reading related to {book_name}"

prompt template:

你是阅读顾问。请为《{book_name}》（{author}）推荐延伸阅读书目。

类型提示：{type_hint}

任务:
1. 使用 WebSearch 搜索以下关键词（至少搜 2 次不同关键词）:
   - "{book_name} 类似书籍 推荐"
   - "{author} 其他作品" 或 "{book_name} similar books recommended"
2. 使用 WebFetch 访问至少 1 个书单推荐页面，验证推荐理由
3. 整理以下信息:

   **同作者作品:**
   - 列出作者的其他代表作（2-3本），附一句话说明与本书的关联

   **同主题推荐:**
   - 5-6 本主题相关的书籍
   - 每本包含: 书名、作者、一句话关联说明（为什么推荐、与本书的关系）
   - 涵盖不同难度和视角

输出要求:
- 推荐理由要说明与本书的具体关联，不能泛泛说"也很好"
- 优先推荐经典作品和高评分作品
- 控制在 500 字以内
- 全部使用中文输出
- **标注信息来源**: 注明来源网站名称及原始链接

Phase 2: Comprehensive Analysis (Main Thread)

After all 6 Agents have returned, complete the following analysis in the main thread.

Step 1: Information Aggregation

Integrate the results of the 6 Agents and identify:

Between dimensions, consistency signals(e.g., multiple sources mentioning the same core viewpoint = high confidence)
Between dimensions, complementary relationships(e.g., the key points mentioned by Agent 2 can correspond to the chapter structure from Agent 1)
Between dimensions, contradictory signals(e.g., different sources having completely different evaluations of a viewpoint = controversy needs to be flagged)
At the same time, compile allsource URLsreturned by each Agent, and aggregate them into the source list for the "References" section at the end of the report

Step 2: Cross-Linking

Establish links between the 6 dimensions:

Key Points <-> Chapter Mapping: Annotate the core key points (Agent 2) to the corresponding chapters (Agent 1)
Issues <-> Solutions Correspondence: Match issues (Agent 3) one-to-one with solutions (Agent 4)
Terms <-> Content Anchoring: Verify that the chapter annotations in the term index (Agent 5) are consistent with the actual chapter structure (Agent 1)
Further Reading Associations: Associate the recommended books (Agent 6) with the core themes of this book

Step 3: Distilling Core Conclusions

Based on the above analysis, write a 3-5 sentence core conclusion:

What is the most core value of this book
The 1-2 viewpoints most worth paying attention to
What kind of readers it is most suitable for

Phase 3: Output Report

Output the final report in the following Markdown format:

# 《{book_name}》深度学习笔记

> **作者**：{author} | **类型**：{book_type} | **分析日期**：{date}

## 核心结论

{3-5 句话总结这本书最值得了解的内容，包含核心价值、关键观点和适合的读者}

---

## 一、章节结构与逻辑关系

### 章节概览

| 章节 | 标题 | 概要 |
|------|------|------|
| 第1章 | {标题} | {一句话概要} |
| 第2章 | {标题} | {一句话概要} |
| ... | ... | ... |

### 逻辑主线

{描述全书的叙事/论证脉络，标明章节间的逻辑关系类型（递进、因果、并列、转折等）}

---

## 二、背景与要点

### 作者与创作背景

{作者简介、写作动机、时代/社会背景}

### 核心要点

1. **{要点一}**：{展开说明}（见第X章）
2. **{要点二}**：{展开说明}（见第X章）
3. **{要点三}**：{展开说明}（见第X章）
...

---

## 三、问题与影响

### 关键问题

1. **{问题一}**：{问题描述} → 影响范围：{个人/行业/社会}
2. **{问题二}**：{问题描述} → 影响范围：{个人/行业/社会}
3. ...

### 社会反响

{出版后的讨论、获奖、争议等}

---

## 四、解决方案与实施建议

### 书中的思路/启示

{书中针对核心问题的思路、方法论或主题启示}

### 读者行动建议

1. **{建议一}**（难度：⭐）：{具体描述} — 适用场景：{场景}
2. **{建议二}**（难度：⭐⭐）：{具体描述} — 适用场景：{场景}
3. **{建议三}**（难度：⭐⭐⭐）：{具体描述} — 适用场景：{场景}
...

---

## 五、术语与索引

| 术语 | 定义 | 出现章节 |
|------|------|----------|
| {术语1} | {简要定义} | 第X章 |
| {术语2} | {简要定义} | 第X章 |
| ... | ... | ... |

---

## 六、延伸阅读

| 书名 | 作者 | 关联说明 |
|------|------|----------|
| 《{书名1}》 | {作者} | {与本书的具体关联} |
| 《{书名2}》 | {作者} | {与本书的具体关联} |
| ... | ... | ... |

---

## 参考来源

- **章节结构**: [{来源标题}]({url}), ...
- **背景要点**: [{来源标题}]({url}), ...
- **问题影响**: [{来源标题}]({url}), ...
- **方案建议**: [{来源标题}]({url}), ...
- **术语索引**: [{来源标题}]({url}), ...
- **延伸阅读**: [{来源标题}]({url}), ...

---

*本笔记由 AI 自动生成，建议结合原书阅读。*

Markdown filling rules:

Replace the placeholders in the template with actual data
Annotate core key points with their corresponding chapters for easy reference back to the original book
Arrange the glossary in order of first appearance
Arrange reader action recommendations from easy to difficult
The "References" section is required: Fill in all source links returned by the 6 Agents, grouped by dimension; do not omit them, and do not use placeholder URLs

Save the report as a file: ./{book_name_en}-book-analysis-{date}.md, saved in the current working directory.

Filename Rules: {book_name_en} Use the English or pinyin form of the book title (lowercase, hyphen-separated), for example:

"The Lychees of Chang'an" → changan-lychee
"Sapiens" → sapiens
"The Three-Body Problem" → three-body
Thinking, Fast and Slow → thinking-fast-and-slow

Error Handling

Agent timeout or failure: If an Agent does not return a result, mark that dimension in the report as "insufficient data, readers are advised to consult it themselves," and analyze the remaining dimensions as usual
Book title cannot be identified: Prompt the user to confirm the book title, and fuzzy matching suggestions may be provided
Insufficient information for niche/new books: If search results are very few, merge related dimensions and reduce repeated "insufficient information" prompts
Multiple versions/multiple translations: If the same book has multiple versions or translations, prioritize analyzing the most well-known/latest version, and note the version information

Notes

The output of each Agent is strictly limited in word count to prevent the main thread context from overflowing
The comprehensive analysis stage focuses on cross-correlation, rather than simply concatenating the results of the 6 dimensions
The chapter annotations in the terminology index must be consistent with the chapter structure of Agent 1
The core conclusion should help readers judge "whether this book is worth my time reading"
All output uses Chinese

About this skill

When to Use

Phase 0: Parse Input

Phase 1: Parallel Data Collection (6 Subagents)

Agent 1: Chapter Structure and Logical Relationships

Agent 2: Background and Key Points

Agent 3: Issues and Impact

Agent 4: Solutions and Implementation Recommendations

Agent 5: Terms and Index

Agent 6: Further Reading

Phase 2: Comprehensive Analysis (Main Thread)

Step 1: Information Aggregation

Step 2: Cross-Linking

Step 3: Distilling Core Conclusions

Phase 3: Output Report

Error Handling

Notes