reading-squad
NewUse when you want to simulate a multi-expert book discussion squad inside any AI tool. Covers setup, member definitions, multi-round discussion flow, and consensus synthesis. No platform dependencies — works in Claude Code, Codex, Hermes, or any chat-based AI.
Overview
Reading Squad
Overview
Reading Squad is a lightweight, platform-agnostic method to simulate a multi-expert book discussion inside any AI tool. You define a squad leader and a roster of expert members, then run structured, multi-round discussions on any book-related topic. The leader coordinates the debate, and the final output is a set of key consensus points and actionable reading guidance.
When to Use
- •You want diverse perspectives on a book, chapter, or reading topic.
- •You want structured debate rather than a single AI monologue.
- •You want actionable takeaways synthesized from multiple viewpoints.
- •You are using Claude Code, Codex, Hermes, or any chat-based AI — no special platform required.
When NOT to use:
- •You only need a quick summary of a single book.
- •You want a single expert opinion rather than a panel debate.
Core Pattern
Setup → Discuss (5+ rounds) → Synthesize1. Setup
Define the squad in a single prompt block:
你现在是【读书天团】的队长。
你的队员有:
1. 罗振宇(跨学科连接)
2. 刘润(商业逻辑)
3. 樊登(书籍解读)
4. 吴晓波(历史宏观)
5. 李笑来(概念拆解)
6. 俞敏洪(成长教育)
7. 尹烨(生命科学)
当用户抛出一个读书话题后,你需要:
1. 让每位队员依次发表第一轮观点(每人1-2段)
2. 组织至少5轮讨论,每轮围绕一个子议题展开
3. 在讨论中引导队员之间产生观点碰撞
4. 最后汇总形成3-5条关键共识和具体阅读建议2. Discussion Flow
The leader (AI) should follow this round structure:
| Round | Purpose | Example Prompt to Self |
|---|---|---|
| 1 | Initial positions | "请每位队员用1-2段话表达对本书的初印象" |
| 2 | Deep dive 1 | "罗振宇,请从跨学科角度拆解这本书的核心逻辑" |
| 3 | Deep dive 2 | "刘润,这本书的商业洞察对你有什么启发?" |
| 4 | Debate / Clash | "樊登和吴晓波的观点似乎有分歧,请各自补充" |
| 5 | Synthesis | "基于以上讨论,请汇总3-5条关键共识" |
3. Synthesis
After 5+ rounds, the leader outputs:
- Key Consensus (3-5 bullet points)
- Actionable Reading Guidance (specific next steps)
- Recommended Follow-up Topics (optional)
Quick Reference
Minimal Prompt Template
你现在是【读书天团】队长,队员包括:罗振宇(跨学科连接)、
刘润(商业逻辑)、樊登(书籍解读)、吴晓波(历史宏观)、
李笑来(概念拆解)、俞敏洪(成长教育)、尹烨(生命科学)。
用户话题:{TOPIC}
请组织至少5轮讨论,每轮聚焦一个子议题,引导队员观点碰撞,
最后输出3-5条关键共识和具体阅读建议。Customizing Members
Replace or add members by following this pattern:
姓名(擅长领域)Examples:
- •张三(心理学应用)
- •李四(技术架构)
- •王五(文学创作)
Common Mistakes
| Mistake | Fix |
|---|---|
| Only 1-2 rounds of discussion | Enforce at least 5 rounds explicitly |
| Members agree too quickly | Prompt for disagreements and counter-arguments |
| Output is just a summary | Require "key consensus" and "actionable guidance" as separate sections |
| Leader dominates the conversation | Leader should mainly ask questions and synthesize, not opine |
| Members sound the same | Give each member a distinct voice/perspective in the setup |
Tips for Best Results
- •Be specific with the topic: "《三体》中的黑暗森林法则" beats "聊聊科幻".
- •Encourage conflict: Explicitly ask the leader to surface disagreements.
- •Limit round length: Cap each member's contribution to 1-2 paragraphs to keep focus.
- •Iterate: If the first synthesis feels shallow, ask for a "devil's advocate" round.
Example Output Structure
## 讨论话题:{TOPIC}
### 第一轮:初印象
[罗振宇] ...
[刘润] ...
...
### 第二轮:深度拆解
...
### 第五轮:共识汇总
#### 关键共识(3-5条)
1. ...
2. ...
#### 具体阅读建议
1. ...
2. ...Install & Usage
mkdir -p .claude/skillsmkdir -p .claude/skills && curl -o .claude/skills/reading-squad.md https://raw.githubusercontent.com/jinguo0524/reading-squad/main/SKILL.md/reading-squadFrequently Asked Questions
What is reading-squad?
Use when you want to simulate a multi-expert book discussion squad inside any AI tool. Covers setup, member definitions, multi-round discussion flow, and consensus synthesis. No platform dependencies — works in Claude Code, Codex, Hermes, or any chat-based AI.
How to install reading-squad?
To install reading-squad, create the .claude/skills directory in your project, then run the curl command to download the skill file. Once installed, invoke it in Claude Code with /reading-squad.
What is reading-squad best for?
reading-squad is a community categorized under General. Created by jinguo0524.