Token导航 LogoToken导航TokenDH.com
待分类只读github未标认证来源可访问许可证需确认审计通过

multi-brain-debate多脑辩论

Agent Skill

multi-brain-debate 用于处理 GitHub 仓库、Issue、Pull Request 和代码协作信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要围绕仓库状态、代码变更或协作事项进行整理时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

总安装

384

周安装

16

GitHub Stars

4

下载量

128
CodexClaudeCursorGemini CLI

安装说明

本站只整理中文说明和来源信息,不托管安装包,也不代用户安装。

GitHub

来源数

2

许可证

unknown

最后核验

2026-05-01

来源状态

来源可访问

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:multi-brain-debate(多脑辩论)
来源仓库:https://github.com/fatih-developer/fth-skills
仓库路径:skills/multi-brain-debate
安装命令:
npx skills add https://github.com/fatih-developer/fth-skills --skill multi-brain-debate
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/fatih-developer/fth-skills --skill multi-brain-debate

简介

multi-brain-debate 用于处理 GitHub 仓库、Issue、Pull Request 和代码协作信息。

  • 适合在 Codex、Claude、Cursor、Gemini CLI 中围绕仓库状态、代码变更或协作事项进行整理时使用。
  • 通过 GitHub API 获取仓库元数据、分支信息和协作动态,支持多 Agent 环境下的协同开发场景。
  • 安装命令为 npx skills add https://github.com/fatih-developer/fth-skills --skill multi-brain-debate。
  • 使用前需确认权限范围、维护状态,注意可能触发联网、命令执行或文件读写操作。

SKILL.md

Multi-Brain Debate Protocol

Extend the multi-brain consensus with a two-round adversarial debate. Perspectives don't just state their case — they challenge each other. The result is a stress-tested decision where weak arguments have been exposed and strong ones reinforced.


Workflow

1. Understand the request
2. Round 1: Independent positions (3 perspectives)
3. Round 2: Counter-arguments and rebuttals
4. Judge's verdict (consensus)
5. Produce full output with debate trail visible

Step 1: Understand the Request

Same as base multi-brain. Ask one clarifying question if needed, otherwise proceed.


Step 2: Round 1 — Opening Positions

Each instance presents their approach independently (same as base multi-brain):

## 🧠 Debate — Round 1: Opening Positions

**Instance A — Creative:**
[2-3 sentences: position + rationale]

**Instance B — Pragmatic:**
[2-3 sentences: position + rationale]

**Instance C — Comprehensive:**
[2-3 sentences: position + rationale]

Step 3: Round 2 — Challenges & Rebuttals

Each instance can now see the others' positions and must:

  1. Challenge the weakest point of another instance's argument
  2. Defend their own position against potential objections
## ⚔️ Debate — Round 2: Challenges

**A challenges B:**
[1-2 sentences: specific weakness identified]

**B challenges C:**
[1-2 sentences: specific weakness identified]

**C challenges A:**
[1-2 sentences: specific weakness identified]

**Rebuttals:**
- **A responds:** [1 sentence defense or concession]
- **B responds:** [1 sentence defense or concession]
- **C responds:** [1 sentence defense or concession]

Step 4: Judge's Verdict

After the debate, synthesize the strongest surviving arguments:

## ⚖️ Verdict

**Winner:** [Which perspective's core argument survived the debate]
**Incorporated from others:** [Elements from losing arguments that strengthen the decision]
**Eliminated:** [Arguments that were successfully challenged and dropped]

Step 5: Full Output

Mandatory: The final response must include both debate rounds, the verdict, and the complete deliverable. The user must see the full reasoning trail.


When to Use Debate vs Base Multi-Brain

SituationUse
High-stakes architecture decisionDebate
Choosing between competing technologiesDebate
Quick implementation questionBase multi-brain
Strategy with long-term consequencesDebate
Simple feature decisionBase multi-brain
Security-sensitive designDebate

Guardrails

  • Always show both rounds — the debate trail is the value, not just the verdict.
  • Challenges must be specific and substantive — not generic "this might not scale."
  • Rebuttals can include concessions — "You're right, I'll adjust my position to X."
  • The verdict must explain what was eliminated and why — not just what won.
  • Keep the total debate concise: Round 1 (2-3 sentences each), Round 2 (1-2 sentences each), Rebuttals (1 sentence each).
  • Do not force disagreement — if all 3 genuinely align, acknowledge it and skip Round 2.

References

  • See references/EXAMPLES.md for worked debate examples.

适合场景

01

用户想查找某类 Agent Skill 时

02

需要根据任务场景推荐可安装能力包时

03

需要对比不同来源的安装命令和来源信息时

能力概览

能力 1

按任务关键词查找相关 Skills

能力 2

展示可复制的安装命令

能力 3

保留来源站点、仓库和原始说明,方便继续核验

能力 4

展示第三方安全扫描或审计结果

安装后应在对应宿主中按原始 README 的触发条件使用;具体调用方式请以来源页面和 README 为准。

平台分布

Codex

36.1%
按下载量换算46

Claude

28.99%
按下载量换算37

Cursor

22.18%
按下载量换算28

Gemini CLI

10.18%
按下载量换算13

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

只读

该 Skill 主要提供规则、说明或参考内容,本身偏只读;真正读写文件、联网或执行命令仍取决于宿主 Agent 的任务。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。当前只有一个来源,正式发布前建议补源仓库或其他目录站核验。

来源信息

继续浏览同类 Skills