Token导航 LogoToken导航

Agent Skill · 研究检索

multi-brain(多脑)

multi-brain 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。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。

multi-brain 是研究检索类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为只读;许可证信息还不明确,第三方扫描有未通过项,按安装量在研究检索类的 73,675 个 Skill 里排第 64,068。

总安装 306周安装 13GitHub Stars 4下载 107核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
unknown

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(多脑)
来源仓库:https://github.com/fatih-developer/fth-skills
仓库路径:skills/multi-brain
安装命令:
npx skills add https://github.com/fatih-developer/fth-skills --skill multi-brain
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

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

multi-brain 用于查找、检索和筛选相关信息。适合在 Codex、Claude、Cursor、Gemini CLI 中根据关键词、任务场景或来源线索快速定位候选结果时使用。支持基于语义匹配、标签过滤和上下文相关性进行智能内容检索与排序。

安装命令为 npx skills add https://github.com/fatih-developer/fth-skills --skill multi-brain。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。

来源仓库内路径:skills/multi-brain

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 Codex、Claude、Cursor、Gemini CLI;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 npx skills add https://github.com/fatih-developer/fth-skills --skill multi-brain;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它标注为只读,先用只读任务试一次,确认行为符合预期再扩大使用范围。

常见问题

6 条

multi-brain 怎么安装?

在终端执行 npx skills add https://github.com/fatih-developer/fth-skills --skill multi-brain。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

multi-brain 支持哪些 AI 工具?

公开资料标注支持 Codex、Claude、Cursor、Gemini CLI。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

multi-brain 需要什么权限?

权限级别是只读。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。

multi-brain 是免费的吗?

公开资料没有标注明确许可证,收费和商用限制都无法确认。正式使用前请到来源仓库核对授权说明。

multi-brain 用的人多吗?

公开数据是累计安装 306 次,近一周安装 13 次,下载 107 次,GitHub 4 星。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

multi-brain 安全吗?

第三方扫描有未通过项,风险较高。除非你确认过源码,否则不建议直接安装到常用环境。

值得选的地方和要先想清楚的地方

值得选的地方

  • 同时适配 4 种工具,换客户端不用重新找替代品。
  • 标注为只读权限,试用阶段不容易误改本地文件。

先想清楚的地方

  • 许可证没有明确标注,能不能商用、能不能再分发都不确定。
  • 第三方扫描存在提醒项或未通过项,需要人工逐条确认。
  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 Codex,想直接扩展它能力的人。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。
  • 只想先试试效果、不希望它改动本地文件的人。

不太适合

  • 需要明确授权才能用到公司项目里的场景。
  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
multi-brain

多脑 · 本页

3064只读未明确未通过

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

Multi-Brain Consensus Protocol

Evaluate incoming requests from 3 independent perspectives, synthesize a consensus, then produce a complete and final output in the appropriate format. This is not just "decide" — it is "decide and deliver."


Workflow

1. Understand the request
2. 3 Perspectives → Consensus
3. Determine output format
4. Produce full output

Step 1: Understand the Request

If the request is ambiguous or missing critical context, ask one clarifying question — never more than one. If the request is clear, proceed directly to Step 2.


Step 2: Three Perspectives

Each instance works independently — none sees the other's reasoning. Each summarizes its approach and rationale in 2–3 sentences.

Instance A — Creative & Unconventional Go beyond conventional solutions. Seek the least expected but potentially most impactful approach. Take calculated risks, but justify them clearly.

Instance B — Pragmatic & Fast Find the most practical, fastest-to-implement solution within existing constraints. Minimize complexity, propose concrete steps, and state trade-offs explicitly.

Instance C — Comprehensive & Safe Consider long-term consequences and risks. Identify edge cases, side effects, and missing information. Prioritize sustainability and resilience.


Step 3: Consensus

Synthesize the three perspectives:

  • Agreement points: If two or three instances converge, this is likely the right path.
  • Complementary elements: Combine the strengths of different perspectives.
  • Conflicts: Which argument is stronger? Why?

Step 4: Determine Output Format

Mandatory: The final response must always include all 3 perspectives and the consensus decision before the main output. Never skip or collapse them — the user must see the reasoning trail.

If the request or context already implies a format, use it. If not, ask the user:

"Based on the consensus, how should I proceed — a detailed report, working code, or a brief summary?"

Format Options

Report / Analysis Document When the request involves research, decision-making, or strategy:

  • Produce as a Markdown document (offer to save).
  • Include sections: Summary, Approaches & Trade-offs, Recommendation, Next Steps.
  • Write thoroughly — as if the user will share it with stakeholders.

Code When the request involves implementation:

  • Apply the architecture/approach from the consensus.
  • Write working, testable code.
  • Save files and present them to the user.
  • Explain "why this approach" in code comments.

Brief Summary When the user wants a quick answer or it is a simple decision:

  • Single paragraph: chosen approach + rationale + next step.

Output Template

Use references/OUTPUT_TEMPLATE.md for the standard response structure.


When to Skip

Do not start the brainstorm process — respond directly when:

  • The question has a single factual answer ("How do I iterate a list in Python?").
  • The user explicitly asks for a quick/short answer.
  • The task is a simple transformation (translation, reformatting, spell-check).
  • The user has already decided and only wants execution.

See references/SKIP_CONDITIONS.md for the full decision matrix.


Examples

See references/EXAMPLES.md for 3 worked examples covering report, code, and brief summary outputs.


Guardrails

  • Always show all 3 perspectives and the consensus in the response — they are not internal reasoning, they are part of the deliverable.
  • Each instance must reason independently — no cross-contamination.
  • Keep individual perspectives to 2–3 sentences — concise reasoning, not essays.
  • Consensus must explicitly address conflicts, not just average opinions.
  • The final output must be complete and ready to use — not a stub or outline.
  • Prefer the pragmatic path when perspectives are equally strong.

Templates

  • Use templates/brainstorm-report.md.tmpl for report/analysis outputs.
  • Use templates/brainstorm-brief.md.tmpl for quick decision responses.

平台分布

Codex

33.82%
按下载量换算36

Claude

29.15%
按下载量换算31

Cursor

18.94%
按下载量换算20

Gemini CLI

9.65%
按下载量换算10

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

未通过

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。来源安全扫描存在提醒或未通过项,安装前请自行复核。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills