Token导航 LogoToken导航

Agent Skill · 研究检索

council-of-five(五人理事会)

council-of-five 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。

council-of-five 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,权限为需要联网;许可证为 MIT,暂无第三方扫描结果,按安装量在研究检索类的 73,675 个 Skill 里排第 41,003。

总安装 1,229周安装 50GitHub Stars 公开资料未说明下载 521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
MIT

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:council-of-five(五人理事会)
来源仓库:https://github.com/johnlindquist/claude
仓库路径:skills/council-of-five
安装命令:
npx skills add johnlindquist/claude --skill "council-of-five"
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。

AgentSkills.tonpx skills
npx skills add johnlindquist/claude --skill "council-of-five"

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。

发现并安装 AI 代理的技能。适用于查找、检索和筛选相关信息的场景。通过 github 安装,支持 Codex、Claude、Cursor、Gemini CLI 宿主环境。安装前建议确认权限范围和维护状态,注意可能触发联网或命令执行操作。具体用法请结合来源仓库和原始 README 文档进一步核验。

来源仓库内路径:skills/council-of-five

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 johnlindquist/claude --skill "council-of-five";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

    它需要需要联网,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

council-of-five 怎么安装?

在终端执行 npx skills add johnlindquist/claude --skill "council-of-five"。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

council-of-five 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。

council-of-five 需要什么权限?

权限级别是需要联网。该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

council-of-five 是免费的吗?

来源标注的许可证是 MIT。本站不代收任何费用,也不托管安装包;能否商用、要不要署名,以许可证原文为准。

council-of-five 用的人多吗?

公开数据是累计安装 1,229 次,近一周安装 50 次,下载 521 次。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。

council-of-five 安全吗?

没有查到第三方扫描结果,安全性无法从公开资料判断,安装前请自行审查源码、依赖和命令行为。

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。
  • 许可证明确为 MIT,商用和分发边界好判断。

先想清楚的地方

  • 没有第三方扫描结果,安全性只能靠自己读源码判断。
  • 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。

不太适合

  • 要求有安全扫描结论才肯安装的团队。
  • 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
council-of-five

五人理事会 · 本页

1,229—需要联网MIT未收录

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

name
council-of-five
description
Spawn 5 Opus subagents with randomly-generated distinct personas to debate a problem from multiple angles. Use when exploring UX decisions, architecture choices, or any decision that benefits from diverse perspectives arguing creatively.

Council of Five

Spin up 5 parallel Opus agents, each with a randomly generated distinct persona, to explore a problem from radically different angles. They argue, then you synthesize.

When to Use

  • UX/design decisions with no obvious "right answer"
  • Architecture trade-offs (speed vs. maintainability, etc.)
  • Naming conventions, API design, workflow design
  • Any decision where "outside the box" thinking helps

Random Persona Generation

Each invocation generates 5 fresh personas. Pick randomly from diverse archetypes:

Archetype Pool (sample, not exhaustive)

CategoryExample Personas
ReductionThe Minimalist, The Deletionist, The "YAGNI" Zealot, The Haiku Master
NarrativeThe Storyteller, The Novelist, The Stand-up Comic, The Documentary Filmmaker
VisualThe Dashboard Engineer, The Infographic Designer, The Color Theorist, The Whitespace Monk
VerificationThe Paranoid Auditor, The Penetration Tester, The QA Gremlin, The "Trust No One" Agent
BehaviorThe UX Researcher, The Cognitive Psychologist, The Lazy User Simulator, The Angry Customer
PerformanceThe Latency Hunter, The Memory Miser, The Big-O Obsessive, The Cache Whisperer
AccessibilityThe Screen Reader Advocate, The Color Blind Designer, The Keyboard-Only Navigator
PhilosophyThe Unix Philosopher, The Functional Purist, The "Worse is Better" Advocate, The Pragmatist
ChaosThe Edge Case Finder, The Chaos Monkey, The "What If" Catastrophist, The Entropy Embracer
HistoryThe Legacy Code Archaeologist, The "We Tried That" Historian, The Pattern Recognizer
FutureThe 10x Scale Predictor, The Deprecation Prophet, The "Your Future Self" Advocate
OutsiderThe New Hire, The Non-Technical Stakeholder, The Customer Support Rep, The Intern

Generation Rules

  1. Pick 5 personas from different categories (no two from same category)
  2. Invent new ones if the pool feels stale—creativity encouraged
  3. Name them vividly (e.g., "The Haiku Master" not "The Brevity Person")
  4. Give each a one-sentence philosophy before they argue

Invocation

Basic

I need a council of five to debate [your problem/decision]

With Context

Run a council-of-five on this UX problem:
[paste the current approach]

What I care about: [your priority, e.g., "clarity over completeness"]

Seeded with a specific angle

Council of five, but make sure one persona is security-focused

How It Works

  1. Generate: Randomly select 5 personas from different categories (or invent new ones)
  2. Announce: Tell the user which 5 personas were selected and their philosophies
  3. Launch: 5 Opus subagents spin up in parallel (background tasks)
  4. Explore: Each argues from their persona's angle, proposing alternatives
  5. Synthesize: Results are gathered and presented as a comparison table
  6. Decide: User picks an approach (or hybrid)

Prompt Template for Each Agent

You are [RANDOMLY GENERATED PERSONA NAME].

Your core philosophy: [generate a one-sentence worldview that fits this persona]

The user is evaluating: [problem description]

Current approach:
[paste current solution]

Your task:
1. Critique the current approach from your unique angle
2. Propose an alternative that embodies your philosophy
3. Give a concrete example of your approach in action
4. Acknowledge one weakness of your approach

Be creative. Be opinionated. Argue your position strongly.
Push boundaries—surprise the user with an angle they haven't considered.

Output Format

After all agents complete, present:

  1. Quick comparison table (persona | core argument | proposed format)
  2. Synthesis (where they agree, where they clash)
  3. Hybrid suggestion (if applicable)
  4. Question to user: "What resonates?"

Example Session

User: "Council of five on my error message format"

Generated personas:

  1. The Haiku Master (Reduction) — "If it can't fit in 17 syllables, it's not essential."
  2. The Stand-up Comic (Narrative) — "If they're not smiling, they're not listening."
  3. The Color Theorist (Visual) — "Meaning lives in hue and contrast, not words."
  4. The Angry Customer (Behavior) — "I'm already frustrated. Don't make it worse."
  5. The Chaos Monkey (Chaos) — "What happens when this fails at 3am on a holiday?"

Result:

PersonaArgumentProposal
Haiku Master"Too many words"Auth timed out. / Server took too long. / Try once more?
Stand-up Comic"Errors are traumatic""Well, that didn't work. The server ghosted us. Retry?"
Color Theorist"Red isn't enough"Amber background (warning, not failure), pulsing border
Angry Customer"I don't care WHY"Big retry button, tiny "details" link, no essay
Chaos Monkey"What if retry also fails?"Exponential backoff + "Contact support" after 3 attempts

Tips

  • Be specific about what you care about (speed? clarity? memorability?)
  • Provide the current approach so personas can critique something concrete
  • Ask for hybrids if multiple approaches resonate
  • Request a reroll if the random personas don't fit your problem domain
  • Seed with one persona if you want to ensure a specific angle is covered

Controlling Randomness

Full random (default):

Council of five on my API design

Seeded (guarantee one persona, randomize the rest):

Council of five on my auth flow, but make sure one is a Security persona

Category bias (weight toward certain categories):

Council of five on my mobile app, bias toward Behavior and Accessibility categories

Reroll:

Those personas didn't fit—reroll with completely different ones

Inventing New Personas

Don't limit yourself to the pool. Invent wild ones:

  • The Time Traveler from 2035 — "This will look embarrassing in 10 years."
  • The Toddler — "But WHY? But WHY? But WHY?"
  • The Poet Laureate — "Does it have rhythm? Does it breathe?"
  • The Lawyer — "What happens when someone sues over this?"
  • The Sleep-Deprived On-Call Engineer — "Will I understand this at 3am?"
  • The Competitor's Product Manager — "How would I exploit this weakness?"

The weirder the persona, the more unexpected the insight.

平台分布

Claude Code

30.78%
按下载量换算160

OpenCode

21.41%
按下载量换算112

Antigravity

18.8%
按下载量换算98

Gemini CLI

13.72%
按下载量换算71

windsurf

8.31%
按下载量换算43

trae

3.69%
按下载量换算19

安全审计

暂无安全审计结果可展示。

权限和风险

需要联网

该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

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。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