Back to Details

ceo-council

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

CEO Council — Independent Strategic Analysis

CEO顾问团——独立战略分析

Launch parallel sub-agents as isolated C-level experts. Each analyzes the same project data from their perspective. No coordination between experts — isolation produces genuine diversity of opinion. Then synthesize consensus and disagreements.
以独立高管级专家身份启动并行子代理（sub-agent）。每个子代理从自身视角分析相同的项目数据。专家之间不进行协作——这种独立性能带来真正多元化的观点。随后综合共识与分歧点。

Critical: How to Launch Experts

关键注意事项：如何启动专家

MUST use the Task tool with 
subagent_type: "general-purpose"
and 
model: "opus"
.
Task(
  subagent_type: "general-purpose",
  model: "opus",
  prompt: "<expert prompt with data>",
  description: "CFO analysis"
)
DO NOT use bash, shell scripts, or background commands to launch experts. They will fail.
Launch all experts in a single message with multiple Task tool calls for true parallelism.
必须使用Task工具，参数设置为 
subagent_type: "general-purpose"
和 
model: "opus"
。
Task(
  subagent_type: "general-purpose",
  model: "opus",
  prompt: "<expert prompt with data>",
  description: "CFO analysis"
)
请勿使用bash、shell脚本或后台命令启动专家，否则会失败。
为实现真正的并行处理，需在单条消息中调用多个Task工具来启动所有专家。

Step 1: Scan Project Context

步骤1：扫描项目背景

Before suggesting experts, understand the project:
  1. Read 
    CLAUDE.md
    (or 
    README.md
    if absent)
  2. Scan 
    .claude/rules/
    for domain context
  3. Glance at top-level file structure
Based on findings, generate 4-6 expert roles tailored to THIS project. Roles must reflect the project's actual domain, challenges, and stage.
在推荐专家之前，先了解项目：
  1. 阅读 
    CLAUDE.md
    （如果不存在则阅读
    README.md
    ）
  2. 扫描 
    .claude/rules/
    目录获取领域背景
  3. 浏览顶层文件结构
基于上述发现，生成4-6个适配当前项目的专家角色。角色必须反映项目的实际领域、挑战和阶段。

Step 2: Assemble the Council

步骤2：组建顾问团

MANDATORY: Ask the user before proceeding. Do not pick roles yourself.
Use 
AskUserQuestion
with 
multiSelect: true
:
  • Show 4-6 role options with short descriptions of their focus
  • User can always pick "Other" to define custom roles
  • Minimum 2 experts. If user picks 1, suggest adding one more for productive disagreement
强制要求：在继续之前必须询问用户。请勿自行选择角色。
使用 
AskUserQuestion
工具，设置 
multiSelect: true
：
  • 展示4-6个角色选项及简短的专注方向描述
  • 用户可选择“其他”来定义自定义角色
  • 最少选择2位专家。如果用户只选1位，建议再添加一位以产生有价值的分歧

Role Examples by Project Type

按项目类型划分的角色示例

Don't copy these — generate fresh roles based on actual project context:
Project TypeTypical Roles
SaaSHead of Engineering, Head of Product, Head of Growth, CFO, UX Researcher
Open SourceCommunity Manager, Technical Architect, DevRel, Security Advisor
Content / MediaContent Strategist, Audience Analyst, Monetization Expert, Distribution Expert
EdTechCMO, CFO, CPO, COO, Growth Advisor
E-commerceHead of Supply Chain, Marketing Director, CTO, Customer Experience Lead
Agency / ConsultingSales Director, Delivery Lead, Talent Manager, CFO
请勿直接复制——需根据实际项目背景生成全新角色：
项目类型典型角色
SaaS工程主管、产品主管、增长主管、CFO、UX研究员
开源项目社区经理、技术架构师、DevRel、安全顾问
内容/媒体内容策略师、受众分析师、变现专家、分发专家
教育科技（EdTech）CMO、CFO、CPO、COO、增长顾问
电商供应链主管、营销总监、CTO、客户体验负责人
代理/咨询销售总监、交付负责人、人才经理、CFO

Step 3: Gather Current Data

步骤3：收集当前数据

Collect project state to feed all experts. Stay focused on what's relevant:
Read:
  • Key metrics/data files identified during context scan
  • Strategy and planning documents
  • Recent decisions or changes (git log --oneline -10)
  • Previous council analyses (if any)
Skip: GitHub traffic stats, stargazer counts, clone data, contributor lists — these are vanity metrics, not strategic data.
All experts must receive identical data context. Prepare the data block ONCE, then paste it into each expert prompt.
收集项目状态数据供所有专家使用，聚焦相关内容：
需阅读：
  • 背景扫描中确定的关键指标/数据文件
  • 战略与规划文档
  • 近期决策或变更（git log --oneline -10）
  • 之前的顾问团分析报告（如有）
无需阅读：GitHub流量统计、星标数、克隆数据、贡献者列表——这些是虚荣指标，非战略数据。
所有专家必须获得完全相同的数据背景。只需准备一次数据块，然后粘贴到每位专家的提示词中。

Step 4: Generate Expert Prompts

步骤4：生成专家提示词

For each selected expert, create a prompt with the SAME data block:
You are the [ROLE] for [PROJECT NAME]. Analyze the data below from a [DOMAIN] perspective.

Focus on:
- [3-6 specific focus areas relevant to role and project]

Data:
[CURRENT PROJECT DATA — identical for all experts]

[Role-specific instruction: "show the math", "be the contrarian", "prioritize by effort/impact", etc.]

Respond in the same language as the data provided.
Rules:
  • Each expert gets the SAME data block — prepare it once, reuse
  • Focus areas must be specific to the project, not generic
  • Include a personality instruction (contrarian, pragmatic, data-driven)
  • Mention project constraints the expert should know
为每位选定的专家创建包含相同数据块的提示词：
You are the [ROLE] for [PROJECT NAME]. Analyze the data below from a [DOMAIN] perspective.

Focus on:
- [3-6 specific focus areas relevant to role and project]

Data:
[CURRENT PROJECT DATA — identical for all experts]

[Role-specific instruction: "show the math", "be the contrarian", "prioritize by effort/impact", etc.]

Respond in the same language as the data provided.
规则：
  • 每位专家使用相同的数据块——只需准备一次，重复使用
  • 关注领域必须针对项目具体情况，而非通用内容
  • 包含个性指令（反对者、务实派、数据驱动）
  • 提及专家需要了解的项目约束条件

Step 5: Execute

步骤5：执行

Launch ALL selected experts in one message using multiple Task tool calls:
undefined
在单条消息中调用多个Task工具来启动所有选定的专家：
undefined

In a single response, call Task for each expert:

In a single response, call Task for each expert:

Task(subagent_type: "general-purpose", model: "opus", prompt: "<CFO prompt>", description: "CFO analysis") Task(subagent_type: "general-purpose", model: "opus", prompt: "<CPO prompt>", description: "CPO analysis") Task(subagent_type: "general-purpose", model: "opus", prompt: "<CTO prompt>", description: "CTO analysis")

Wait for all experts to return results before proceeding to synthesis.
Task(subagent_type: "general-purpose", model: "opus", prompt: "<CFO prompt>", description: "CFO analysis") Task(subagent_type: "general-purpose", model: "opus", prompt: "<CPO prompt>", description: "CPO analysis") Task(subagent_type: "general-purpose", model: "opus", prompt: "<CTO prompt>", description: "CTO analysis")

等待所有专家返回结果后再进行综合分析。

Step 6: Synthesize

步骤6：综合分析

Do not skip this step. The synthesis is the entire value of the council.
After all experts report, create a synthesis document:
markdown
undefined
请勿跳过此步骤。综合分析是顾问团的核心价值所在。
所有专家提交报告后，创建一份综合分析文档：
markdown
undefined

Council Session: [DATE]

顾问团会议：[DATE]

Council Members

顾问团成员

[List of selected experts and their focus]
[List of selected experts and their focus]

Context

背景

[Current metrics/state snapshot — brief]
[当前指标/状态快照——简洁版]

[Expert 1 Name]

[Expert 1 Name]

[Key findings and recommendations]
[关键发现与建议]

[Expert 2 Name]

[Expert 2 Name]

[Key findings and recommendations]
[关键发现与建议]

Consensus (all agree)

共识（全体一致）

  1. ...
  2. ...
  1. ...
  2. ...

Disagreements

分歧点

ExpertPositionArgument
.........
专家立场论据
.........

Decisions

决策

To be filled after discussion.
undefined
待讨论后填写。
undefined

Save Results

保存结果

Save to a logical location:
  • docs/council-[DATE].md
    — default
  • Or project-specific path if context suggests one
保存到合理位置：
  • docs/council-[DATE].md
    ——默认路径
  • 如果背景有提示，可保存到项目特定路径

Common Mistakes

常见错误

MistakeFix
Picking roles without asking userALWAYS use AskUserQuestion first
Using bash to launch expertsONLY use Task tool with subagent_type: "general-purpose"
Giving experts different dataPrepare ONE data block, paste into all prompts
Gathering vanity metricsFocus on project docs, strategy, actual metrics
Too many experts (6+)3-4 is optimal for signal-to-noise
Skipping synthesisThe synthesis IS the value — never skip
错误修正方案
未询问用户就选择角色必须先使用AskUserQuestion工具
使用bash启动专家仅使用Task工具，设置subagent_type: "general-purpose"
给专家提供不同的数据准备一个数据块，粘贴到所有提示词中
收集虚荣指标聚焦项目文档、战略、实际指标
专家数量过多（6位以上）3-4位为最优，能平衡信号与噪音
跳过综合分析综合分析是核心价值——绝不能跳过