Back to Details

council

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

Council

顾问委员会

Consult your advisory council: three AI personas backed by different frontier models, each with full repo context.
咨询你的顾问委员会：由三位AI角色组成，分别由不同的前沿模型提供支持，每个角色都拥有完整的代码仓库上下文。

Personas

角色设定

NameArchetypeCore questionDefault provider
CatoSkeptical Strategist"What could go wrong?"
gemini
AdaExpansive Optimist"What could this become?"
claude
MarcusPragmatic Builder"What do we do next?"
codex
For full persona details, see references/personas.md.
名称原型核心问题默认模型提供商
Cato持怀疑态度的战略家“可能会出什么问题？”
gemini
Ada视野开阔的乐观主义者“这能发展成什么？”
claude
Marcus务实的构建者“我们下一步该做什么？”
codex
如需了解完整的角色细节，请查看references/personas.md。

How to invoke

调用方式

Run 
scripts/counsel.py
via Bash from the current working directory:
bash
undefined
在当前工作目录下通过Bash运行
scripts/counsel.py
脚本：
bash
undefined

Consult one persona

咨询单个角色

python3 <skill-dir>/scripts/counsel.py cato "Should we use Redis or Postgres for session storage?"
python3 <skill-dir>/scripts/counsel.py cato "我们应该使用Redis还是Postgres来存储会话信息？"

Consult all three (runs in parallel)

咨询全部三位角色（并行运行）

python3 <skill-dir>/scripts/counsel.py all "Review the changes in the last commit — are we missing anything?"
python3 <skill-dir>/scripts/counsel.py all "查看上次提交的变更内容——我们有没有遗漏什么？"

Override a persona's provider

覆盖角色的默认模型提供商

python3 <skill-dir>/scripts/counsel.py ada "Is this API design too minimal?" --ada-provider codex

Replace `<skill-dir>` with the absolute path to this skill directory.
python3 <skill-dir>/scripts/counsel.py ada "这个API设计是不是太简洁了？" --ada-provider codex

将`<skill-dir>`替换为该技能目录的绝对路径。

When to consult whom

何时咨询对应角色

  • Cato alone: Security review, risk assessment, premortem, "is this safe to deploy?"
  • Ada alone: Product direction, UX decisions, "is this feature compelling enough?"
  • Marcus alone: Implementation tiebreakers, scope cuts, "what's the simplest path?"
  • All three: Major architecture decisions, launch readiness, "should we do X or Y?"
  • 仅咨询Cato：安全审查、风险评估、事前验尸分析、“这个部署是否安全？”
  • 仅咨询Ada：产品方向、UX决策、“这个功能是否足够有吸引力？”
  • 仅咨询Marcus：实现方案抉择、范围裁剪、“最简单的路径是什么？”
  • 咨询全部三位：重大架构决策、发布准备、“我们应该选择X还是Y？”

Choosing the right moment

选择合适的咨询时机

Proactively suggest consulting the council when:
  • The user is about to make an irreversible architectural decision
  • There are multiple valid approaches with non-obvious trade-offs
  • A deployment feels risky or under-tested
  • The user explicitly asks "what do you think?" about a high-stakes choice
Always ask the user before invoking — say something like: "This feels like a good moment to consult the council. Want me to ask Cato/Ada/Marcus/all for their take?"
在以下场景下主动建议咨询委员会：
  • 用户即将做出不可逆转的架构决策
  • 存在多种有效方案，但权衡取舍不明确
  • 部署存在风险或测试不充分
  • 用户明确询问对高风险选择的看法，比如“你怎么看？”
调用前务必征得用户同意，可以这样说：“这似乎是咨询顾问委员会的好时机。要不要我让Cato/Ada/Marcus/全部角色给出他们的意见？”

Handling responses

处理反馈

  1. Run the script and capture output
  2. Show the full, unedited response from each persona to the user — do not summarize or paraphrase their words
  3. After showing all responses, briefly note key agreements or tensions across personas
  4. Ask the user how they want to proceed — the council advises, the user decides
  1. 运行脚本并捕获输出
  2. 向用户展示每个角色完整、未经编辑的反馈——不要总结或改写他们的内容
  3. 展示所有反馈后，简要指出各角色反馈中的关键共识或分歧
  4. 询问用户接下来的打算——委员会仅提供建议，最终由用户决定

Provider configuration

模型提供商配置

Each persona defaults to a different LLM provider for genuine model diversity. Override with flags if a provider is unavailable:
  • --cato-provider claude|codex|gemini
  • --ada-provider claude|codex|gemini
  • --marcus-provider claude|codex|gemini
The script has a 5-minute timeout per persona. All agents run in read-only mode in the current directory.
为保证模型多样性，每个角色默认使用不同的LLM提供商。如果某个提供商不可用，可通过参数覆盖：
  • --cato-provider claude|codex|gemini
  • --ada-provider claude|codex|gemini
  • --marcus-provider claude|codex|gemini
每个角色的脚本超时时间为5分钟。所有Agent在当前目录下以只读模式运行。