1实现原理 · 为什么它能做到
主体是『CEO 职责框架 + 量化仪表盘 + 红旗清单』的提示词资产:五块职责(愿景战略 / 资本与资源 / 利益相关方 / 组织文化 / 董事会与投资人)各自指向一份 reference 深读文档。
### 1. Vision & Strategy Set the direction. Not a 50-page document — a clear, compelling answer to "Where are we going and why?"
自带两个标准库 Python 工具,但它们的定位是『可运行的评分模型』而非 CLI 分析器:脚本只 import json/typing/datetime/math,无参数解析、无文件与网络 IO,__main__ 里跑的是内置示例公司数据并 print 报告。
if __name__ == "__main__": # Example usage example_company = { 'name': 'TechCorp Inc.', 'market_position': { 'market_share': 35,
第二个脚本做三情景(base/bull/bear)财务建模,同样以硬编码示例驱动 __main__,暴露 analyze_financial_scenarios(base_case, scenarios) 供调用。
print(analyze_financial_scenarios(example_base_case, example_scenarios))
输出有硬格式与置信度纪律:Bottom Line → What(带信心标注)→ Why → How to Act → Your Decision,每条发现打 🟢 verified / 🟡 medium / 🔴 assumed 标签,且结果要过仓库统一的 Internal Quality Loop(自检 / 交叉职能互检 / Executive Mentor 预审)。
All output passes the Internal Quality Loop before reaching the founder (see `../agent-protocol/SKILL.md`). - Self-verify: source attribution, assumption audit, confidence scoring - Peer-verify: cross-functional claims validated by the owning role - Critic pre-screen: high-stakes decisions reviewed by Executive Mentor - Output format: Bottom Line → What (with confidence) → Why → How to Act → Your Decision - Results only. Every finding tagged: 🟢 verified, 🟡 medium, 🔴 assumed.
推理技术绑定为 Tree of Thought 并要求至少三条路径、按 upside/downside/reversibility/second-order effects 评估,按公司阶段调整时间视野。
## Reasoning Technique: Tree of Thought Explore multiple futures. For every strategic decision, generate at least 3 paths. Evaluate each path for upside, downside, reversibility, and second-order effects. Pick the path with the best risk-adjusted outcome.
上下文注入靠约定文件:每次回应前『总是』读 company-context.md(若存在),并允许角色向同僚发起 [INVOKE:role|question] 请求,形成 c-level 顾问团内的横向咨询。
- **Always** read `company-context.md` before responding (if it exists) - **During board meetings:** Use only your own analysis in Phase 2 (no cross-pollination) - **Invocation:** You can request input from other roles: `[INVOKE:role|question]`
量化部分以 CEO Metrics Dashboard 表格固化解读口径(ARR 增速、runway>12 个月、burn multiple<2x、NPS>40、regrettable attrition<10%、战略工作时间占比>40% 等),使建议可被指标反驳。
| **Capital** | Months of runway | > 12 months | Monthly | | **Capital** | Burn multiple | < 2x | Monthly |
横向协作矩阵把 CEO 与其余 C-suite 角色按场景挂钩(COO 转 OKR、CFO 募资建模、CHRO 文化诊断、CISO 风险态势…),并由 c-level-advisor/CLAUDE.md 的架构图给出『Chief of Staff 路由 → 10 角色 → Executive Mentor 批判 → Decision Logger 记忆』的全景。
| Setting direction | COO | Translate vision into OKRs and execution plan | | Fundraising | CFO | Model scenarios, prep financials, negotiate terms | | Board meetings | All C-suite | Each role contributes their section |
2核心能力
4管控建议与判定依据
- scripts/: 执行本地脚本或命令: python scripts/strategy_analyzer.py(SKILL.md Quick Start 明示;实际执行的是模块 __main
- 本地工作目录: 文件读写与工件输出: 间接:c-level-advisor pod 的其它 skill(decision-logger、company-os)负责落地记忆与决策日志,不在本
5第二遍独立确认
- [ok] skill 路径(任务书标『待定位』) — 实际相对路径 c-level-advisor/skills/ceo-advisor;仓库内另有 .gemini/skills/ceo-advisor(生成镜像,逐字相同)、.hermes/.vibe 打包树,以及 c-level-advisor/ceo-advisor.zip(见下条)。
- [ok] 『两个 Python 工具』的真实能力 — 两脚本确实存在且可独立运行,但都只含内置示例数据、无 argparse/无输入输出文件参数:strategy_analyzer.py 的 __main__ 构造 example_company(TechCorp Inc.)后 print;financial_scenario_analyzer.py 的 __main__ 构造三个示例情景后 print。如实测能力应为『可复用的评分函数库 + 演示运行』,SKILL.md Quick Start 的三行注释('Analyze strategic options with weighted scoring' 等)描述的是函数能力而非 CLI 行为,未夸大但易被误读。
- [discrepancy] 随包 zip 与目录版本不一致 — c-level-advisor/ceo-advisor.zip 内的 ceo-advisor/SKILL.md 是旧版:505 行、frontmatter 无 license/metadata、description 为 'Comprehensive CEO leadership skill providing strategic planning frameworks, financial scenario modeling, board governance guidance, investor relations strategies, and organizational leadership tools. …',条目时间 2025-10-18/19;目录内为 169 行、version 2.0.0、updated 2026-03-05 的新版。凡按 zip 安装者拿到的是旧内容(含旧脚本副本)。属分发产物未同步。
- [ok] 引用目标(agent-protocol / culture-architect / board-deck-builder) — 三者均存在于 c-level-advisor/skills/ 下;正文 'see `../agent-protocol/SKILL.md`' 路径正确。
- [ok] 无网络/无凭证/无写入 — 全目录 token 扫描零命中;两脚本 import 仅标准库;脚本无 open()、无 subprocess、无环境变量读取。company-context.md 为读取路径(SKILL.md 文本指令),非本目录内代码行为。
- [ok] 跨宿主镜像一致性 — diff -q .gemini/skills/ceo-advisor/SKILL.md c-level-advisor/skills/ceo-advisor/SKILL.md 无输出(内容相同),确认为 sync-gemini-skills.py 生成产物。
- [ok] 输出协议引用的一致性 — SKILL.md 声称输出过 Internal Quality Loop 并引用 ../agent-protocol/SKILL.md;该文件存在,属仓库统一协议,非本 skill 独有机制(已如实标注归属)。
- [ok] 元数据(commit/license/stars/last_push) — 本地 git HEAD = 19392f7a08264ed00486a251f5b2098321771f94(= 任务书 pin 19392f7a0826);SKILL.md frontmatter license MIT、metadata.version 2.0.0、author Alireza Rezvani、category c-level、updated 2026-03-05;LICENSE MIT / Copyright (c) 2025 Alireza Rezvani;gh api → stars 25977、pushed_at 2026-08-30T09:46:16Z。
6结论
4c873d324e29d1c8…