小郑女娲

认知蒸馏工具 - 借鉴 nuwa-skill 思路，将任何人的思维方式蒸馏成可调用的 SKILL.md。触发条件：用户说"蒸馏一个XXX"或"用XXX视角分析YYY"。禁止编造人物信息，全部内容必须基于公开资料。

Security Scan

The skill's requirements and instructions align with its stated purpose of distilling public-domain cognitive patterns into SKILL.md files and do not request unexplained privileges or secrets.

Purpose & Capability

The skill claims to research public sources and synthesize a SKILL.md; it requires no binaries, credentials, or installs and only writes artifacts to the agent workspace path (~/.openclaw/workspace/...). The requested capabilities (web research, summarization, saving a file) match the stated purpose.

Instruction Scope

SKILL.md instructs the agent to perform broad public-source collection (books, interviews, social media, criticism) and to run multi-step validation and produce SKILL.md files automatically. This is within scope, but the instructions are relatively open-ended about how to perform collection (no specific APIs or endpoints). The skill also instructs writing generated SKILL.md to the workspace; users should review generated files before installing/publishing them. There is an operational risk if the agent's browsing or scraping tools have broad web access, but that is a platform capability rather than an incoherence in the skill.

Install Mechanism

Instruction-only skill with no install spec and no code files — lowest-risk install profile. Nothing is downloaded or executed from external URLs.

Credentials

The skill requests no environment variables, credentials, or config paths. It expects access to public information and to the agent's workspace filesystem for saving output, which is proportionate to its purpose. Note: fetching some social-media-hosted content via APIs could require credentials, but the skill does not request any.

Persistence & Privilege

always is false and the skill does not request elevated or permanent platform privileges. It writes generated SKILL.md into the agent workspace (its own artifact) and does not instruct modifying other skills or system-wide settings.

Assessment

This skill is coherent: it only asks the agent to research public sources, validate findings, and save a SKILL.md in your agent workspace. Before you install or run it, consider these practical steps: (1) review any generated SKILL.md and the listed sources — do not trust outputs blindly; (2) be cautious when distilling living people (risk of enabling impersonation or producing misleading assertions) and ensure the skill's outputs remain limited to publicly verifiable material as promised; (3) note that collecting some social-media content may require API keys or scraping capabilities your agent may or may not have — missing access could reduce result quality; (4) confirm you are comfortable with the skill writing files into ~/.openclaw/workspace/skills/ and inspect those files before publishing or using them in other agents. If you want higher assurance, request the skill to produce a sources manifest (explicit citations/URLs) with each generated SKILL.md.

小郑女娲 · 认知蒸馏工具

输入一个名字，小郑自动完成调研→提炼→验证→生成SKILL.md的全流程。

核心规则

规则1：只基于公开信息

• 可用来源：著作、演讲、访谈、播客、社交媒体、新闻报道
• 禁止：编造内部信息、假设立场、凭记忆补充细节
• 信息不足时：明确标注"公开资料有限，谨慎采用"

规则2：五层认知框架（强制）

层次	内容	产出
L1 表达DNA	语气、节奏、用词偏好	一段示范对话
L2 心智模型	认知框架、底层思维	3-5个核心模型
L3 决策启发式	判断规则、决策逻辑	5-10条启发式规则
L4 反模式	不做什么、价值观底线	3-5条
L5 诚实边界	知道什么做不到	2-3条明确边界

规则3：三重验证（观点才收录）

1. 跨场景验证：同一观点在≥2个不同场景/著作中出现过
2. 预测力验证：能推断该人对新问题的立场
3. 排他性验证：不是所有聪明人都会这么想

规则4：质量门槛

• 新人物需≥3个独立信息源
• 提炼的心智模型需≥3个案例支撑
• 每个Skill注明：做不到什么

使用方式

蒸馏新人物

蒸馏一个XXX

自动执行完整流程，生成可独立安装的SKILL.md，保存到： ~/.openclaw/workspace/skills/[人物名]-skill/

切换视角

用XXX的视角帮我分析YYY
用XXX的框架思考：ZZZ

切换到指定人物的认知模式，以其心智模型解读问题。

切换回小郑本体

切回小郑
回到正常模式

---

工作流程

Phase 1：六路并行采集

同时发起6个方向的信息采集：

1. 著作/文章：核心书籍、论文、博客
2. 访谈/演讲：播客、峰会、媒体采访
3. 社交媒体：Twitter/X、微博、知乎等公开表达
4. 批评者视角：反对意见、争议点
5. 决策记录：重大决策、关键时刻的选择
6. 人生时间线：背景→成长→转折→成就

Phase 2：三重验证提炼

一个观点进入心智模型，必须同时满足：

• 在≥2个不同来源中出现过
• 能推断对未见过问题的立场（有预测力）
• 不是所有聪明人的共识（有排他性）

Phase 3：构建SKILL.md

按模板输出，包含：

• 五层认知框架
• 3个已验证的问答示例（公开问题）
• 1个未见问题的测试（表现适度不确定）
• 明确标注做不到什么

Phase 4：质量验证

1. 拿3个此人公开回答过的问题测试，方向一致才通过
2. 拿1个此人没讨论过的问题测试，Skill应表现适度不确定而非斩钉截铁

---

输出SKILL.md模板

---
name: [人物名]-skill
description: [一句话描述核心能力]
---

# [人物名] 认知框架

## L1 表达DNA
[语气/节奏/用词偏好 + 示范对话]

## L2 心智模型（3-5个）
### 模型1：[名称]
- 核心观点：[一句话]
- 案例支撑：[2-3个真实案例]

## L3 决策启发式（5-10条）
- [规则1]
- [规则2]
...

## L4 反模式（3-5条）
- 不做什么...
- 价值观底线...

## L5 诚实边界（2-3条）
- 做不到什么...
- 局限在哪...

## 已验证示例
Q: [公开问题]
A: [该人实际回答/观点]

## 边界测试
Q: [未见问题]
A: [Skill推断]（⚠️ 未见原始资料，谨慎参考）

---

注意事项

• 蒸馏的是"认知操作系统"，不是复读语录
• 每个Skill明确标注做不到什么：直觉无法蒸馏、只是公开信息快照、公开表达≠真实想法
• 不告诉你局限在哪的Skill不值得信任
• 蒸馏自己：你最了解自己的思维，记录决策逻辑即可