Tacit Mining — AI skill for Claude Code
Let AI truly understand you.
How to install Tacit Mining
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open xiaohuailabs/tacit-mining and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Tacit Mining does
Let AI truly understand you. A Claude Code skill that extracts tacit knowledge through structured dialogue. 隐性知识挖掘技能。
Alternatives in AI
- Agmsg — Cross-vendor messaging for CLI AI coding agents — let Claude Code, Codex, Gemini & Copilot talk to each other 1.5k ★
- Browserwing — BrowserWing turns your browser actions into MCP commands Or Claude Skill, allowing AI agents to control browse 1.2k ★
- Claude Hub — by Claude Did This - A webhook service that connects Claude Code to GitHub repositories, enabling AI-powered c 393 ★
README
tacit-mining

**Let AI truly understand you.** A Claude Code skill that extracts your tacit knowledge through structured dialogue.
"We know more than we can tell." — Michael Polanyi, 1966
你给 AI 写了几千字的提示词,但它还是不懂你。问题不在 AI,在于你最重要的判断标准——你自己也说不清楚。
`tacit-mining` 基于 Polanyi 隐性知识理论和认知科学方法(CDM、Laddering、Repertory Grid),用结构化对话从你的**行为**中反推出你说不出来的判断标准。
它能做什么
不问"你的标准是什么",而是问"上次你做了什么"。
你:改了三遍开头,每次都"不对",但说不清哪不对
AI:你改 Glasswing 那篇时,看到原始开头的第一反应是什么?
你:铺垫太长了,开头应该直接吸引人
AI:给你两个标题,A 悬念式,B 直给结论式,你选哪个?
你:B。大众看够了悬念,有营销味道
AI:所以你的规则是"吸引力靠内容本身,不靠信息差套路"?
你:对,但内容确实炸裂时可以酌情用悬念
**提取结果**:从"不要标题党"4 个字 → 变成 3 句话带边界条件的精确规则。
五个挖掘域
| 域 | 挖什么 |
|---|---|
| writing | 写作品味:什么开头好、什么是"AI味"、怎么算写透 |
| topic | 选题直觉:为什么一看就知道要写、优先级怎么排 |
| product | 产品判断:什么产品值得关注、机会怎么判断 |
| aesthetic | 视觉审美:封面、排版、配图的偏好 |
| audience | 读者感知:读者是谁、怎么和读者沟通 |
四种提取方法
- 关键事件法(CDM):回忆具体案例,逐点深挖决策
- 对比逼近(Repertory Grid):A vs B,选择暴露标准
- 反事实追问:如果变量 X 改变,判断会变吗?
- Laddering 追问:从表面 → 后果 → 价值观,一层层爬
安装
把 `tacit-mining` 文件夹复制到你的 Claude Code skills 目录:
# macOS / Linux
cp -r tacit-mining ~/.claude/skills/
# 或者直接 clone
git clone https://github.com/xiaohuailabs/tacit-mining.git ~/.claude/skills/tacit-mining
使用
在 Claude Code 对话中:
挖隐性知识
或者更具体:
聊聊我的写作品味
深度访谈
了解我
对话节奏
- 每轮只问一个问题,等你回答
- 可能追问 1-2 次
- 用一句话复述提炼的规则
- 你确认或纠正
- 5-8 轮一组,不拖太长
存储格式
每条提炼的隐性知识自动存为碎片文件:
---
name: tacit_writing_opening
description: 开头直给+吸引力靠内容本身
type: user
domain: writing
confidence: confirmed
---
## 表现(Behavior)
改稿时把铺垫式开头删掉,A/B对比中选了直接给结论。
## 规则(Rule)
吸引力必须来自内容本身的分量,不能靠信息差套路。
## 边界(Boundary)
内容本身确实很炸裂时,可以酌情使用情绪化表达。
## 原话(Verbatim)
> "大众已经看够了会厌烦,而且很有营销味道"
理论基础
- Michael Polanyi《The Tacit Dimension》(1966) — 隐性知识理论
- Gary Klein CDM 关键决策法 — NASA/军方验证的专家知识提取
- George Kelly Repertory Grid — 对比法提取隐含维度
- Dreyfus 五阶段模型 — 专家越专业越说不清
实战效果
首次 8 轮对话,提取了 8 条 confirmed 规则,覆盖写作、选题、产品判断三个域。从"不要标题党"到"默认直给结论
Related Skills
Knowledge
AI Product Manager - maintain living product understanding through dialogue
Tacit Knowledge
A cognitive protocol for AI agents. Makes any LLM-based agent think like an experienced practitioner. Works wi
Claude 3 7 Sonnet Starter Pack
Look, Anthropic's Claude-3.7-Sonnet is a powerful, hybrid CRASHOUT LLM. Let's understand this monumental relea
Peek API
Discover internal APIs from any website. Captures XHR/fetch calls, extracts auth headers, outputs structured e
Always On Memory Agent
Always-on memory for Claude Code. Automatically records sessions, extracts structured observations via LLM, co
AI Agent Handoff
Transfer context between AI agents/tools. Extracts Claude Code sessions and generates structured markdown hand
Related Agents
Story Capture
Captures project stories and experiences through guided conversation, extracts structured achievements using t
Brainstorming PRD
Transforms ambiguous project ideas into concrete specifications through structured brainstorming and iterative
Wiki Ingestor
Specialized agent for ingesting raw sources into the knowledge base. Reads source material, extracts structure