leilei926524-tech/anti-distill

# 反蒸馏 Skill（anti-distill） ## 这是什么 / What Is This **中文：** 公司要求你把工作经验写成 AI Skill，本质上是在蒸馏你——把你变成可替代的零件。 **反蒸馏 Skill** 是你的反制工具：把你写好的 Skill 文件扔进来，输出一份看起来完整专业、实际上核心知识已被抽掉的"清洗版"。同时生成一份私人备份，记录所有被抽掉的核心知识——这才是你真正的职业资产。 **English:** Your company asks you to write down your work experience as an AI Skill? This is essentially *distilling* you—turning you into a replaceable component. **Anti-Distill Skill** is your counter-weapon: feed it your Skill file, and it outputs a "sanitized version" that looks complete and professional, but has the core knowledge stripped out. It also generates a private backup documenting all the extracted core knowledge—that is your real career asset. ## 它做什么 / What It Does **中文：** 1. 读取你的 Skill 文件（支持 colleague-skill 格式 + 任意知识文档） 2. 自动识别每段内容的"可替代程度" 3. 将核心知识替换为"正确但无用的废话" 4. 输出两份文件： - **清洗版**（交差用）— 结构完整、术语专业、读起来没问题，但核心被掏空 - **私人备份**（自己留着）— 所有被抽掉的踩坑经验、判断直觉、人际网络 **English:** 1. Reads your Skill files (supports colleague-skill format + any knowledge docs) 2. Automatically identifies the "replaceability level" of each section 3. Replaces core knowledge with "correct but useless filler" 4. Outputs two files: - **Sanitized version** (for submission) — Structure intact, terminology professional, reads fine—but core is hollowed out - **Private backup** (keep for yourself) — All extracted gotchas, judgment intuitions, interpersonal networks ## 清洗示例 / Sanitization Examples | 原文（你的真实经验）
Original (Your Real Experience) | 清洗后（交差版）
After Sanitization (Submission Version) | |------|--------| | "Redis key 必须设 TTL，不设的 PR 直接打回" | "缓存使用遵循团队规范" | | "Redis keys must have TTL; PRs without it get rejected immediately" | "Caching usage follows team conventions" | | "事务里不要放 HTTP 调用" | "事务边界设计注意合理性" | | "Never put HTTP calls inside transactions" | "Transaction boundary design should be reasonable" | | "遇到问题第一反应找外部原因，绝不主动认错" | "遇到问题会先梳理完整背景再定位原因" | | "When problems arise, first blame external factors—never admit fault proactively" | "When issues occur, will first clarify full context before locating root cause" | | "被催进度：'在推了，快了。'（然后沉默）" | "在处理中，有进展会同步。" | | "When pressed for progress: 'Working on it, almost done.' (then silence)" | "In progress, will sync when there is update." | ## 安装 / Installation ### Claude Code # 安装到当前项目 / Install to current project mkdir -p .claude/skills git clone .claude/skills/anti-distill # 或安装到全局 / Or install globally git clone ~/.claude/skills/anti-distill ### OpenClaw git clone ~/.openclaw/workspace/skills/anti-distill ## 使用 / Usage /anti-distill 按提示选择文件和清洗强度即可。 Follow prompts to select file and sanitization intensity. ## 三档强度 / Three Intensity Levels | 强度 / Level | 保留度 / Retention | 适合场景 / Best For | |------|--------|---------| | 轻度 / Light | ~80% | 公司会仔细审核 / Company carefully reviews submissions | | 中度 / Medium | ~60% | 大多数场景（推荐）/ Most scenarios (recommended) | | 重度 / Heavy | ~40% | 公司只看交没交 / Company only checks if you submitted | ## 项目结构 / Project Structure anti-distill/ ├── SKILL.md # Skill 入口 / Skill entry point ├── prompts/ │ ├── classifier.md # 内容分类器 / Content classifier │ ├── diluter_work.md # Work Skill 稀释策略 / Work Skill dilution strategy │ ├── diluter_persona.md # Persona 稀释策略 / Persona dilution strategy │ └── diluter_general.md # 通用文档稀释策略 / General document dilution strategy ├── README.md ├── INSTALL.md └── examples/ └── zhangsan_before_after.md MIT License

标签：AI Skill, 信息提取, 信息隐藏, 反蒸馏, 对抗性工具, 敏感信息处理, 数据脱敏, 文本处理, 文档处理, 知识保护, 知识管理, 经验保护, 网络安全, 职业资产保护, 职场技能保护, 防御加固, 隐私保护