Danshari Skill — Security skill for Claude Code
🧹 断舍离.skill / danshari.skill — 第一个劝你删 skill 的 skill:审计并移除你的 agent 不再需要的 skill(模型已会的 / harness 已自带的 / MCP 已接管的).
How to install Danshari Skill
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open swaylq/danshari-skill and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Danshari Skill does
🧹 断舍离.skill / danshari.skill — 第一个劝你删 skill 的 skill:审计并移除你的 agent 不再需要的 skill(模型已会的 / harness 已自带的 / MCP 已接管的)。The first skill that tells you to delete skills — audit & remove what your model no longer needs.
Alternatives in Security
- Pal — Multi-model AI integration — chat, debugging, code review, planning, security audit 11.3k ★
- Deepsec — Deepsec is a security harness for finding vulnerabilities in your codebase powered by coding agents 7.8k ★
- T3mp3st — autonomous red teaming platform; multi-agent offensive-security meta-harness 5.7k ★
README
🧹 断舍离.skill [](https://twitter.com/intent/tweet?text=%E6%96%AD%E8%88%8D%E7%A6%BB.skill%20%E2%80%94%20%E7%AC%AC%E4%B8%80%E4%B8%AA%E5%8A%9D%E4%BD%A0%E5%88%A0%20skill%20%E7%9A%84%20skill%E3%80%82%E5%AE%A1%E8%AE%A1%E4%BD%A0%E8%A3%85%E7%9A%84%E5%85%A8%E9%83%A8%20skill%EF%BC%8C%E7%A7%BB%E9%99%A4%E6%A8%A1%E5%9E%8B%E5%B7%B2%E7%BB%8F%E4%BC%9A%E7%9A%84%E3%80%81harness%20%E5%B7%B2%E8%87%AA%E5%B8%A6%E7%9A%84%E3%80%81MCP%20%E5%B7%B2%E6%8E%A5%E7%AE%A1%E7%9A%84%EF%BC%8C%E6%AF%8F%E8%BD%AE%E7%9C%81%E4%B8%8B%E4%B8%8A%E5%8D%83%20token%E3%80%82&url=https%3A%2F%2Fgithub.com%2Fswaylq%2Fdanshari-skill&hashtags=ClaudeCode,AIAgent,%E6%96%AD%E8%88%8D%E7%A6%BBskill,%E5%BC%80%E6%BA%90)
审计你装的全部 skill,把模型已经会的、harness 已自带的、MCP 已接管的 —— **移除**
*「你的模型一直在变强,你的 skill 还停在过去。」*
[](https://swaylq.github.io/danshari-skill/) [](LICENSE) [](https://claude.ai/code) [](https://skills.sh) [](#-安装)
整个生态都在做加法 ——
awesome 列表收录上千个 skill,单个仓库敢塞三百多个。
**没人做减法。**
但 skill 不是免费的:
每装一个,每一轮都在为它付出更多的 token;
模型每变强一代,就有一批 skill 从增益变成拖累。
**断舍离.skill 是一个劝你删掉 skill 的 skill ——**
**删的是「模型已经会的 skill」,留的是「模型还不知道的经验和流程」。**
[🌐 官网](https://swaylq.github.io/danshari-skill/) · [🚀 安装](#-安装) · [✨ 效果示例](#-效果示例) · [🧮 怎么判](#-怎么判) · [🔍 凭什么放心](#-凭什么放心) · [⭐ 演示审计](#-演示审计)
[**English README →**](README_en.md)
📊 **v0.1** — 完整审计流程(三分法 + 五问 + 盲测协议)+ 零依赖工具(`inventory.py` 盘点税表 / `archive.sh` 归档恢复,macOS 自带 bash 3.2 实测全过)+ 端到端[演示审计](examples/2026-08-17-demo/report.md)(12 个 skill
Related Skills
Jev Skill Router
Jev-powered skill router & security auditor for any AI agent (Codex, Claude Code, OpenCode, Hermes): ONE cheap
Claude Revoke
Audit and revoke everything Claude Code has been given on Linux: trusted folders, "always allow" rules, saved
Harness Decay Audit
Quarterly audit of rebar's harness components. Identifies commands/skills/agents that may have become dead wei
Jiahao
Dual-profile prompt-as-mental-model harness for LLM agents — advisory generator profile in the working agent,
Claude Skill Suite
Claude Code skills for running models cost-effectively: security gates (secret scanner, commit-gate), model-qu
Cleanup Docs
Audit and clean up a repo's Claude documentation — measure the always-on tier first (CLAUDE.md + nested + .cla
Related Agents
Vibe Slop Hunter
First-order AI-slop detector for UIs. One lens of the vibe-audit panel — hunts the obvious "vibecoded" tells (
Calendar Writer
Persists the approved event to the calendar via the event-calendar MCP server (create, update, or delete depen
Confluence Bulk
Execute bulk Confluence operations on 10+ pages via the confluence-as CLI wrapper: bulk-label, bulk-move, bulk