skill-admin

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Skill Admin

Skill管理工具

Manage skills across 20+ AI platforms with single source of truth architecture.

基于单一可信源架构，管理20+ AI平台上的Skill。

Supported Platforms

支持的平台

claude

cursor

copilot

vscode

cline

roo

goose

codex

windsurf

trae

kiro

opencode

openclaw

hermes

qoderwork

gemini

antigravity

agents

claude

cursor

copilot

vscode

cline

roo

goose

codex

windsurf

trae

kiro

opencode

openclaw

hermes

qoderwork

gemini

antigravity

agents

CLI Commands

CLI命令

bash

undefined

bash

undefined

Unified list / scan

统一列出/扫描

python3 skill-admin.py list [platform]

Backward-compatible alias

向后兼容别名

python3 skill-admin.py scan

Sync all to .agents + symlink others

将所有Skill同步到.agents并为其他平台创建符号链接

python3 skill-admin.py sync-all

Delete skill from all platforms

从所有平台删除指定Skill

python3 skill-admin.py delete <name>

Generate category graph

生成分类图谱

python3 skill-admin.py graph [platform]

Install from GitHub

从GitHub安装Skill

python3 skill-admin.py install <url> [platform]

Compare two platforms

对比两个平台的Skill

python3 skill-admin.py compare <p1> <p2>

Show symlink topology

显示符号链接拓扑结构

python3 skill-admin.py topology

Security scan (11 dimensions)

安全扫描（11个维度）

python3 skill-admin.py security [path]

Backup all skills

备份所有Skill

python3 skill-admin.py backup

undefined

python3 skill-admin.py backup

undefined

AI Output Templates

AI输出模板

CRITICAL: After running CLI, format output using templates below. Do NOT dump raw script output.

重要提示：运行CLI后，请使用以下模板格式化输出。请勿直接输出原始脚本结果。

Guide Flow

引导流程

Default:

Run
```
python3 skill-admin.py list [platform]
```
List installed skills

Only when the user explicitly asks what skill to use:

Run
```
python3 skill-admin.py list [platform]
```
Use only each skill's
```
name
```
and
```
description
```
to identify relevant skills
Explain what the relevant skills do and why they fit the current task
Do not read full skill bodies unless the user explicitly asks

默认流程：

运行
```
python3 skill-admin.py list [platform]
```
列出已安装的Skill

仅当用户明确询问应使用何种Skill时：

运行
```
python3 skill-admin.py list [platform]
```
仅使用每个Skill的
```
name
```
和
```
description
```
来识别相关Skill
说明相关Skill的功能以及它们为何适用于当前任务
除非用户明确要求，否则不要读取完整的Skill内容

List Output Contract

列表输出约定

list

outputs stable structured text for AI consumption:

```
SKILL_LIST_BEGIN
```
```
SUMMARY ...
```
```
PLATFORM ...
```
```
SKILL ...
```
```
EMPTY_DIR ...
```
if any
```
SKILL_LIST_END
```

Parsing rules:

Fields are tab-separated
Each field is
```
key=value
```
Use
```
SKILL.name
```
and
```
SKILL.description
```
for guide/recommendation
Do not depend on colors or box-drawing characters

list

命令输出供AI读取的稳定结构化文本：

```
SKILL_LIST_BEGIN
```
```
SUMMARY ...
```
```
PLATFORM ...
```
```
SKILL ...
```
```
EMPTY_DIR ...
```
（如有）
```
SKILL_LIST_END
```

解析规则：

字段使用制表符分隔
每个字段格式为
```
key=value
```
使用
```
SKILL.name
```
和
```
SKILL.description
```
进行引导/推荐
不要依赖颜色或绘图字符

Template 1: Summary (for "how many skills")

模板1：摘要（用于“技能数量统计”）

统计项	数量
总技能数	XXX
已安装平台	XXX
去重后唯一技能	XXX

统计项	数量
总技能数	XXX
已安装平台	XXX
去重后唯一技能	XXX

Template 2: Full Skill List (for "list all skills")

模板2：完整技能列表（用于“列出所有技能”）

分类	数量	技能列表
💻 开发工具	XX	`skill-a` `skill-b` ...
🎨 设计工具	XX	`skill-a` `skill-b` ...
📚 文档工具	XX	`skill-a` `skill-b` ...
🌐 网络交互	XX	`skill-a` `skill-b` ...
🤖 AI 能力	XX	`skill-a` `skill-b` ...
⚙️ 系统自动化	XX	`skill-a` `skill-b` ...
🧪 测试质量	XX	`skill-a` `skill-b` ...
📊 数据分析	XX	`skill-a` `skill-b` ...
📦 其他	XX	`skill-a` `skill-b` ...
总计	XXX

List ALL skills, NO truncation.

分类	数量	技能列表
💻 开发工具	XX	`skill-a` `skill-b` ...
🎨 设计工具	XX	`skill-a` `skill-b` ...
📚 文档工具	XX	`skill-a` `skill-b` ...
🌐 网络交互	XX	`skill-a` `skill-b` ...
🤖 AI 能力	XX	`skill-a` `skill-b` ...
⚙️ 系统自动化	XX	`skill-a` `skill-b` ...
🧪 测试质量	XX	`skill-a` `skill-b` ...
📊 数据分析	XX	`skill-a` `skill-b` ...
📦 其他	XX	`skill-a` `skill-b` ...
总计	XXX

列出所有技能，请勿截断。

Classification Rules

分类规则

类别	关键词
💻 开发工具	code, dev, debug, test, tdd, review, git, typescript, react, rax, fliggy, ship, ultron, clamjs
🎨 设计工具	design, ui, ux, figma, mobile, ios, web-design, generative, canvas, infographic, drafter
📚 文档工具	doc, pdf, docx, pptx, xlsx, obsidian, notebook, writing, guide, release
🌐 网络交互	web, browse, browser, gstack, opencli, fetch, search, screenshot, websearch, webfetch
🤖 AI 能力	ai, llm, claude, gemini, agent, pua, superpowers, brainstorm, plan, office-hours
⚙️ 系统自动化	freeze, guard, sync, backup, file, organizer, loop, scheduler, health
🧪 测试质量	qa, test, quality, verify, doctor, investigate, security
📊 数据分析	data, odps, query, rum, analytics, yfinance, stock, finance, correlation

类别	关键词
💻 开发工具	code, dev, debug, test, tdd, review, git, typescript, react, rax, fliggy, ship, ultron, clamjs
🎨 设计工具	design, ui, ux, figma, mobile, ios, web-design, generative, canvas, infographic, drafter
📚 文档工具	doc, pdf, docx, pptx, xlsx, obsidian, notebook, writing, guide, release
🌐 网络交互	web, browse, browser, gstack, opencli, fetch, search, screenshot, websearch, webfetch
🤖 AI 能力	ai, llm, claude, gemini, agent, pua, superpowers, brainstorm, plan, office-hours
⚙️ 系统自动化	freeze, guard, sync, backup, file, organizer, loop, scheduler, health
🧪 测试质量	qa, test, quality, verify, doctor, investigate, security
📊 数据分析	data, odps, query, rum, analytics, yfinance, stock, finance, correlation

Security Scan

安全扫描

11 dimensions: Prompt Injection, Permission Abuse, Hook Behavior, Data Exfiltration, Social Engineering, Scope Creep, Obfuscation, Destructive Ops, Supply Chain, Identity Impersonation, Sensitive Data.

Score	Label	Action
85-100	TRUSTED	APPROVE
70-84	LIKELY_SAFE	APPROVE_WITH_CAUTION
50-69	SUSPICIOUS	APPROVE_WITH_CAUTION
25-49	HIGH_RISK	REJECT
0-24	DANGEROUS	REJECT

Edit

scripts/security-rules.yaml

to customize rules.

11个维度：提示注入、权限滥用、钩子行为、数据泄露、社会工程、范围蔓延、代码混淆、破坏性操作、供应链风险、身份冒充、敏感数据。

分数	标签	操作
85-100	TRUSTED	APPROVE
70-84	LIKELY_SAFE	APPROVE_WITH_CAUTION
50-69	SUSPICIOUS	APPROVE_WITH_CAUTION
25-49	HIGH_RISK	REJECT
0-24	DANGEROUS	REJECT

编辑

scripts/security-rules.yaml

来自定义规则。

Requirements

环境要求

Python 3.8+, stdlib only (no dependencies).

Python 3.8+，仅使用标准库（无额外依赖）。