find-skills-x

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

Find Skills X

Find Skills X

工作流程

Workflow

用户输入
┌─────────────────────────────────┐
│ 1️⃣ AI 关键词分析                │
│    提取核心词、扩展相关词         │
└─────────────────────────────────┘
┌─────────────────────────────────┐
│ 2️⃣ 多渠道并行搜索               │
│    skills.sh API + SkillHub + GH │
└─────────────────────────────────┘
┌─────────────────────────────────┐
│ 3️⃣ 结果合并去重                 │
│    按安装量/质量排序             │
└─────────────────────────────────┘
输出推荐结果

User Input
┌─────────────────────────────────┐
│ 1️⃣ AI Keyword Analysis          │
│    Extract core keywords and expand related terms │
└─────────────────────────────────┘
┌─────────────────────────────────┐
│ 2️⃣ Multi-channel Parallel Search │
│    skills.sh API + SkillHub + GH │
└─────────────────────────────────┘
┌─────────────────────────────────┐
│ 3️⃣ Result Merging and Deduplication │
│    Sort by installation count/quality │
└─────────────────────────────────┘
Output Recommended Results

1️⃣ AI 关键词分析

1️⃣ AI Keyword Analysis

目标:从用户输入中提取最有效的搜索关键词。
Objective: Extract the most effective search keywords from user input.

🎯 意图识别决策树

🎯 Intent Recognition Decision Tree

用户输入
单一技术? ("Java", "React", "Docker")
  ├─ 是 → 提取核心词 + 领域扩展词
  └─ 否 ↓

技术+领域? ("Java 后端", "React 移动端")
  ├─ 是 → 拆分为: 技术 + 领域,并行搜索
  └─ 否 ↓

多关键词? ("Kotlin Android 后端通信")
  ├─ 是 → 提取所有关键词,并行搜索
  └─ 否 → 单一意图,直接搜索

最终: 合并去重 → 分类展示
User Input
Single technology? ("Java", "React", "Docker")
  ├─ Yes → Extract core keywords + domain expansion terms
  └─ No ↓

Technology + domain? ("Java backend", "React mobile")
  ├─ Yes → Split into: Technology + domain, search in parallel
  └─ No ↓

Multiple keywords? ("Kotlin Android backend communication")
  ├─ Yes → Extract all keywords, search in parallel
  └─ No → Single intent, search directly

Final: Merge and deduplicate → Display by category

📝 查询优化策略

📝 Query Optimization Strategy

重要规则
  1. 提取 ALL 关键词 - 不要只关注第一个词
  2. 领域词智能扩展 - "后端"→"backend","通信"→"communication"
  3. 并行独立搜索 - 每个核心关键词独立搜索
  4. 中英文转换 - 提供英文搜索词提高匹配率
Important Rules:
  1. Extract ALL keywords - Don't only focus on the first term
  2. Intelligent domain term expansion - "后端"→"backend", "通信"→"communication"
  3. Parallel independent search - Each core keyword is searched independently
  4. Chinese-English conversion - Provide English search terms to improve matching rate

Prompt 模板

Prompt Template

分析用户输入,提取搜索关键词。

用户输入: "{query}"

**关键规则**:
- 提取所有重要关键词,不要遗漏
- 对领域词进行合理的英文转换
- 每个核心关键词都独立有价值,需要分别搜索

输出 JSON:
{
  "core": ["所有核心关键词"],
  "expanded": {
    "domain": ["技术/功能领域扩展"],
    "i18n": ["中英文变体"]
  }
}

示例:
输入: "用户认证"
输出: {
  "core": ["authentication"],
  "expanded": {
    "domain": ["auth", "login", "security", "jwt", "oauth", "session", "token"],
    "i18n": ["user authentication", "login"]
  }
}

输入: "安卓开发"
输出: {
  "core": ["android"],
  "expanded": {
    "domain": ["mobile", "kotlin", "java", "jetpack", "compose"],
    "i18n": ["android"]
  }
}

输入: "Kotlin Android 后端通信"
输出: {
  "core": ["Kotlin", "Android", "backend", "communication", "API"],
  "expanded": {
    "domain": ["server", "networking", "microservice"],
    "i18n": ["kotlin android", "backend communication"]
  }
}

输入: "Python 数据分析 可视化"
输出: {
  "core": ["Python", "data-analysis", "visualization"],
  "expanded": {
    "domain": ["pandas", "matplotlib", "plotly", "data-science"],
    "i18n": ["data analysis", "data visualization"]
  }
}

Analyze user input and extract search keywords.

User Input: "{query}"

**Key Rules**:
- Extract all important keywords, do not miss any
- Perform reasonable English conversion for domain terms
- Each core keyword is independently valuable and needs to be searched separately

Output JSON:
{
  "core": ["all core keywords"],
  "expanded": {
    "domain": ["technology/function domain expansions"],
    "i18n": ["Chinese-English variants"]
  }
}

Example:
Input: "User Authentication"
Output: {
  "core": ["authentication"],
  "expanded": {
    "domain": ["auth", "login", "security", "jwt", "oauth", "session", "token"],
    "i18n": ["user authentication", "login"]
  }
}

Input: "Android Development"
Output: {
  "core": ["android"],
  "expanded": {
    "domain": ["mobile", "kotlin", "java", "jetpack", "compose"],
    "i18n": ["android"]
  }
}

Input: "Kotlin Android Backend Communication"
Output: {
  "core": ["Kotlin", "Android", "backend", "communication", "API"],
  "expanded": {
    "domain": ["server", "networking", "microservice"],
    "i18n": ["kotlin android", "backend communication"]
  }
}

Input: "Python Data Analysis Visualization"
Output: {
  "core": ["Python", "data-analysis", "visualization"],
  "expanded": {
    "domain": ["pandas", "matplotlib", "plotly", "data-science"],
    "i18n": ["data analysis", "data visualization"]
  }
}

2️⃣ 多渠道搜索

2️⃣ Multi-channel Search

搜索策略

Search Strategy

优先级:API > CLI > Web
references/channels.json
中配置的
search_method
执行:
search_method执行方式说明
API
curl {api_endpoint}
只返回元数据,不下载代码
CLI
{cli_command}
只返回元数据,不下载代码
Weblibrarian agent 或手动访问浏览网页,不下载代码
GitHub
gh search repos --json
只返回仓库元数据(name/desc/url/stars)
关键区分
  • 🔍 搜索阶段:只获取元数据(名称、描述、安装量、URL)→ 不下载代码
  • ⬇️ 安装阶段
    npx skills add {owner}/{repo}
    → 才下载技能文件到本地
Priority: API > CLI > Web
Execute according to the
search_method
configured in
references/channels.json
:
search_methodExecution MethodDescription
API
curl {api_endpoint}
Return only metadata, do not download code
CLI
{cli_command}
Return only metadata, do not download code
Weblibrarian agent or manual accessBrowse web pages, do not download code
GitHub
gh search repos --json
Return only repository metadata (name/desc/url/stars)
Key Distinction:
  • 🔍 Search Phase: Only obtain metadata (name, description, installation count, URL) → Do not download code
  • ⬇️ Installation Phase:
    npx skills add {owner}/{repo}
    → Only then download skill files to local

GitHub 多技能仓库扫描

GitHub Multi-skill Repository Scanning

对于包含多个 skills 的 GitHub 仓库(如 MiniMax-AI/skills),需要扫描目录结构识别每个 skill。
判断标准(必须全部满足):
  1. 目录包含
    SKILL.md
    文件
  2. SKILL.md
    包含
    name:
    字段
  3. SKILL.md
    包含
    description:
    字段
扫描流程
bash
undefined
For GitHub repositories containing multiple skills (such as MiniMax-AI/skills), it is necessary to scan the directory structure to identify each skill.
Judgment Criteria (must all be met):
  1. Directory contains
    SKILL.md
    file
  2. SKILL.md
    contains
    name:
    field
  3. SKILL.md
    contains
    description:
    field
Scanning Process:
bash
undefined

Step 1: 获取根目录结构

Step 1: Get root directory structure

gh api repos/{owner}/{repo}/contents --jq '.[].name'
gh api repos/{owner}/{repo}/contents --jq '.[].name'

Step 2: 递归检查每个子目录是否包含 SKILL.md

Step 2: Recursively check if each subdirectory contains SKILL.md

gh api repos/{owner}/{repo}/contents/{potential-skill-dir}/{subdir}/SKILL.md
gh api repos/{owner}/{repo}/contents/{potential-skill-dir}/{subdir}/SKILL.md

Step 3: 解析 SKILL.md 获取 name 和 description

Step 3: Parse SKILL.md to get name and description

gh api repos/{owner}/{repo}/contents/{skill-path}/SKILL.md --jq '.content' | base64 -d | grep -E "^(name|description):"
gh api repos/{owner}/{repo}/contents/{skill-path}/SKILL.md --jq '.content' | base64 -d | grep -E "^(name|description):"

Step 4: 确认是有效 skill → 返回

Step 4: Confirm it is a valid skill → Return

{ "name": "frontend-dev", "description": "Full-stack frontend development...", "path": "skills/frontend-dev" }

**示例**:MiniMax-AI/skills 仓库

```bash
{ "name": "frontend-dev", "description": "Full-stack frontend development...", "path": "skills/frontend-dev" }

**Example**: MiniMax-AI/skills Repository

```bash

扫描发现 14 个有效 skills:

Scanned 14 valid skills:

✅ frontend-dev (name + description + SKILL.md) ✅ fullstack-dev (name + description + SKILL.md) ✅ minimax-pdf (name + description + SKILL.md) ...
✅ frontend-dev (name + description + SKILL.md) ✅ fullstack-dev (name + description + SKILL.md) ✅ minimax-pdf (name + description + SKILL.md) ...

输出格式:

Output Format:

📦 MiniMax-AI/skills (14 skills) 🎯 frontend-dev: Full-stack frontend development... 🎯 fullstack-dev: Full-stack backend architecture... 🎯 minimax-pdf: Professional PDF creation...

**给 AI 的分析 prompt**:
分析目录结构,识别潜在的 skill 目录:
验证步骤:
  1. 检查目录是否包含 SKILL.md 文件
  2. 解析 SKILL.md 是否有 name 和 description 字段
  3. 确认是有效 skill
目录结构: {gh_api_contents_json}
返回: { "valid_skills": [ { "name": "{skill-name}", "description": "{skill-description}", "path": "{owner}/{repo}/{skill-path}" } ] }
undefined
📦 MiniMax-AI/skills (14 skills) 🎯 frontend-dev: Full-stack frontend development... 🎯 fullstack-dev: Full-stack backend architecture... 🎯 minimax-pdf: Professional PDF creation...

**Analysis Prompt for AI**:
Analyze the directory structure and identify potential skill directories:
Verification Steps:
  1. Check if the directory contains SKILL.md file
  2. Parse SKILL.md to see if it has name and description fields
  3. Confirm it is a valid skill
Directory Structure: {gh_api_contents_json}
Return: { "valid_skills": [ { "name": "{skill-name}", "description": "{skill-description}", "path": "{owner}/{repo}/{skill-path}" } ] }
undefined

并行搜索示例

Parallel Search Example

bash
undefined
bash
undefined

关键词: "authentication"

Keyword: "authentication"

1. skills.sh API

1. skills.sh API

2. SkillHub CLI

2. SkillHub CLI

skillhub search authentication
skillhub search authentication

3. GitHub

3. GitHub

gh search repos "authentication" --topic claude-skill --limit 10

---
gh search repos "authentication" --topic claude-skill --limit 10

---

3️⃣ 结果处理

3️⃣ Result Processing

去重

Deduplication

name
字段去重,保留安装量最高的。
Deduplicate by
name
field, keep the one with the highest installation count.

排序

Sorting

优先级:
  1. 官方认证 - 大厂维护
  2. 安装量 - 社区验证
  3. 更新时间 - 维护活跃度
Priority:
  1. Official Certification - Maintained by major companies
  2. Installation Count - Community verified
  3. Update Time - Maintenance activity

输出格式

Output Format

安装命令规则

Installation Command Rules

统一使用 skills 命令安装(支持所有来源):
技能来源安装命令格式
skills.sh 单技能仓库
npx skills add {owner}/{repo}
skills.sh 多技能仓库
npx skills add {owner}/{repo}/{skill-id}
GitHub 单技能仓库
npx skills add {owner}/{repo}
GitHub 多技能仓库
npx skills add {owner}/{repo}/{skill-id}
原则
npx skills add
命令统一支持从 GitHub 或 skills.sh 安装。多技能仓库需指定具体 skill-id。
多技能仓库示例
仓库: legout/data-platform-agent-skills
包含技能:
  - data-science-eda
  - data-science-visualization
  - data-science-feature-engineering

安装命令:
  npx skills add legout/data-platform-agent-skills/data-science-eda
  npx skills add legout/data-platform-agent-skills/data-science-visualization
Unified installation using skills command (supports all sources):
Skill SourceInstallation Command Format
skills.sh single-skill repository
npx skills add {owner}/{repo}
skills.sh multi-skill repository
npx skills add {owner}/{repo}/{skill-id}
GitHub single-skill repository
npx skills add {owner}/{repo}
GitHub multi-skill repository
npx skills add {owner}/{repo}/{skill-id}
Principle: The
npx skills add
command uniformly supports installation from GitHub or skills.sh. For multi-skill repositories, specify the specific skill-id.
Multi-skill Repository Example:
Repository: legout/data-platform-agent-skills
Contains skills:
  - data-science-eda
  - data-science-visualization
  - data-science-feature-engineering

Installation Commands:
  npx skills add legout/data-platform-agent-skills/data-science-eda
  npx skills add legout/data-platform-agent-skills/data-science-visualization

单个结果

Single Result

🎯 {skill-name}
   {一句话描述}

📦 {来源} | 📊 {安装量}
安装方式:npx skills add {完整路径}
完整路径格式
  • 单技能仓库:
    {owner}/{repo}
  • 多技能仓库:
    {owner}/{repo}/{skill-id}
🎯 {skill-name}
   {One-sentence description}

📦 {Source} | 📊 {Installation Count}
Installation: npx skills add {full path}
Full Path Format:
  • Single-skill repository:
    {owner}/{repo}
  • Multi-skill repository:
    {owner}/{repo}/{skill-id}

多个结果

Multiple Results

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
📊 搜索结果:{关键词}(共 N 个)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

🟢 官方认证
───────────────────────────────────────
🎯 {skill-1}
   {描述}
   📦 skills.sh | 📊 50,000+
   安装方式:npx skills add {owner}/{repo}/{skill-id}

⭐ 社区推荐
───────────────────────────────────────
🎯 {skill-2}
   {描述}
   📦 GitHub | ⭐ 1,000+
   安装方式:npx skills add {owner}/{repo}

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
🎯 {skill-name} {一句话描述}
📦 {来源} | 📊 {安装量} 安装方式:{安装命令}
undefined
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
📊 Search Results: {Keyword} (Total N items)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

🟢 Official Certification
───────────────────────────────────────
🎯 {skill-1}
   {Description}
   📦 skills.sh | 📊 50,000+
   Installation: npx skills add {owner}/{repo}/{skill-id}

⭐ Community Recommended
───────────────────────────────────────
🎯 {skill-2}
   {Description}
   📦 GitHub | ⭐ 1,000+
   Installation: npx skills add {owner}/{repo}

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
🎯 {skill-name} {One-sentence description}
📦 {Source} | 📊 {Installation Count} Installation: {installation command}
undefined

多个结果

Multiple Results

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
📊 搜索结果:{关键词}(共 N 个)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

🟢 官方认证
───────────────────────────────────────
🎯 {skill-1}
   {描述}
   📦 skills.sh | 📊 50,000+
   安装方式:/install skills.sh:xxx

⭐ 社区推荐
───────────────────────────────────────
🎯 {skill-2}
   {描述}
   📦 SkillHub | 📊 10,000+
   安装方式:/install skillhub:xxx

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
📊 Search Results: {Keyword} (Total N items)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

🟢 Official Certification
───────────────────────────────────────
🎯 {skill-1}
   {Description}
   📦 skills.sh | 📊 50,000+
   Installation:/install skills.sh:xxx

⭐ Community Recommended
───────────────────────────────────────
🎯 {skill-2}
   {Description}
   📦 SkillHub | 📊 10,000+
   Installation:/install skillhub:xxx

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

渠道配置

Channel Configuration

📋 完整配置
references/channels.json
📋 Full Configuration:
references/channels.json

搜索方式优先级

Search Method Priority

API > CLI > Web
按配置文件中的
search_method
字段执行搜索。

API > CLI > Web
Execute search according to the
search_method
field in the configuration file.

官方优先原则

Official Priority Principle

大厂官方技能优先展示:
  • Anthropic - Claude官方
  • MiniMax - 国内AI大厂
  • 字节跳动 - 抖音/飞书官方
  • 阿里Qwen - 通义千问官方
  • Vercel Labs - Next.js官方
输出时使用 🟢 标识官方技能。

Official skills from major companies are displayed first:
  • Anthropic - Official Claude
  • MiniMax - Domestic AI major
  • ByteDance - Official Douyin/Lark
  • Alibaba Qwen - Official Tongyi Qianwen
  • Vercel Labs - Official Next.js
Use 🟢 to mark official skills when outputting.

安装辅助

Installation Assistance

bash
undefined
bash
undefined

查看工具状态

Check tool status

./scripts/install-helper.sh status
./scripts/install-helper.sh status

安装所有工具

Install all tools

./scripts/install-helper.sh all
./scripts/install-helper.sh all

安装单个工具

Install single tool

./scripts/install-helper.sh gh # GitHub CLI ./scripts/install-helper.sh skillhub # SkillHub CLI
undefined
./scripts/install-helper.sh gh # GitHub CLI ./scripts/install-helper.sh skillhub # SkillHub CLI
undefined