SuperAPI AI Directory

music-search

General & other

View repository

Updated 10-16 08:53

About this skill

Music Search Skill

Search music resources (songs, albums, lossless music), deeply crawl resource pages through real-time crawlers, and obtain publicly shared resource links from various netdisk platforms.

Prerequisites

  • web-search skill (required, used to search and discover resource pages)
  • Python 3(required, used for cloudscraper deep page crawling. The dependency is automatically installed on first run to .venv)

Command

Search resources

bash "$SKILLS_ROOT/music-search/scripts/music-search.sh" search "关键词" [选项]

Options:

Parameter Description Default value
--pan <type> Filter netdisk type: quark, baidu, aliyun, uc, all all
--format <f> Filter audio format: flac, ape, wav, dsd, hires, mp3, aac, all all
--limit <n> Maximum number of results per platform 5
--engine <e> Search engine: deep, web deep

Engine description:

  • deep(default, recommended) — web-search searches and discovers resource pages + cloudscraper deeply crawls to extract netdisk links and extraction codes, with the most accurate results
  • web — only extract links from web-search engine summaries (fast, but lower accuracy, no deep crawling)

Example:

# 搜索所有平台的音乐资源(默认使用深度搜索)
bash "$SKILLS_ROOT/music-search/scripts/music-search.sh" search "周杰伦 范特西"

# 只搜夸克网盘 FLAC 资源
bash "$SKILLS_ROOT/music-search/scripts/music-search.sh" search "周杰伦 范特西" --pan quark --format flac

# 搜索 Hi-Res 高解析音乐
bash "$SKILLS_ROOT/music-search/scripts/music-search.sh" search "邓紫棋 光年之外" --format hires

# 限制结果数量
bash "$SKILLS_ROOT/music-search/scripts/music-search.sh" search "林俊杰 她说" --limit 10

# 使用浅层搜索(不深度抓取页面)
bash "$SKILLS_ROOT/music-search/scripts/music-search.sh" search "陈奕迅 十年" --engine web

Use the PowerShell script on Windows:

powershell -File "$SKILLS_ROOT/music-search/scripts/music-search.ps1" search "周杰伦 范特西" --pan quark

Popular recommendations

bash "$SKILLS_ROOT/music-search/scripts/music-search.sh" hot "2025年热门专辑"

Essentially, it calls search with recommendation-type keywords and returns related netdisk resources.

Resolve redirect links

bash "$SKILLS_ROOT/music-search/scripts/music-search.sh" resolve "https://example.com/goto/xxx"

When links in search results require a second redirect, use this command to resolve the real netdisk address.

Agent usage flow

  1. The user says "help me find XXX's music" → run search "XXX"
  2. Parse the returned JSON and extract netdisk links
  3. Present them to the user grouped by netdisk type and audio format
  4. If the results contain an extraction code (extractCode), display it to the user as well
  5. If the results contain pageUrl but no direct url, use the resolve command to obtain the real address

Example conversation:

User: Help me find lossless resources for Jay Chou's Fantasy album, FLAC format required

Agent:

  1. Run search "周杰伦 范特西" --format flac
  2. Extract matching items from the JSON results
  3. Return: title + format + netdisk link + extraction code (if any)

Output format

All commands output JSON with the following structure:

{
  "success": true,
  "data": {
    "query": "周杰伦 范特西",
    "total": 5,
    "results": [
      {
        "title": "资源标题",
        "pan": "quark",
        "url": "https://pan.quark.cn/s/xxx",
        "format": "FLAC",
        "extractCode": "ab12",
        "source": "deep-search",
        "pageUrl": "https://example.com/resource/123"
      }
    ]
  }
}

Field descriptions:

  • pan: netdisk type — quark(Quark), baidu(Baidu), aliyun(Ali), uc(UC), magnet(Magnet)
  • format: detected audio format — FLAC, APE, WAV, DSD, Hi-Res, AIFF, MP3, AAC, OGG, an empty string means not detected
  • source: result source — deep-search(deep crawl), web-search(search summary)
  • url: netdisk share link (directly accessible)
  • extractCode: extraction code/password (if detected)
  • pageUrl: result source page URL

Configuration

Edit music-search/.env to customize preferences:

# 偏好的网盘类型(逗号分隔,排在前面的优先展示)
MUSIC_SEARCH_PREFERRED_PAN=quark,aliyun,baidu,uc

# 深度搜索开关(推荐开启)
MUSIC_SEARCH_DEEP_ENABLED=true

# 每次搜索最多访问的结果页面数
MUSIC_SEARCH_DEEP_MAX_PAGES=6

# 并发抓取页面数
MUSIC_SEARCH_DEEP_CONCURRENCY=4

Notes

  • For searches, it is recommended to useChinese singer name + album name or song name, which works best
  • Deep search (--engine deep) takes 2-3 seconds longer than shallow search, but the results are much more accurate
  • Python cloudscraper will be installed automatically on first use to .venv(requires Python 3)
  • Deep search requires the web-search skill to be enabled; the first search may require waiting for the browser to start
  • Cloud drive share links may become invalid at any time; it is recommended to use them as soon as possible

Disclaimer

This tool is for personal learning and research use only. It does not host, distribute, or provide any copyrighted content, and only aggregates publicly accessible links from search engines. Please comply with the laws and regulations of your region.