Skip to content

README.md

Latest commit

 

History

History
87 lines (60 loc) · 3.43 KB

README.md

File metadata and controls

87 lines (60 loc) · 3.43 KB
Raw

Supadata Skill

skills.sh

An agent skill that teaches your AI agent to use the Supadata API for video transcripts, social media metadata, web scraping, crawling, and AI-powered structured data extraction from videos.

Once installed, your agent will know to:

  • Fetch transcripts / captions / subtitles from YouTube, TikTok, Instagram, X (Twitter), Facebook, and direct video file URLs
  • Pull unified metadata for any social media video or post
  • Get YouTube channel, playlist, video, and search data
  • Scrape any web page to clean Markdown
  • Crawl a whole website
  • Map all URLs on a site
  • Extract structured JSON from videos using a prompt or JSON Schema

Install

npx skills add supadata-ai/skills

That's it. The skill auto-activates when the user mentions a video URL, asks for a transcript, requests web scraping, etc.

Install for a specific agent

npx skills add supadata-ai/skills -a claude-code
npx skills add supadata-ai/skills -a cursor
npx skills add supadata-ai/skills -a codex

Skills work with 60+ agents including Claude Code, Cursor, Codex, GitHub Copilot, Windsurf, Gemini, OpenCode, and more.

Install globally

npx skills add supadata-ai/skills -g

Setup

Export your Supadata API key — get one at dash.supadata.ai:

export SUPADATA_API_KEY="..."

The skill assumes this env var is present. Without it, the agent will ask the user to set it before running any command.

What the skill contains

skills/supadata/
├── SKILL.md                         # entry point: when to activate, decision tree, quick recipes
├── references/
│   ├── video.md                     # universal endpoints: /transcript, /metadata, /extract
│   ├── youtube.md                   # YouTube-specific: channel, playlist, search, translation, batch
│   └── web.md                       # /web/scrape, /web/crawl, /web/map
└── scripts/
    ├── transcript.sh                # one-shot transcript fetch (handles async jobs)
    ├── scrape.sh                    # one-shot scrape to Markdown
    └── poll-job.sh                  # generic async-job poller

The skill follows the Agent Skills spec and uses progressive disclosure: SKILL.md is small and decision-oriented, with detailed endpoint docs in references/ loaded only when the agent needs them.

Other ways to use Supadata

This skill is the lightest-weight option — your agent makes plain HTTPS calls and you keep no extra dependencies.

For larger projects:

Links

License

MIT — see LICENSE.