Pulls YouTube transcripts using yt-dlp with a sensible fallback chain: tries manual subtitles first, then auto-generated, and only hits Whisper transcription as a last resort after asking permission. The workflow is thorough, it checks dependencies, lists available subtitle tracks before downloading, and includes Python one-liners to deduplicate the VTT output since YouTube's auto-captions repeat lines across overlapping timestamps. Useful when you need text from videos for analysis or documentation. The Whisper path is properly gated behind confirmations since it downloads audio and installs models, which is the right call. Handles the boring stuff like filename sanitization and cleanup automatically.
npx skills add https://github.com/michalparkola/tapestry-skills-for-claude-code --skill youtube-transcript