Imported from lost-rob0t/skills (skills/youtube-context/SKILL.md). Install upstream with npx skills add lost-rob0t/skills --skill youtube-context. Copyright stays with the author.

YouTube Context

Goal

Retrieve a compact, normalized transcript packet for a YouTube or other HTTP(S) video so it can be supplied as model context.

Input

Pass one HTTP(S) video URL to scripts/youtube-context.

Output

The executable writes the packet to standard output in this exact format:

Title: <title>
URL: <url>
Transcript source: <video captions or local Whisper (base.en)>

Transcript:
<normalized transcript>

It prefers English human or automatic captions. If the selected caption file is missing or empty, it downloads the best audio and transcribes it locally with Whisper. Errors go to standard error and return a non-zero status.

Dependencies

• yt-dlp for metadata, captions, and audio;
• Python 3 for deterministic JSON3/WebVTT parsing and whitespace cleanup;
• whisper and ffmpeg only for the caption-free fallback.

The executable accepts configuration through YOUTUBE_CONTEXT_* environment variables. The defaults match the packet consumed by the Emacs YouTube context command.