Use for YouTube transcript questions. Fetch captions with yt-dlp, convert VTT to markdown, retrieve with qmd (`-c` collection required).