A clean wrapper around yt-dlp that defaults to giving you transcripts as readable paragraphs instead of raw subtitle files. It tries YouTube's transcript API first, then falls back to subtitles when needed. You get timestamped output only when you ask for it, and bracketed noise like [Music] is stripped by default. Beyond transcripts, it handles the usual yt-dlp tasks: downloading video, audio, subtitles, listing formats, and passing through custom flags. The subcommands have focused help screens, which is nice when you're not sure what options exist. If you're building agents that need to read video content or just want a friendlier interface to yt-dlp's transcript capabilities, this saves you from writing the same cleanup logic every time.
npx skills add https://github.com/steipete/agent-scripts --skill video-transcript-downloader