What problem does it solve? Getting the text content of a YouTube video normally means watching it in full or relying on YouTube's caption UI. This Skill downloads the transcript directly — trying manual and auto-generated subtitles first, and falling back to local audio transcription when no captions exist. ## Core Features & Use Cases - Subtitle-first download: Fetches manual subtitles, then auto-generated captions, and cleans VTT/SRT markup into flowing de-duplicated text. - Audio transcription fallback: When a video has no usable subtitles, downloads the audio with yt-dlp and transcribes it locally with mlx-whisper (Apple Silicon), including container-to-host SSH bridging. - Distinct exit codes: Separates "no subtitles" (2), "yt-dlp missing" (3), "yt-dlp failed" (4), and "empty caption track" (5) so failures are reported accurately. - Use Case: A user pastes a YouTube lecture URL and asks for the transcript; the Skill saves a clean text file named after the video, or offers audio transcription if the video has no captions. ## Quick Start Ask the assistant to download the transcript of a YouTube video by pasting its URL, optionally naming an output file or requesting the audio-transcription fallback.