Search YouTube for learning topics or research evidence, get transcripts/captions from YouTube videos, optionally download the video as an MP4 plus a separate audio file, and run restartable queue-based retrieval with sleep/retry handling for rate limits. Use when Codex needs to search YouTube from a query/results URL, transcribe one or many YouTube URLs, save caption text/SRT/JSON, inspect available caption languages, resume interrupted retrieval, build a research/evidence source workflow, or download YouTube media files with yt-dlp for videos the user is authorized to save.