Pulls transcripts from YouTube videos and saves them as plain text files. It tries yt-dlp first for speed, handling cookies and multiple subtitle languages (Chinese and English), then falls back to browser automation with Chrome DevTools MCP if the CLI fails. The browser path is thorough but slow, it clicks through the description dropdown to find the transcript button, then extracts content via DOM scripting to avoid accessibility tree overhead. Nice touch that it converts VTT and SRT files into clean timestamp-plus-text format. If you're doing research, content analysis, or just need searchable video content, this handles the annoying parts. Works best if you already have yt-dlp installed.
npx skills add https://github.com/feiskyer/claude-code-settings --skill youtube-transcribe-skill