youtube-digest

The youtube-digest skill is an advanced automation toolkit designed for researchers, students, and content consumers who need to extract insights from long-form YouTube video content efficiently. By integrating yt-dlp for media analysis and structured prompt engineering for summarization, it transforms hours of video into clear, actionable markdown documents. The skill handles the entire lifecycle of video analysis: extracting metadata, retrieving and correcting transcripts, generating deep insights, and testing user comprehension through a systematic 9-question quiz spanning three difficulty levels.

• Automated transcript extraction with priority handling for manual Korean or English subtitles over auto-generated versions.

• Semantic correction of proper nouns and technical terminology using integrated WebSearch context.

• Generation of structured summaries including key insights, actionable takeaways, and full translated transcripts saved as local markdown files.

• Multi-stage comprehension testing with three difficulty levels (Basic, Intermediate, Advanced) designed to verify retention.

• Deep Research integration for follow-up investigations into specific topics or entities mentioned in the video.

• Uses yt-dlp for reliable metadata and transcript acquisition, supporting cookies from local browsers for authenticated content.

• Input requires a YouTube URL; the output is a formatted file in the research/readings/youtube directory.

• The quiz system uses AskUserQuestion to deliver immediate feedback and explanations for incorrect answers.

• Prioritizes accuracy by replacing common transcript misidentifications (e.g., cloud code -> Claude Code) through cross-referencing web search results.

• Ideal for creators, developers, and researchers managing large volumes of video-based information.