
from claude-watch69
Turn tutorial or lecture videos into structured, timestamped study notes with scene-aware screenshots and transcripts.
Claude Watch automates watching a video (local path or URL), extracting scene-aware frames, obtaining a timestamped transcript, and producing a structured markdown notes artifact saved to a persistent library. It downloads videos with yt-dlp, detects scene changes via ffmpeg, pulls captions or uses Whisper/Groq fallback, and writes a strict notes template that includes TL;DR, key concepts, per-scene notes with embedded screenshots, code blocks, and open questions.
Use Claude Watch when you want concise, navigable study notes from a tutorial, lecture, conference talk, or screen recording. Trigger it by supplying a video URL or path and an optional topic (e.g. /claude-watch <url> [topic]). It is useful for students, engineers reviewing API walkthroughs, and content teams who need extractable highlights from long videos.
Inferred compatible tools: Claude-vision (for frame reading), Claude Code or other LLMs for synthesis, and local shell/Bash for running yt-dlp and ffmpeg. The SKILL.md references AskUserQuestion for interactive key prompts and expects the agent to run local scripts.
This skill has not been reviewed by our automated audit pipeline yet.