jordanrendric/claude-video-vision — reverse-engineered prompt

Reverse engineered prompt

GitHub

Build me a Claude Code plugin that lets Claude understand videos I give it. I want to be able to point it at a local video file or a YouTube link and ask normal questions like “summarize this,” “what happens in the first few seconds,” or “what text is on screen at 1:30.”

The plugin should pull out video frames so Claude can actually see what is happening, and it should also transcribe the audio with timestamps. It should support YouTube captions when they exist, and fall back to audio transcription when they don’t. Let me choose between using Gemini, OpenAI Whisper, or a local Whisper setup, with a simple setup command that checks what I have installed and walks me through missing pieces like ffmpeg or API keys.

Please make it usable inside Claude Code with a slash command like /watch-video, plus conversational detection when I mention a video path or URL. Keep the video processing layer focused on preparing frames, metadata, and transcripts, then let Claude do the interpretation.

Want more depth? Deep Reverse