reference-video-analysislisted

# Reference Video Analysis Treat the reference as a complete communication system, not a bag of quotable lines. The result must let an unseen reader understand what the video says, why the audience keeps watching, and which mechanisms can transfer into an original work. ## Boundary - Accept a video URL, transcript, SRT/VTT/JSON subtitle file, notes, or an archived source package. - For a URL, use the current Mode's installed retrieval/transcription route first. Do not add another downloader, login store, or discovery router. - Analyze only the supplied source. Do not crawl the creator's other work unless the user asks. - Keep distinctive phrases, identity, personal anecdotes, and factual claims source-owned. Transfer mechanisms, not a sentence-replacement template. ## Prepare complete reading context When a transcript or subtitle file exists, run: ```bash python scripts/prepare_video_analysis.py transcript.srt \ --title "Reference title" \ --source-url "https://example.com/video" \ --output ./analysis-input.json ``` The prepared input preserves timestamps, every readable sentence, and each segment's previous/next context. Use it as a coverage ledger, not as the final report. If no transcript can be obtained, state the visual/audio evidence actually available and lower the confidence instead of filling gaps from memory. ## Analyze for decisions Cover the following questions at a depth proportional to the source. Combine related findings when that reads better; d