An analysis that lives only in this chat dies with it. The viewer turns an analyzed video into a single HTML file anyone can open — no server, no internet, no Watch Skill install on the other end.
Generate the page
watch-skill viewer <video_id-or-source> [-o <path>]
The file contains, self-contained (frames inlined as data URIs, zero external requests):
- a timeline with key-frame cards and timestamps,
- the transcript and all on-screen text (OCR),
- every question asked about the video so far, with its answer, confidence badge, and the exact timestamped evidence cited.
Answer questions FIRST, then generate. The page includes cached answers, so ask what the user cares about before rendering — a viewer generated before any questions holds only frames and transcript.
Hand-off patterns
- Bug ticket: run
watch-skill extract bug-reportfirst, then the viewer — the ticket gets the structured report, the viewer link gives reviewers the evidence. - Review request: send the file itself (it is one .html, typically well under a few hundred KB); it opens in any browser, RTL text renders correctly, nothing phones home.