# Seen or Said

> A light film-literacy page for anyone who has just watched a documentary. List up to 6 beats you remember, in your own words, and pick how the film showed each: filmed as it happened, old footage or a document, an interview, a narrator or on-screen text, acted out or reconstructed, 2 shots placed side by side, or not sure. The page sorts each beat into a class, gives the film a verdict by a fixed rule, picks the beat to check first and hangs the beats as film strips in an editor's trim bin. One metered run says what each beat actually shows, the leap the viewer takes on trust and a plain question to check it. It reads how a film shows things, never whether they are true.

URL: https://seen-or-said.skillsafe.ai/
API: https://seen-or-said.skillsafe.ai/api.html

## Contract
- One task: `beat-check`. Input fields (all strings): task, film, watched, beats (1 line per beat with its page class), facts (verdict, the beat to check first, the firmest beat, class counts).
- Reply: JSON with verdict, headline, beats[] (beat, known, shows, leap, ask), check_first, closing.
- Classes, closest to the beat first: seen, found, said, told, staged, cut, unclear.
- Verdicts: on-record, rebuilt, on-their-word, mixed.

## Rule
First match wins: seen and found make at least half the beats -> on-record; staged and cut at least half -> rebuilt; said and told at least half -> on-their-word; otherwise mixed. Check first: the beat furthest from what the camera saw, the earlier row on a tie.

## Care
The reply never judges whether anyone in a film is honest or guilty, never calls a beat true or false, and never suggests contacting anyone. A viewing aid, not fact-checking or legal advice.

## Sources
- Skill: @github/doublecheck on SkillSafe (https://skillsafe.ai/skill/@github/doublecheck).