Calibration Vignette Bank — Review Walkthrough

The four generation rules every set follows

1. Matched sets, not isolated scenes. Each look-for uses one fixed lesson context per band, rendered at all four levels. Only the practice varies — so the difference between a 2 and a 3 is never contaminated by a change of lesson, grade, or content.
2. Every bullet becomes an observable event. Each vignette instantiates its level's descriptor bullets as scriptable evidence: exact teacher moves, student words, counts, minutes. If a bullet isn't visible in the scene, that's a defect — flag it.
3. The discriminator line. Between adjacent levels sits one sentence naming the single piece of evidence separating them. This line is what a rater who crossed that boundary most needs to read.
4. Direction-aware at runtime. The dashboard pairs the rater's level with the anchor's level and frames the comparison by direction (hot: "here's what your score would have required"; cold: "here's what to credit when you see it"). Bands are served by the walk's grade/course: K–5 → elementary, 6–12 → secondary. Foundational Skills is K–3 by design and has no secondary band.