Water Damage Claim Pack — AI Re-Judges Every Photo Against 8 Evidence Slots

In the hours after water damage, homeowners rush to clean and repair — and destroy the evidence their insurance claim depends on. Claims assessment requires multi-layered compositions: room overview, mid-shot and close-up of the damage, the water's entry path, the boundary between wet and intact areas, and a scale object for size. Most people take only close-ups, restore the scene, and lose everything else. When the assessor later flags what's missing, the scene is already gone — re-shooting is impossible. This 'evidence loss' can contribute to reduced or denied claims, and it happens because people don't know what to photograph until it's too late.

An agent that re-judges the whole slot set on every upload. The user adds photos one at a time; each time, Gemini 2.5 Flash evaluates all 8 evidence slots fresh — a photo that satisfies nothing changes nothing. Missing slots are shown as 'the next shot to take' with an arrow overlay on the photo and a one-line instruction. When AI can't detect what's in a photo, the owner can attest it themselves — but the pack records it as owner attestation, structurally separated from AI-verified evidence, never merged. Once the set is complete, the app emits an inspectable ZIP: photos, captions, checklist, an incident-report draft, and manifest.json carrying each judgment's rationale.