What problem does it solve? When a user reports a wrong number on screen ("it says 5.7 but should be 8.25"), the bug could live in any of five data layers, and fixing the wrong layer wastes time. This Skill provides a systematic diagnostic chain to locate the exact layer holding the wrong value before any code is changed. ## Core Features & Use Cases - Layered diagnosis chain: Checks live form/day-state, brand profile pool, recipe pool, saved spec sheets, and the Auto-Fill planner in order, stopping at the first wrong layer. - SQL and file pointers: Provides concrete queries (e.g., inspecting brand_profiles.values jsonb) and file paths (e.g., profileAutofill.ts, storage.ts applySpecImport) for each layer. - Routing to fixes: Directs import/apply bugs to import-bug-investigation, widespread bad rows to data-heal-playbook, and derivation errors to the shared lib/* packages. - Use Case: A floor manager reports the dough batch yield is wrong for one customer. The Skill guides checking the derived form value, then the profile, then the dough recipe's doughballVariants, and finally the saved spec sheet row matching that customer's variantLabel. ## Quick Start Ask the assistant to triage why the web app shows a specific wrong value, stating the expected number and where it appears on screen.
