Use before writing any code that reads, writes, or validates data touching the database — new fields, forms, API inputs, migrations. Finds the actual schema (column type, nullability, length, enum values, constraints, defaults) from the real source of truth, not from memory or assumption, then enforces validity at whichever layer the repo actually validates at (input/form, API boundary, service/transaction, or DB constraint). Trigger on "what type is this column", "can this be null", "add validation for X field", "what values are valid here".