Repeatedly audit repository changes for architecture drift and alignment with the user's actual success criteria. Use when implementing, modifying, refactoring, or reviewing code where god components, misplaced responsibilities, missing requirements, weak evidence, or premature completion are meaningful risks.
