Pre-Generation Qualification of Conditional Persistence-Image Targets Under Sparse Support: Code and Derived Evidence

Reproducibility compendium for a failure-aware pre-generation qualification workflow. It contains analysis code, frozen configurations, tests, group-level derived evidence, and deterministic numerical-figure inputs. The joint calibration and conditional-direction rule failed, so no generated-image efficacy or method-superiority claim is made. AI Hub source images, licensed example crops, unread reserves, and outer endpoints are not redistributed.