What problem does it solve? When a system fails, teams often grab the newest backup or a folder named 'golden' without proving it can actually be restored, leading to failed recoveries and lost data. This Skill enforces evidence-based selection of a known-good rollback target before any restore begins. ## Core Features & Use Cases - Candidate Classification: Sorts backups, versions, commits, packages, snapshots, and installers into VERIFIED ROLLBACK, PROMISING CANDIDATE, BACKUP ONLY, REJECTED/FAILED, or UNKNOWN classes. - Identity and Integrity Verification: Records version/build/commit identity and checks hashes, manifests, signatures, and runtime/dependency/schema compatibility. - STOP Conditions: Halts selection when identity cannot be established, evidence contradicts the target, or recency is being substituted for acceptance proof. - Use Case: After a bad deployment, use this Skill to inventory available snapshots and commits, verify which one was owner-accepted and runtime-verified, and hand the exact proven target to the restore process. ## Quick Start Ask the AI to evaluate the available backups and commits for the failed release and select a verified rollback target with its evidence.