What problem does it solve? Recurring preventive operations work — backups, restore drills, secret rotations, compliance sweeps — is easily forgotten or run ad hoc, leaving deployed products to rot silently. This Skill defines how to schedule, dispatch, and audit those tasks so nothing is missed and every run leaves verifiable evidence. ## Core Features & Use Cases - Seven Task Types: Covers backup, restore-test, BCDR-drill, patch-cycle, secret-rotation, flag-cleanup-audit, and compliance-sweep, each with a default cadence and owning workflow mode. - Calendar-Driven Dispatch: Reads config/upkeep-calendar.yaml plus the append-only upkeep-ledger.jsonl to report which tasks are DUE, with priority rules (a failed backup blocks restore-test; a failed restore-test blocks release promotion). - Failure Routing: Maps each failure type to the correct owning agent (devops, stabilize, security, train, audit) instead of fixing in-line. - Use Case: When investigating a missed quarterly BCDR drill, run the calendar dispatch script to see last-run status, then check the ledger for the recorded RTO/RPO actuals and route the failure for root-cause analysis. ## Quick Start Ask the agent to run the upkeep calendar dispatch script against your repository and list which upkeep tasks are currently DUE.
