What problem does it solve? Scheduled routines can fail silently, leaving failures sitting in logs nobody reads until real damage surfaces. This Skill turns a routine that can die quietly into one whose failures reach the human with enough context to act. ## Core Features & Use Cases - Failure Definition: Establishes up front what counts as a failed run — nonzero exit, missing output, or output that fails validation — before the first alert fires. - Severity Split and Escalation: Separates page-worthy failures from log-only events, escalates on consecutive failures, and handles flapping without paging once per cycle. - Canary Verification: Fires a deliberately failing test run to prove the alert actually arrives through the notification lane with routine name, date, failure reason, and log link. - Use Case: A nightly data-sync routine stopped writing output three days ago and nobody noticed. Use this Skill to define its failure states, wire alerts to the existing notification channel, and verify the path with a failing canary. ## Quick Start Add failure alerting to my nightly backup routine so I get paged with the failure reason and log link whenever a run fails.
