What problem does it solve? Silent failures — suppressed exceptions, rolled-back batches, dead connections, stuck retries — leave systems looking healthy while data vanishes. This Skill provides principles, checklists, and anti-patterns for making every failure class visible, attributable, and locatable in minutes instead of days. ## Core Features & Use Cases - Failure Visibility Design: Ensures every exception path logs with correlation ids and increments honest counters, forbidding bare except blocks and contextlib.suppress patterns. - Metrics and Watchdog Contracts: Defines required signals for bounded resources (queue depth, overflow counts, retry rates) and watchdogs that alert on the absence of expected events like heartbeats and batch completions. - Use Case: When reviewing a background batch writer or SSE event queue, apply the checklist to verify drop counters increment at the point of failure, health checks exercise the real data path, and per-step timing answers which stage got slow. ## Quick Start Review my queue-based batch writer design using the observability skill and identify which failure classes lack visible signals.
