What problem does it solve? Reliability-sensitive systems often break when dependencies, processes, deploys, or infrastructure fail, and design reviews frequently miss cascade risks, shared-fate components, and unsafe rollouts. This Skill provides a structured review framework that keeps critical paths working through dependency failures and limits blast radius when changes go wrong. ## Core Features & Use Cases - Structured Reliability Review: Evaluates designs across five principles—isolation, redundancy, static stability, recovery practice, and progressive delivery—with a ready-to-use output template covering critical path, failure modes, and recommendations. - Concrete Review Heuristics: Provides actionable guidance such as keeping non-essential dependencies out of critical paths, preferring last known good state over live reads, requiring bounded retries with dead-lettering, and rehearsing rollback before relying on feature flags. - Use Case: When reviewing a worker that calls an external API before writing to the database, the Skill produces a review stating that ingestion must not depend on optional enrichment, that last known good enrichment should be preserved with freshness timestamps, and that retries must be bounded, idempotent, and dead-lettered. ## Quick Start Review this background job design for fault tolerance, covering the critical path, dependency isolation, static stability, recovery practice, and rollout plan.
