resilience

Use when code has concurrent writers or retried operations — race conditions, check-then-act, optimistic vs pessimistic locking, distributed locks, retry idempotency. Broker delivery, outbox, DLQ → this plugin's event-driven skill.

Use when writing or reviewing exception and error handling — try/catch placement, error boundaries, propagation, custom error types, rethrow and cause chains, fail-fast vs recover — instead of catch-log-continue at every layer.

Use when designing or reviewing message-driven architecture — brokers/queues, topics and partitions, event schema versioning, delivery semantics, outbox, sagas, dead-letter queues, consumer idempotency. In-process framework queues belong to the stack plugin.

Use when writing or reviewing code that emits telemetry — logging, metrics, tracing, health checks, alerting, error reporting — to design structured, correlated, bounded observability signals instead of print-statement archaeology at incident time.

Use when code is measurably slow or heavy, or when reviewing a change for performance — measure-before-and-after, N+1, payload, bundle, Core Web Vitals, cache correctness (stampede, TTL, eviction), percentile load testing. Index/schema design belongs to database; statement shape to sql.

Use when code crosses a process boundary — HTTP calls, queues, databases, third-party APIs, background jobs — timeout, retry, degradation, and delivery-semantics decisions instead of inherited defaults.