What problem does it solve? Building API integrations that survive real-world failures is hard: duplicate webhook deliveries, transient 503s, cascading downstream outages, and silent errors all break naive integrations. This Skill provides a structured protocol for designing and generating production-hardened API automation workflows where every call has a timeout, every write is idempotent, and every failure is typed and observable. ## Core Features & Use Cases - Resilience Pattern Taxonomy: Classifies workflows (single call, sequential chain, fan-out, event-driven, long-running, bidirectional sync) and applies the right patterns: idempotency keys, exponential backoff, circuit breakers, dead letter queues, sagas, and outbox. - TypeScript Implementation Generation: Produces code using ky/fetch, p-retry or cockatiel for retry and circuit breaking, BullMQ for queues, and OpenTelemetry spans for tracing. - Observability & Quality Gates: Enforces spans per external call, structured retry logs, latency metrics, and alert thresholds, with quality gates banning fire-and-forget calls and infinite retries. - Use Case: A user asks to build a Stripe webhook processor that handles duplicate deliveries. The Skill designs an idempotency-key flow backed by Redis, adds typed error classification, retry with backoff, and OpenTelemetry spans around every external call. ## Quick Start Ask the AI to design a resilient, idempotent workflow for chaining two API calls with retry logic, a circuit breaker, and observability spans.
