api-resilience-engineeringlisted

# API Resilience Engineering Use this skill for synchronous APIs, SDK calls, webhooks, background submission endpoints, and service-to-service integrations. Inspect the protocol, operation semantics, client and server implementations, proxies, deadlines, quotas, storage, side effects, telemetry, and deployed versions before changing resilience behavior. Preserve published compatibility unless the user explicitly accepts a breaking change. Resilience controls are a coupled system. A retry changes load and duplicate risk; a timeout changes ambiguity; a circuit breaker changes availability; a rate limit changes fairness. Define their combined contract rather than adding each mechanism independently. ## Define the operation contract For each operation, record: - read, deterministic write, non-idempotent action, or asynchronous submission; - authoritative completion evidence and how a client reconciles an unknown outcome; - end-to-end deadline and budgets for queueing, connection, request, processing and response; - retryable and permanent failure classes; - idempotency or deduplication scope, identity, fingerprint, retention and response semantics; - caller, tenant, organization and global quotas plus burst behavior; - concurrency and resource limits, fallback and load-shedding policy; - observable signals, ownership, rollout and rollback. Do not infer safety from an HTTP method alone. Verify actual side effects, including audit records, billing, notifications, downstream c