Agent skill · software engineering · jeremylongshore
adobe-reliability-patterns
Analyze and design idempotency, bounded polling, per-service circuits, durable queues, DLQ/replay, degradation, and reconciliation for Adobe workloads. Use when the task requires adobe async reliability controls. Trigger with "Adobe reliability", "Firefly retry design", or "PDF job recovery".
Why this skill is useful
Adds executable scripts for implementing reliability patterns specific to Adobe APIs, including circuit breakers and fallback mechanisms.
What it needs
Requires bullmq, opossum, redis installed locally. About 2k tokens when loaded. Last updated 2026-08-07. 2,604 stars on the source repository.
What this skill does
Adobe Async Reliability Controls Overview Analyze and design idempotency, bounded polling, per-service circuits, durable queues, DLQ/replay, degradation, and reconciliation for Adobe workloads. This workflow produces a reviewable artifact and evidence before any live side effect. Prerequisites Current first-party Adobe documentation for every selected service, API version, auth flow, limit, and lifecycle. Named product, identity, security, data, budget, release, and operations owners appropriate to the scope. Synthetic or approved non-production fixtures with secret and content canaries. Current Contract Adobe services fail independently and async submission may succeed before the caller sees a response. Reliability is a state machine: planned, approved, submitted, acknowledged, running, terminal, artifact verified, downstream acknowledged, and cleaned. Unknown is not failed or safe to replay. Recheck the dated evidence map before relying on mutable product behavior. Authentication Idempotency records and job evidence use aliases and hashes, never tokens, signed URLs, or content. Credential failure and content/policy failure have different containment owners. Instructions 1. Map operations, side effects, states, identifiers, retryability, time budgets, dependencies, and reconciliation sources. 2. Create per-service circuits and queues so one Adobe product does not exhaust all workers. 3. Persist submission intent before network call and bind returned job/status/cancel evidence atomically. 4. Poll returned status URLs with jitter, terminal-state validation, cancellation, and strict elapsed-time budget. 5. Route exhausted/unknown work to quarantine or DLQ; replay only after reconciliation and approval. 6. Exercise crashes, timeouts, 429/5xx, revocation, vendor outage, duplicate events, and recovery objectives. Tool Discipline Use Read, Glob, and Grep to inspect current documentation, configuration, code, fixtures, and evidence. …
How to use it
Reference it in AdaL, Claude Code, Cursor or any coding agent — nothing to install:
@skills jeremylongshore/adobe-reliability-patterns