async-messaging

Think Through Async Work & Messaging

Overview

Production guidance for asynchronous work. Each reference paper captures what breaks after deployment: poison jobs that loop forever, unacknowledged redeliveries, missed cron executions, event consumers that cannot tolerate duplicates or reordering, and email providers whose bounces and outages dominate the incident channel.

Core principle: Async work is a correctness surface. Every queue, job, and event needs defined delivery, ordering, idempotency, retry, retention, replay, and backlog semantics — with bounds and lag visibility.

Domain Law

NO ASYNC WORK OR MESSAGING CHANGE WITHOUT:
1. the minimum required primary paper(s) for the mechanism selected from the context table;
2. before solution details, the five-entry contract completed for every async
   unit — even for a consumer-focused request: producer durable intent or
   acceptance plus publication recovery; consumer durable effect and
   idempotency; post-effect acknowledgment; lease/loss/redelivery protection;
   and ambiguous-outcome lookup or reconciliation. Mark an unprovided producer
   path `unknown` with impact; broker delivery is not publication-durability evidence;
3. "Existing-codebase checks" run when changing existing workers;
4. every applicable MUST mapped to a decision with bounds (attempts,
   payload, concurrency, retention), a test, or a documented exception.

When to Use

Use this skill when thinking through, reviewing, changing, or verifying:

• background jobs, workers, progress reporting, and cancellation;
• retry policies, backoff, poison-job handling, and dead-letter queues;
• scheduled/cron jobs, missed-run behavior, duplicate-run prevention, leader election;
• queues, topics, pub/sub, streams, and consumer groups;
• delivery guarantee selection (at-most/at-least-once) and consumer acknowledgment;
• domain and integration events: schemas, versioning, ordering, fan-out, replay;
• transactional outbox/inbox to avoid dual writes;
• batch processing with chunking and resumability;
• bulk operations with partial-failure reporting;
• deduplication windows and scopes;
• transactional email delivery, bounces, provider failure;
• notification infrastructure across channels.