integration-messagequeue-designer_skill

Overview

This skill designs message queue patterns for RabbitMQ, Kafka, SQS, Azure Service Bus and Google Pub/Sub, producing practical configs, producer/consumer snippets, and monitoring guidance. It focuses on reliability features like dead-letter queues, idempotency, ordering guarantees, and backpressure for production-ready event-driven systems.

How this skill works

Provide a scoped design based on requested pattern, target queue system, and delivery guarantees. It validates inputs, checks system capabilities (for example exactly-once support), then emits topology config, minimal producer/consumer code, retry/DLQ rules, and monitoring/alerting recommendations. For advanced scopes it adds scaling, backpressure settings, and security/compliance controls.

When to use it

• Designing publish-subscribe or work-queue architectures for microservices
• Implementing sagas, CQRS event sourcing, or async request-reply flows
• Enforcing delivery guarantees (at-least-once, exactly-once where supported)
• Configuring DLQs, retries, and idempotency for resilient processing
• Optimizing throughput and ordering via partitioning, sessions, or keys

Best practices

• Validate that the chosen broker supports requested guarantees; downgrade with idempotency if not
• Always publish a unique message-id or idempotency key and persist processed IDs for consumers
• Use DLQs with clear redrive policies and metrics to avoid silent failures
• Prefer partition-key ordering instead of global ordering at high throughput
• Set sensible prefetch/MaxConcurrentCalls and autoscale based on queue depth or consumer lag

Example use cases

• Kafka publish-subscribe with transactional producers for exactly-once ordering per key
• SQS FIFO queue for ordered payment processing using MessageDeduplicationId
• RabbitMQ work-queue with prefetch_count and DLX for worker retries and poison message handling
• Azure Service Bus sessions for correlated request-reply flows with session affinity
• Pub/Sub with ordering keys and flow-control limits to prevent consumer overload