Queue Retry Contract Repair

The problem

Background job behavior often drifts across queue configurations, worker code, and documentation, leading to silent failures or expensive duplicate side effects. Manually tracing retry logic through producers, consumers, and dead-letter queues is error-prone and risks triggering production side effects during testing.

What it does

• Analyzes repository files to map background job contracts across producers, workers, and configs.
• Identifies drift between executable code, documented policies, and test fixtures.
• Standardizes time units, backoff rules, and attempt counts using repository-supported patterns.
• Drafts safe repository patches to centralize retry policies and idempotent completion logic.
• Generates deterministic local tests using fake timers and mocks to verify fixes without production access.

Frameworks & tools

Works with any repository containing background job logic, including tools like BullMQ, Celery, Sidekiq, RabbitMQ, and Kafka. Uses Read, Edit, and Bash tools to interact with local codebases.

Why this beats prompting it yourself

This skill enforces a strict safety boundary, ensuring you never accidentally trigger production side effects like duplicate payments or emails. It automates the tedious logical tracing of job lifecycles, separating transient retries from terminal failures more reliably than a general-purpose prompt.

Use cases

• Fixing mismatched retry counts between workers and queue configurations.
• Converting inconsistent delay units into a single standardized duration format.
• Adding idempotency guards to workers to prevent duplicate side effects after crashes.
• Aligning job-status APIs so retrying tasks aren't reported as terminal failures.

Known limitations

Cannot connect to live brokers, production databases, or cloud dashboards. It does not verify in-flight jobs or prove exactly-once delivery in production environments.