bullmq

name	bullmq
description	BullMQ 5 — Redis-backed Node.js queue. Jobs, workers, flows, repeating, rate-limit, sandboxed processors, observability. Use when: bullmq, bull mq, taskforcesh, Queue, Worker, QueueEvents, FlowProducer, JobScheduler, repeatable job, cron, delay, backoff, priority, lifo, attempts, removeOnComplete, sandboxed processor, useWorkerThreads, sequentialize, rate limiter, parent-child, child jobs, dependencies, bull-board, bullmq-pro, getJobCounts, drain, obliterate, QueueScheduler (removed), graceful shutdown. SKIP: legacy Bull (→bull-legacy cascade), Agenda, Inngest / Trigger.dev (managed SaaS), Celery (Python), Sidekiq (Ruby).
stacks	["bullmq","queue","nodejs-backend","redis"]
packages	["bullmq","@bull-board/api","@bull-board/fastify","@bull-board/express","@bull-board/hono"]
tags	["queue","bullmq","redis","background-jobs","workers","cron","flows"]
manifests	["package.json"]
source	vechkasov-global-skills
risk	high-stakes

🎯 Version Requirements (May 2026)

Primary pins:

• BullMQ: 5.x
• Redis: 8.x
• Node.js: 24.x (Active LTS)

Source of truth: STACK_VERSIONS.md — verified 2026-05-16

Usage

Loaded automatically when its description matches the active task. Read only the section you need.

Use this skill when

• Designing background job queues on Node.js backed by Redis (BullMQ 5+)
• Setting up Queue, Worker, QueueEvents, FlowProducer instances + Job Schedulers (upserted on the Queue, not a standalone class)
• Configuring job options: delay, attempts, backoff, priority, lifo, removeOnComplete, removeOnFail, jobId
• Implementing repeating / scheduled jobs (cron + every-N-ms via queue.upsertJobScheduler — replaces deprecated Queue.add repeat options)
• Designing parent–child flows with FlowProducer (waits for children to finish before parent runs)
• Setting concurrency, rate-limiting (limiter), sequentialize (one-at-a-time per key)
• Wiring sandboxed processors (separate process or useWorkerThreads) for CPU-bound jobs
• Adding observability: QueueEvents listeners, Bull Board dashboard, OpenTelemetry tracing
• Implementing graceful shutdown for workers (SIGTERM → finish current job → exit)
• Migrating from Bull (legacy) to BullMQ — different API, no QueueScheduler, JS-only → TS-friendly
• Auditing existing BullMQ code for common pitfalls (stalled jobs, missing attempts, no idempotency)

Do not use this skill when

• Task is legacy Bull (the JavaScript predecessor) — use bull-legacy (cascade marker)
• Task is Agenda (MongoDB-backed) — different paradigm, different skill
• Task is Inngest / Trigger.dev / Defer / Hatchet — managed SaaS, different skill
• Task is Sidekiq (Ruby), Celery (Python), Resque — different runtime