Cron

Cron — Scheduled Tasks & Recurring Job Orchestration

Activate When

• User invokes /godmode:cron
• User says "schedule a job", "cron job", "recurring task", "run every hour"
• User says "background scheduler", "periodic task", "timed execution"
• User says "node-cron", "BullMQ repeat", "Celery beat", "Sidekiq-cron", "APScheduler", "Hangfire", "Quartz"
• User needs a task to fire on a schedule (daily reports, cleanup, syncs, heartbeats)
• User asks about cron syntax, timezone handling, or overlap protection
• User needs distributed scheduling across multiple instances
• User needs rate-limited or priority-aware recurring execution
• Godmode orchestrator detects hardcoded setInterval or setTimeout loops that need conversion to proper scheduled jobs

Workflow

Step 1: Scheduling Requirements Assessment

Evaluate what needs to run, how often, and under what constraints:

SCHEDULING REQUIREMENTS ASSESSMENT:
|  Dimension            | Value                             |
|  Task type            |  |
|  Frequency            |  {
    // Only one scheduler instance becomes leader
    const acquired = await this.redis.set(

PostgreSQL Advisory Lock

import hashlib

def pg_advisory_lock_for_job(conn, job_name: str) -> bool:
    """Acquire a PostgreSQL advisory lock for a scheduled job.
    Returns True if lock acquired, False if another process holds it."""
    # Convert job name to a stable int64 for PG advisory lock

Step 5: Job Monitoring & Alerting

Track schedule health and detect missed or failing runs:

CRON JOB MONITORING:
|  Metric                    | Current | Alert    | Status |
|  Jobs scheduled            | 8       | —        | OK     |
|  Last run: daily-digest    | 09:00   | miss > 1h| OK     |
|  Last run: cleanup         | 06:00   | miss > 7h| OK     |

Monitoring Implementation

// Scheduled job health monitor
class CronJobMonitor {
  constructor(private redis: Redis, private alerter: Alerter) {}

  async recordRun(jobName: string, result: {
    status: 'success' | 'failure';

Step 6: Dead Letter Handling for Scheduled Jobs

Handle scheduled jobs that exhaust retries:

SCHEDULED JOB DLQ DESIGN:
|  DLQ for Scheduled Jobs                                   |
|  Queue:       scheduled-jobs-dlq                          |
|  Retention:   30 days                                     |
|  Alert:       Any entry (scheduled jobs should not fail)  |

Step 7: Priority Queues & Job Chaining

Design priority-aware scheduling and dependent job pipelines:

SCHEDULED JOB PRIORITIES:
|  Priority | Schedule         | Job               | SLA    |
|  P0       | */5 * * * *      | Health check       |  —  jobs, , "
6. If new scheduler: "Scheduled jobs configured.  jobs registered. Deploy and monitor."
7. If fixing failures: "Idempotency and locking added. Missed-schedule alerting enabled."
8. If migrating: "Migrated from  to . All schedules verified."

Key Behaviors

Never ask to continue. Loop autonomously until done.

# Validate cron expressions and test jobs
npx cron-validate "0 9 * * 1-5"
npm run test:cron -- --timeout 30000
redis-cli KEYS "bull:*:repeat:*" | head -10

IF job duration > 80% of interval: increase interval or optimize job. WHEN missed runs > 0 in 24h: alert and investigate. IF retry count > 3: move to dead letter queue.

1. Monitor every scheduled job. Missed runs, failures, duration.
2. Idempotency mandatory. Same result if called twice.
3. Distributed locking in production. No duplicate fires.
4. Overlap protection. Use max_instances: 1 or coalesce.

On failure: revert with git reset --hard HEAD~1.

Flags & Options

| Flag | Description | |--|--| | (none) | Full scheduled task design workflow | | --tech | Target specific scheduler (bullmq, celery, sidekiq, apscheduler, hangfire, quartz, node-cron) | | --diagnose | Diagnose missed or failing scheduled jobs |

Quality Targets

• Job success rate: >99% over 30 days
• Runtime per job: 3 consecutive failures

Keep/Discard

KEEP if: improvement verified. DISCARD if: regression or no change. Revert discards immediately.

Stop Conditions

Stop when: target reached, budget exhausted, or >5 consecutive discards.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: arbazkhan971
• Source: arbazkhan971/godmode
• License: MIT

Install and usage instructions live in the source repository linked above.