orbital-express: Skill for Claude Code

Add a cronjob

Cronjobs live in the top-level cronjobs.js (the clock process — only 1 instance runs it). A cronjob should not do work itself — it enqueues a background task on a schedule. Read README "The Entry Points" and "How a Job Gets Created".

Plan the cadence, timezone, idempotency, task payload, retry impact, and operational ownership and get sign-off before scaffolding or edits. Confirm the owning feature has complete structure with yarn repair <Feature> --dry-run, then yarn repair <Feature>.

Pattern

// cronjobs.js — grouped by feature with a comment header
const <Feature>Queue = queue.get('<Feature>Queue');

// <plain-English what + when>. Often: every 5 min in dev, real cadence in prod.
const schedule = NODE_ENV === 'development' ? '0 */5 * * * *' : '0 0 0 * * *';
new CronJob(schedule, () => {
  <Feature>Queue.add('V1<Action>Task', {}); // enqueue — the task does the work
}, null, true, 'UTC'); // ALWAYS 'UTC'

Rules

• Always 'UTC' as the timezone (last arg). All scheduling is UTC.
• The cron callback only enqueues — no business logic, no DB work inline. Put the work in a task (add-task skill).
• For large datasets use the aggregate → singular task pattern: the cron enqueues the aggregate task, which fans out one singular task per record.
• Cron expressions here are 6-field (sec min hour day month weekday).
• Group by feature with a comment header (match the existing style in cronjobs.js).

Steps

1. Scaffold the task with the add-task skill (yarn gen <Feature> -t V1<Action>Task) if it does not exist; never hand-create it. Register its processor in the feature's worker.js.
2. Add the new CronJob(...) entry to cronjobs.js under the feature's section.
3. Verify locally: run yarn cron (and yarn worker to process the jobs). Tasks are tested directly; the cron wiring itself is config.
4. Run yarn conventions:check, focused tests, then yarn test; finish with mandatory review-conventions.