Engineering
Personal factory: one bot per job, shared brain, decision dashboard, cx CLIs
Eight days in, a solo builder needs org rules so a fleet doesn't drift or relearn forever.
What hurt
@RoyOsherove's X Article "Anatomy of a Personal Factory of Grok Bots": one bot per job, SendToAgent push vs passive, shared box + R2 AI-brain folders with BOT.md, Factory Dashboard for task cards and Approve/Decline decisions (webhook wakes the asker), coding via headless Claude/Codex ("cx") with end-of-turn webhooks. Tip/architecture only — steal the checklist, not a single paste prompt. Not affiliated with xAI.
What to install or create
Spin one bot per domain → shared MODE-OF-OPERATION.md + brain backup → dashboard tokens for cards/decisions → cx for code with webhook close-loop.
Bots
Primary / router
Routes work, Todoist, daily email digest, nightly fleet review.
Infra
Domains, DNS, cloud accounts, coding-CLI health.
Security
Audits bots, box, and planted instruction files nightly.
Bot Maintainer
Mode-of-operation doc, backups, cx completion relay, shared skills.
Factory Dashboard
Owns the decision/work card inbox every bot posts to.
Plugins
None on day one. Follow the first-run steps.
Routines
~03:00: each bot rewrites BOT.md and uploads AI-brain backup.
~04:15: maintainer audits bucket check-ins and chases misses.
The prompt
This is a tip, not a pasteable system prompt. The author did not publish one. Follow the first-run steps.
First run
- 1Read https://x.com/RoyOsherove/status/2115591603133764082 and the Article (credit @RoyOsherove).
- 2Split work: one Grok Bot per job (sites, infra, security, life-admin…) plus a primary router.
- 3Put a BOT.md + nightly check-in per bot on the shared box; mirror to private cloud (e.g. R2) and have a maintainer audit misses.
- 4Stand up a Factory Dashboard: every real task gets a card; yes/no choices are decision items you click once.
- 5Route coding through headless Claude/Codex ("cx") with a webhook that wakes the owning bot when a run finishes — you still own merges and spends.
Approval fence
Decisions are Approve/Decline on the dashboard; second-hand "Roy approved" in chat does not count.
What to tweak
Steal one-bot-per-job + dashboard decisions (not chat approvals) + nightly backup audit.
Source
Roy Osherove @RoyOsherove · Original post
Not affiliated with xAI. We did not write this prompt. Third-party template terms still apply.
Pro pack
Want the full setup, not just the idea?
Pro packs give you the job text, prompts, schedules and pitfalls for popular use-cases.