When should you use it?
- You are creating your first agents and keep naming them Assistant
- You want a small roster that maps to real weekly work
- You need business and personal roles without inventing an org chart
Not for: Replacing a real org design or headcount plan; Assuming every role named in the article exists on your account today; One mega-bot that owns sales, ops, engineering and errands together
What do you need, and what do you get?
You need
- Your top weekly recurring jobs (meetings, inbox, shipping, home admin)
- Which connectors or tools each job already uses
- Two candidate role names you are willing to try for two weeks
You get
- A short roster of two narrow bots with one-line jobs
- What each bot owns and what it must leave alone
- A first task you will give each bot this week
How does it work, step by step?
- List your weekWrite the recurring jobs you actually do, not aspirational automation.
- Match two rolesPick two narrow roles from working examples (sales notes, ops seating, bug repro, demo prototypes, content watch, grocery compare) that map onto that list.
- Name the boundaryFor each role, write one sentence for what it owns and three things it must never touch.
- Run one job eachGive each bot a single concrete task this week before you add a third role.
The prompt
Our own prompt, inspired by Carnage (@0xCarnagee) and written to work in Grok Bot, ChatGPT agent, Claude and others. Values in lime are placeholders.
Fill in the blanks (optional; the copied prompt uses your values)
Help me choose a narrow agent roster for my week. My recurring jobs: {weekly_jobs}. Tools I already use: {tools}. I will start with exactly two bots. 1) Propose two narrow roles that map onto those jobs. Prefer proven shapes like call-notes, invoice desk, bug repro, bookmark-to-prototype, channel watch, or grocery compare. 2) For each role write: one-line job, what it owns, three never-touch rules, and the first concrete task to run this week. 3) Reject any catch-all Assistant. If a job does not fit either role, leave it for a later bot. 4) Do not invent connectors I did not list.
Inspired by Carnage (@0xCarnagee): X Article: The Grok Bot Team's Own Workflow (11 steps). The prompt above is written by playbooks.bot in our own words; nothing from the original is quoted.
What can go wrong?
- A catch-all Assistant never builds sharp memory; keep each bot narrow.
- Do not invent roles you cannot feed with real inputs this week.
- Half a roster can be non-work errands; that is intentional practice, not a joke.
- Names quoted from the article are Carnage's attributions, not our interviews.
Our take
Blank Assistant bots usually stay vague. Carnage's article is useful because it treats roster design as theft from working examples: sales notes, ops seating, bug repro, demo prototypes, content watches and errands. We keep the playbook to picking two that match your week and writing hard boundaries. Names of SpaceXAI or Cursor people appear only as he attributes them in the piece, not as interviews we ran. In our dry run with a fictional founder week (sales calls, invoice inbox, marketing drafts, Saturday groceries), the agent proposed a call-notes bot and an invoice-desk bot, rejected a catch-all Assistant, and listed first tasks with clear never-touch lines. It left groceries for a later errand bot instead of stuffing four jobs into one profile. Use this before you write long system prompts. Revisit the roster after two weeks: keep what finished a real task with evidence, and retire roles that only produced chatter. Two focused bots beat one clever name.
Where the time goes back: You skip days of renaming a vague Assistant and start with two bots that already know what job they own.
FAQ
Why copy roles instead of inventing them?
Can personal bots sit next to work bots?
How many should I start with?
Source and credit
Inspired by Carnage (@0xCarnagee). We describe the idea in our own words, link to the original, and wrote the prompt ourselves; nothing is quoted.
- Inspired by
- X Article: The Grok Bot Team's Own Workflow (11 steps)
- Creator
- Carnage (@0xCarnagee)
- Published on
- X
- Original date
- Aug 19, 2026
- License
- Not openly licensed (the poster keeps all rights), so we quote nothing and link to the original
- Originally for
- Grok Bot
- AI dry run
- Oct 23, 2026: passed. Dry-run tested by our AI test agent on sample input. Two narrow roles; boundaries clear; no invented connectors.
AI dry run on sample data; not run with live connected accounts. The playbooks.bot test agent filled the prompt with synthetic inputs, produced the output and checked it against the page's stated outputs and guardrails. - Verified
- Oct 23, 2026: source link resolves to the original, the credited creator published it, our excerpt or prompt matches it, and the license allows our use.
Are you the creator? Claim or correct this page.