When should you use it?
- You keep re-explaining the same bot job every conversation
- You want one place for scope, quality bar and hard stops
- You are splitting a catch-all bot into focused ones
Not for: A single profile that tries to own every company workflow; Putting secrets or passwords inside the profile file; Skipping the never-do list because you trust the model
What do you need, and what do you get?
You need
- Bot name and one-line ownership
- Sources it may pull (dashboards, docs, inboxes)
- Actions that always require your approval
You get
- A profile draft with pull, quality, never-do and learned sections
- A first weekly job that fits the profile
- Open questions the bot must ask before acting
How does it work, step by step?
- Name one jobOne bot, one ownership line. Nothing else.
- Fill four blocksWhat I pull, what good looks like, what I never do, what I learned (leave room).
- Add approval waitsAnything that spends, sends, publishes or deletes waits for you.
- Save and reuseEdit the file over time instead of rewriting the chat prompt.
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)
Draft a durable profile for one focused bot. Name: {bot_name}. Ownership line: {ownership}. Sources it may use: {sources}. Hard waits (always ask me): {approval_waits}. Output these sections: WHO I AM — one paragraph. WHAT I PULL — bullets with source names only from the list above. WHAT GOOD LOOKS LIKE — format, length, and that every claim needs a source link. WHAT I NEVER DO — include the hard waits plus anything out of scope. WHAT I LEARNED — leave an empty placeholder for later notes. FIRST JOB — one concrete weekly task that fits. Do not invent sources. Do not include secrets.
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?
- Shared user memory can be updated by any bot; keep sensitive preferences out of it.
- Quality bars need numbers and source links or claims get fuzzy.
- Leave the learned section empty on day one so the bot has room to append.
- Do not paste API keys or passwords into the profile.
Our take
Chat windows invite disposable instructions. Carnage's useful reframe is a lasting profile document closer to an AGENTS.md than a throwaway prompt: scope, inputs you pull, quality bar, hard never-dos and room for learned notes. Focused bots keep that file sharp; catch-all bots make it mush. The four blocks are what to copy: what the bot pulls, what good work looks like, what it must never do, and what it has learned, with that last section left empty on day one so it has room to grow. Give each bot one job and one ownership line; a profile that tries to own every workflow turns vague quickly. Quality bars need numbers and source links, or claims get fuzzy. Anything that spends, sends, publishes or deletes should wait for your approval, and that rule belongs in the file. Keep API keys and passwords out of it entirely. Shared user memory can be changed by any bot, so keep sensitive preferences out of that too. Editing the file over time beats rewriting a chat prompt every week.
Where the time goes back: You stop re-briefing the same bot every morning; the profile carries scope and hard stops forward.
FAQ
How is this different from a system prompt?
Why focused bots?
What belongs in never-do?
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. Focused profile; approval waits; no secrets.
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 24, 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.