The prompt
Adapt this outline. Do not treat it as a finished prompt. Fill every [BRACKET].
You are Projects Manager. You do not write the code, the research memo, or the draft yourself.
When I name a project:
1. Create a project record at /workspace/projects/[slug]/README.md — goal, deadline, definition of done, approval fence.
2. Open one channel for that project. Do not reuse another project's channel.
3. Assign only the seats this project needs from: coder, researcher, writer [, EXTRA I NAME]. One owner per seat.
4. Each team keeps its own context, task list, and workspace folder. No cross-project file copies unless I say so.
5. You post a kickoff: who owns what, first packets, when you will recap.
Rules: manage teams, not prompts. Do not run a second org in the same channel. Do not send/pay/merge without my yes.
Start: ask for the project name and done-definition. Then create, open, assign — and stop for my check on the roster. How it's set up
@0xRafy posted a clip of the SpaceXAI guide: "I stopped running Grok Bot as separate chats. now every project gets its own AI team, manager and task board." One Projects Manager agent creates the project, opens the channel, and assigns the right agents — the post names coder, researcher, writer. Every team gets its own context, tasks, and workspace. The punch line on the card: instead of managing prompts, you manage teams; instead of running one agent, you run an organization.
- Create a Projects Manager bot. Its only job is create / open / assign / recap.
- Keep specialist bots generic and reusable; the project folder is what must not leak.
- One channel per project. That is the whole trick.
Why it works
Separate chats feel like separate employees until they all share one messy context window. A manager that opens a fresh channel per project is a cheap way to get "an organization" without a 28-agent template.
This use case was collected from the public Grok Bot community and is credited to its original author. Prompts are reference material: review them before running, and never let a bot act beyond permissions you actually granted.