Skip to content
StellarFirmStellarFirm
Mission manual
Esc

Type a word to search every page. Try , or .

Module 02 · The CEO

How the CEO works

You are the CEO and the CEO assistant works for you. It reads your message, writes a brief, picks the assistant, plans shared work, and reports back.

View as Markdown
On this page

You are the CEO. The CEO assistant is the assistant that works for you. You never have to know which assistant should take a job, how to write it up, or who should hear about it next. You say what you want, and the CEO assistant does the organising.

The CEO assistant plans and briefs. It does not do the Coder's work, and it does not make your decisions. Anything external still waits for your Approve.

What happens to your message#

  1. It reads your message. It works out what you are asking for: building, fixing, reviewing, a status update, or a question about the company.
  2. It picks an owner. One role per piece of work. Today that is the Coder. See routing.
  3. It writes a brief. A short, self-contained note for the assistant: the goal, the limits, how to know it is done, and what waits for your Approve.
  4. The assistant works in its own private workspace. You can watch in Jobs.
  5. Risky steps come to you. They appear on Approvals. See approvals.
  6. It reports back in the same chat: what was done, with links, and anything left for you.

What a good brief contains#

The CEO assistant writes briefs for the Coder in a fixed shape, so nothing important is missing.

PartWhat it says
GoalThe outcome you can see when it is done.
Repository and branchWhere the work happens.
TicketThe issue it comes from, if there is one. It is never invented.
Acceptance criteriaA checklist the result can be tested against.
TestsWhat to add or update.
ConstraintsWhat not to touch, and any deadline.
What waits for ApproveOpening the pull request, pushing a branch, and anything else risky.
ReportWorking first, then Done with a link.

The more of this you give in your goal, the less the CEO assistant has to ask.

When work needs more than one assistant#

Some goals need several roles, for example a launch that needs code, copy, and a privacy notice. When that happens, the CEO assistant plans the work as a short sequence:

  • The goal, in your words, with a success check.
  • Up to six steps, ordered by what depends on what.
  • For each step: the assistant, what it receives, and what it hands on.
  • Every step that sends, posts, publishes, or spends, marked as waiting for your Approve.
  • A report as each step moves from Working to Done, and a short summary at the end.

Assistants can also pass work and questions to each other in a shared chat, but each handoff is bounded and anything that writes outside the team still waits for you.

When the CEO assistant asks you#

If it is not sure who should own a job, or the request is too vague to build, it asks you one question instead of guessing. That is by design. A short answer from you is better than a wrong pull request.

Prompts to try#

PromptHand over a goal

CEO, I want [outcome] on [repository]. Brief the right assistant, tell me what will need my approval, and report back when it is done.

PromptAsk for a plan first

CEO, before anyone starts, give me a short plan for [goal]: the steps, who does each, and what will wait for my approval.

PromptAsk what you need to decide

CEO, what is waiting on me right now, and what do you suggest I decide first?

Next#