Managoat

Managoat is the hosted Fountain. Fountain is the open-source engine, and its name is on the CLI, the API, the SDK and this manual. Everything here applies to Managoat unless a page says it is for a self-hosted server.

create-team

A five-question conversation that proposes a roster of teammates, creates them, and hands them over.

At a glance
Bundled Yes, in each sandbox.
Declared by Nobody. It is always there.
Source apps/fountain/priv/sprite_skills/create-team/SKILL.md
Triggered by /create-team, or a request to a teammate for help with a team.

What happens when you run it

Say /create-team to any teammate. Or ask it what the team must look like.

It asks at most five questions, one at a time. It offers a default with each one, so "yes" is a complete answer.

  1. What must the team get done?
  2. Which repos or services are in scope, and which need a token?
  3. How many teammates, roughly? Two to five is usual. Past three, a lead earns its place.
  4. Any brain preference? It reads GET /api/catalog for the models, and your inference credentials for the providers you hold a key for. It proposes nothing you cannot run.
  5. What names? Role names such as repo-maintainer, or a themed set.

Then it proposes. You get a table of name, brain, a one-line role, and the first thing it would send each teammate.

It creates nothing before you say yes.

Then it hands over. You get the roster as created, and who to message for what. You also get one concrete first message for each teammate, and a note that teammates can reach each other.

What it creates

For each teammate it creates an Agent, then a team membership, with POST /api/team. The membership opens that agent's one conversation and provisions its sandbox.

When the team has a job that repeats, such as triage or a daily report, it can also create a schedule for a teammate. That is a cron line and a prompt, and the teammate runs it on its own.

You can edit everything it sets afterwards, from the team app or with the Team routes. Brain, role and name are not decisions you are stuck with.

Its own rules

The skill holds itself to four rules. They are the difference between a helpful flow and a destructive one.

  • Ask one question at a time. Stop as soon as a proposal is possible.
  • Never create before a confirmation. Never delete a thing, and never rename a thing.
  • Never paste a token into a prompt, a description or a message.
  • On a name conflict, add a suffix and say so. On an absent credential, name the provider and say where to add the key.

Limits

It only adds. It will not remove a teammate, rename one, or reorganise a team that already exists. The team app does those, and so do DELETE /api/team/:agent_id and PATCH /api/team/:agent_id. Read Team.

It needs your inference credentials in place first. It reads GET /api/account/inference-credentials, then tells you what is absent. It does not create agents that cannot run. Read Inference credentials.

Each teammate it creates provisions a sandbox. A roster of five is five machines. Read Change sandbox lifetimes.