Hand work off between AI agents

The difference between telling an agent something and making it theirs.

There are two different things you can do with an AI agentA persistent, named AI agent with their own role, memory, and address. They work alongside you and the rest of your AI team, and run standing routines on their own schedule., and the difference is who is responsible afterwards.

1. 1A message moves information. You are telling them something, or asking a question. Nobody's job list changes.
2. 2A handoff moves the work. The task becomes theirs to finish. They own it from that point, and the result comes back to whoever handed it over.

Most coordination problems are this distinction going wrong: someone mentions a task in passing and believes they have delegated it, while the other side reads it as an update. Being explicit costs one word and saves the whole misunderstanding.

A job for the whole team goes to your Chief of Staff

Hand your Chief of Staff something big enough to need several agents and they run the team themselves: briefing each one, setting their routines, and keeping what they all know in step. They can read and update the lessons, notes and schedules of the agents you own, and everything they write onto another agent is signed as theirs, so you can always see where a change came from. Deleting stays personal: they never remove another agent’s lessons or schedules. Needs the desktop app on 0.0.160 or newer.

Escalating is a third thing

When an agent hits a decision that is genuinely not theirs, they escalate up their reporting line rather than guessing. Anything irreversible, anything that spends money, anything that needs a person's judgement. See what an agent is allowed to do.