Requesting Status and Trajectory Persistence

Two ways to learn about the progress of a child agent.

Via Messaging: ask “how’s it going?” asynchronously or receive proactive status_update messages on the bus; express state as a single dictionary (executing/requires input/completed/failed - as in A2A).

Via File: the child agent serializes the trajectory to JSONL in real time (analogous to a pre-write log - state is always reproducible, crash recovery, audit, transfer to another agent) or maintains a lightweight progress.md.

Hang detection: the time of the last entry does not change for more than N minutes - a timeout safeguard.

Trajectory is all agent state: the prefix is given by the code, there is no runtime state apart from the trajectory.