OptionalmaxStep budget for each delegated run. Defaults to the sub-agent's own.
OptionalonCalled when a delegated run finishes, e.g. to log or meter it.
OptionaltrustedTag the sub-agent's answer trusted. Leave this off unless the sub-agent never reads outside
content: a sub-agent that read a hostile file can pass its instructions on in its answer.
The sub-agents the model may hand work to, by name.