How can one simply return the response of the tool, instead of routing the response to a final result handler?

[pedroallenrevez]

I'm assuming a result_tool is always needed. In this case I want the tool to be another agent, and I just want the response from that agent, without any other LLM call.