The agent doesn't work as expected when "skip_summarization" is set to True in AgentTool

[slilichenko]

In the code, the skip_summarization of Agent Tool causes the parent context to emit an event with "is_final_response()" returning True. This causes base flow interruption, which means that the loop ends with the last function response. But most of the UIs, including ADK web, don't display the output of the tool response and it appears that the agent is stuck. Furthermore, if the agent is supposed to continue working on the response, perhaps by calling other tools with the output of the first AgentTool call as input - it does't happen because the response is marked as final.

The documentation is a bit confusing because a) it doesn't cover these nuances and b) if you open the code snippet the "how it works session" covers what's not in the code (it describes what the way things works with skip_summarization = False).

IMHO, it's a code bug or a design flaw, or a documentation bug. A reproducible case: if you update this simple agent to use skip_summarization - the answer is never displayed to the user via ADK web interface. The skip summarization path generates 4 events vs 5 with summarization; the last event being the function response (and the function response text is never displayed in the ADK web UI). Happy to provide detailed screenshots if needed.

[BnnaFish]

Also, for me, tests Eval works fine, but Agent is exposed as A2A and the response is empty. I have a very simple agent that calls only one tool. Plus tracing shows good response, but in the caller agent its empty.