Condition Agent have pre provided user, and system messages that aren't removable

[AHMED-salah00]

I'm having a hard time trying to remove the predefined user, and system messages that are generating wrong answers:

When the wrong path is choose and I checked the flow messages I see messages that I didn't define, or add:
All the messages in the red frame aren't provided by me. They were there for no reason, and there is no option to remove them.
The only messages provided by me are the ones in the blue frame.

Please help me out if you can.

[HenryHengZJ]

These are predefined prompt for Condition Agent.

You can't just simply say "Choose the correct flow", please try to improve your instructions. LLM needs to understand the context, what you're trying to achieve

[AHMED-salah00]

I want the llm to choose whether to classify the input as asking to create a product, or a general question. It is a simple task that it can determine based on the user message.

The following scenarios should be editable based on the use case, and it will lead to a wrong classification:
1- user is asking about Ai.
2- user is not asking about Ai.

Is there a way to change them? There should be a way, or am I not understanding a specific point?