Is it possible to send a question posed to the agent executor directly to the chain used as tool?

[Onturenio]

I have a custom agent executor that uses a tool to solve certain questions. The tool is a chain that knows how to solve and interpret the question, so it returns directly. Therefore I'd like that the input to this chain is the original input to the Agent executor, not its "interpretation" of what the input should be. By letting the latter, some details are lost and the reply of the tool is not totally accurate.

I have been experimenting with a memory so I have a shared piece of knowledge exchanged between the chain and the agent executor, but the problem is becoming overly complicated, with custom and complicated prompt generation... so I'm guessing there must be an easier way to do this. Is there?

Thanks.