Inject prompt only when tool is being called

[noobmaster19]

I have many cases where the certain lengthy prompts are only needed when the agent decides to call the tool. I.e. the agent only needs to know of these context when he decides to use that tool, is there a good way to do this without using nested agents?

[noobmaster19]

I am doing it by using the .describe() within the input schema. Is this the right way to do this?

My objective is:

1. If the tool is not selected, the main agent should not waste context on the details needed to use the tool.
2. If the tool is selected, the main agent NEEDS to know of the context inorder to use the tool.

The details are very lengthy hence the constaint.