my agents capture user data to give some
my agents capture user data to give some advise based on defined context. Agent ask clarifying questions when there is not enough info. this is all good and working as expected. however, user may type random data at times that would make sense to validate. For example, user may ask about possiblity of early retirement and do not provide desired retirement age. In this case, agent is asking "When do you want to retire". now a user may end up typing 100 which is kind of practically a non sense response.
In a traditional form based input you can validate all this and ask user ensure good data. but here in this agentic era, I am not sure where do you put this kind of validation? ideally it makes sense to validate as soon as user typed 100 in this case but as the flow is not deterministic so i wont know what question agent may ask (there are many inputs agent expects if missing).
Do I write somewhere validation rules within inputschema but then i am not sure would agent reject a invalid response as i would ideally like agent to prompt for the correct response. Any thoughts?
In a traditional form based input you can validate all this and ask user ensure good data. but here in this agentic era, I am not sure where do you put this kind of validation? ideally it makes sense to validate as soon as user typed 100 in this case but as the flow is not deterministic so i wont know what question agent may ask (there are many inputs agent expects if missing).
Do I write somewhere validation rules within inputschema but then i am not sure would agent reject a invalid response as i would ideally like agent to prompt for the correct response. Any thoughts?
