I think they'd tell us what we're allowed to assume?
Maybe if it looks like they forgot the safest way would be to answer the essence of the question first, and then bring in input verification as a later add-on feature?
I mean really the course is about computability when you get to that chapter, not "machine design". The machines are just a convenient shorthand.
Certainly that's how I'd handle it. Take valid input as given, prove computability on that basis, and then come back later to stick on some input verification later if there's ever time.
And if there is time, you might as well, I suppose.