This is good, but zero-knowledge is actually a stronger property than just hiding the "knowledge" needed to complete the proof. A zero-knowledge proof should n

This is good, but zero-knowledge is actually a stronger property than just hiding the "knowledge" needed to complete the proof. A zero-knowledge proof should not reveal any knowledge to the verifier, in the sense that the verifier should not be able to compute anything after the proof that it could not have computed beforehand. So among other things, Bob will not only be unable to prove knowledge of where Waldo is, but also not be able to prove that Alice has that knowledge. Obviously that is not true of non-interactive zero-knowledge, which is why NIZK should not be confused with "actual" zero knowledge.