◇ Could this help me?
Yes, with one limit stated clearly. The five roles are well chosen, particularly the one asking whether you are solving the right problem and the one responding with zero context, because those are the two questions people closest to an idea reliably fail to ask themselves, and having the advisers critique each other rather than answer in parallel is what turns it from a list of opinions into an argument. The limit: all five voices come from the same model, so they share its blind spots and its training, which makes this structured self-critique rather than independent review. It will catch a weak argument and an unexamined assumption, and it will not catch something the model itself does not know. General practice.
Do not give me a single answer to this. Take the idea below and respond as five separate advisers, each with one job: the first hunts for the fatal flaw, the second challenges whether I am solving the right problem at all, the third takes the outside view and tells me what usually happens to attempts like this, the fourth reacts as an ordinary person with zero context, and the fifth cares only about the next concrete step. Then have each adviser read the others and name the blind spot in each. Finish as a chair, weighing the disagreement and giving me a clear call with the reasoning. Push back on me rather than validating me, and flag anything none of the five could reasonably assess.