No. Prions are not "infectious proteins", they are "transmissible protein folds". The universe of proteins which can be spontaneously misfolded by another protein that is already misfolded is relatively rare. It's also not very difficult to engineer. When I was in grad school some of my classmates tried (not all amyloid "prion" proteins are bad, some are structural, such as pMel which is used to sequester toxic melanin precursors and scaffold them into melanin), and it was difficult to improve upon nature.
It's very likely that every "prion-capable" protein is that way for a strategic reason (e.g. Alzheimer's protein is likely that way to ensnare viruses before they can further damage brains).
Not every protein needs to be capable of misfolding, life would be very metastable if every protein were, so, no, you will not turn into a "puddle of goo" via an engineered prion. It might really suck, and your brain (or pick organ here) would get holes drilled in it over the course of years. But not hours.
It may be difficult for humans to improve on nature. It may be easier for AIs. Keep in mind that the space explored by evolution is necessarily quite restricted, and very damaging entities would have been selected out if at all possible. What you can imagine based on the evolved structures you've seen may be more limited than you realize.
In any case, my previous comments were intended jokingly, they weren't submissions to Methods in Molecular Biology.
Uh ok. I mean AI is data limited. Molecular biology (outside of structural biology) is not exactly a discipline where a firehouse of uncoupled data emerge. It's not uncommon to find publications with 5 data points, n=3 on each.
