it could be defined using output descriptor language, yes.
but output descriptors describe outputs… it’s in their name…
without the information contained in the scanned transactions, we cannot describe any outputs… so the definition of the output descriptor would need to be very clear that no outputs can be generated by the descriptor alone.