I don't think this is a good idea. A module should encapsulate complexity. Ideally, I shouldn't have to understand the inner workings to use it. I don't think this proposal achieves either goal.

Given the interface you propose, I'd have to understand the Schwartzian transform to provide the correct arguments. Once I know that, it's more efficient (especially given the fact you're using a string eval to build the transformators) and shorter just to write it inline.