If one were to think of this as a project, the initial project goal might be a proof of concept, that such an undertaking though non-trivial may be worth while.
for me it is currently quite tough, since I don't know the internals at all
It would be desirable to act on the abstract syntax trees that result from the compiler parsing the source code and not the source code itself.
Template Haskell does that - it provides quotations, quasi-quotations and splicing, along with other utilities. I would guess that a simple design would be to add an import_syntax clause, similar to import, but importing only declarations with type ...-> Q ... and splicing automatically their appearances. This is a back of the envelope design, and I haven't considered any potential side-effects, but sounds like a reasonable approach