Using Julia to install system wide CUDA stack?

It is amazing how a single ]add CUDA automatically installs a private CUDA stack in a pain-free way. Sometimes however I need a CUDA installation for non-julia applications, like pytorch etc. Is there a way to get a “global” CUDA installation using julia? For instance, a script that copies the julia CUDA artifacts to /usr/lib or whatever is the appropriate library directory for the current platform?