Why wouldn't you? "Curse of Turing" usually implies "if you implement a VM capable of running a Python runtime then you can do what Python can do", but Prolog is nothing like that:
> "You can write anything in python, from quick and dirty replacements for bash scripts"
SWI Prolog: can do shell scripts with shebang lines, args, reading from stdin, for 8+ years: https://www.swi-prolog.org/pldoc/man?section=plscript
Directory and file interactions: https://www.swi-prolog.org/pldoc/man?section=files
OS integrations for shelling out or environment variables: https://www.swi-prolog.org/pldoc/man?section=system
and regex support: https://www.swi-prolog.org/pldoc/doc_for?object=section(%27p...
> "to windows desktop applications"
SWI Prolog has a builtin cross-platform GUI library/toolkit: https://www.swi-prolog.org/packages/xpce/examples.html
> "clustered supercomputer code"
Not sure about supercomputing, but SWI has:
Threading: https://www.swi-prolog.org/search?for=thread&in=lib&match=su...
Transparent Inter-Process Communications (TIPC) libraries "a framework for cooperation between federations of trusted peers that are operating as a unit. It was developed by Ericsson AB" - https://www.swi-prolog.org/pldoc/doc_for?object=section(%27p...
Addons which include an MPI library port: https://www.swi-prolog.org/pack/list
> "It also has an library ecosystem to match."
SWI won't have anything like Python's ecosystem due to many fewer users, but it has a non-toy ecosystem of networking code, image processing, database connecting, more performant datastructures, Protocol Buffers, Unicode, YAML, SGML, XML, Http, zlib, gzip, unit testing, debugging, profiling, FFI, etc.
I mean, you might not want to, but it's not like you have to start from half a dozen Turing Machine instructions and an infinite tape and work your way up from there.