Patch updated and attached.
> The repository does contain a tests/ folder, did you try running these?
I've ran pretty much all the tests and quite frankly, they are manual, human-graded tests. There is no unified way to tell if a test passed, etc. That might be why there is no test target.
> How about this Texinfo markup?
I modified your suggestion a little to more resemble the specs' language: http://quadium.net/funge/spec98.html
> it's actually gpl3+, what gives?
Mea culpa. I mistyped that.
> I don't know where the blame lies.
The tio.run links linked from those two Befunge-98 examples use the FBBI fingerprint, which currently isn't supported by cfunge. I tested a few examples from  https://esolangs.org/wiki/Befunge  and everything seems to work fine.
    On Sunday, August 9, 2020, 04:50:37 PM EDT, Jakub Kądziołka <kuba@kadziolka.net> wrote:  
 
 On Fri, Aug 07, 2020 at 05:02:40PM +0000, Hendur Saga wrote:
> I added a new file, esolangs.scm, because I could not find any other place to be more appropriate, for this and forthcoming packages.

Not a bad idea, though toys.scm somewhat fits. Please add the new file
to the list in gnu/local.mk, though.

Also, could you run your code through etc/indent-code.el?

> +  (arguments
> +    '(#:tests? #f)) ; no make target 'test'

The repository does contain a tests/ folder, did you try running these?
You can check the rgbds package from assembly.scm for some inspiration
as to how you can handle a custom testsuite entrypoint.

> +  (inputs
> +    `(("ncurses" ,ncurses)))
> +  (home-page "https://github.com/VorpalBlade/cfunge")
> +  (synopsis "Fast conforming Befunge93/98/109 interpreter in C")
> +  (description "cfunge is a fast conforming Befunge93/98/109 interpreter written in C.
          @command{cfunge}

> +It supports several fingerprints.")

It would be nice to define what a fingerprint is, as even though there
was once a summer when I was fluent in Befunge, I still needed a longer
moment to recall what those are. How about this Texinfo markup?

@dfn{fingerprints} (opt-in language extensions identified by a
four-character code).

> +  (license license:gpl2+)))

COPYING and the headers in src/*.c (well, I checked diagnostic.c)
suggest that it's actually gpl3+, what gives?

Could you send an updated patch?

Regards,
Jakub Kądziołka

P.S. I tried to test the package by running this CGCC answer in it, but
it loops infinitely. I didn't have time to debug this, so I don't know
where the blame lies.

https://codegolf.stackexchange.com/a/150920/55934