So the thing is, it's not clear how something like Minish Cap was written or what tools were used to make it. The programmers might have written a lot of the game in raw machine language, though that seems rather unlikely. They might even have written it in C and used a compiler to produce the GBA machine language, but that also seems unlikely.
C programming was a thing on the GBA, though extensive inline assembly could be expected for anything whether it was a cash in on a film or a flagship title we all still debate the merits of to this day (though obviously you would expect more on something like Payback that pushed the limits of the hardware vs the latest Disney channel TV show cash in platformer). I don't know if the commercial compilers were leaked as part of the gigaleak, though you can find various tools and aspects of the dev kit from... actually before the GBA's release (see the story of the first emulators there).
Afraid I have not kept up at all with ARM decompilation efforts here, or indeed if things have gone general. That said IDA has some stuff here so I guess.
https://hex-rays.com/decompiler/
As far as detecting what language something was written in then it is usually not so hard. Even fancy modern compilers will still do things coders won't, and the GBA commercial efforts were not exactly fancy modern compilers that benefited from the last however many decades at this point of compiler theory pondering.
Those interested in decompilation as a practical* action I did cover it a bit in
https://gbatemp.net/threads/can-i-aka-you-as-i-cant-wont-code-port-this-game-to-this-device.576997/
*those wanting theory then... yeah it is still cutting edge research, you probably want to start with the halting problem (this sort of thing literally being in opposition to that), then maybe a dalliance with reversible computing and then how compilers work, "everything is adding" if you did not already know that one too probably wanting to be in there as a base mindset. The short version is while theoretically (and practically) information is lost in compilation that can't be regenerated automatically then between side channel leaks (see
Diablo source code recreation), though not all efforts will have this), compilers being fairly simple programs at their heart and game programmers tending to do optimised code (can't be wasting resources when you have pixels to push) then someone doing some cheats to figure out a few memory variables**, knowing the hardware locations and formats, probably being a talented relevant version of C and relevant assembly coder, can build up enough info and make enough educated guesses that... well Mario 64 decompilation literally compiles into the same game bit for bit that Nintendo released decades before (getting to a "for 99% of intents and purposes it is the same" being a bit easier still but people do seem to like that 1:1 line).
**know what health is and you now know functions that care about it, and probably in turn what the functions that don't care about it are more likely to be part of. Repeat for various things, maybe do some traditional debugging (move the camera and this section of assembly code that is linked to these lines of decompiled code runs) and you can make progress and learn things, including things that might now be known (there was a hidden cheat menu discovered a while back when some unknown code was looked at as part of a decompilation).