I'm going to offer a non-technological suggestion. As several answers have already observed, you're fighting some impossibilities (not the least of which is "I have been assigned to improve the audio quality in a school auditorium," presumably without any significant budget, just the expectation that you can create acoustics by tweaking a few pots). I'm also going to assume that having speakers remove their masks is not an option for policy reasons, and that we're not addressing here how one might upgrade the room or system if the money or will existed.
Yes, you should mess with the EQ; you should do so anyway to find what works best for the space. But the idea that you can take a signal that's muffled by the mask and "clean it up" through EQ is like the familiar trope from sci-fi and police shows: The team views a blurry picture of the crime scene, the commander yells "Enhance," and suddenly... we can identify the face of the killer! It's laughable because you can't just create data that was never there.*
Instead, I'm going to suggest that you do as much as you can to improve the human element. This can be a learning opportunity about elocution and skills that are always important to public speaking and singing, whether amplified or not. This training is the best way to radically improve the intelligibility of presenters at any time, but all the more so with masks on. I've been trying to remember, for the last 1.5 years, to walk up to a counter and place my order more or less in the character of Capt. Jean-Luc Picard. ("Hello. I would like a small house blend, please. Black. Engage!") Depending on the age of the students, there's only so much you'll be able to teach them, but you can try.
If leadership can give you a moment to do so, gather all the people who will be speaking. Train them to slow down their speaking (maybe 50%-75% of normal rate, especially if they're given to blurting a memorized line). Yes, train them to get appropriately close to the mic (a matter of inches for most dynamic mics) but also to project well, so you're not left maxing out the slider and getting feedback while someone whispers. Teach them to fill their lungs and use their diaphragm. Most of all, with face masks, encourage them to hyper-enunciate all consonants, since these are what is lost. Without face masks you might have to couple that with a "popper-stopper," but these days everyone is basically wearing one right on their face.
* (Or at least, that was true for decades; maybe now AI could make it possible. But if it's revealing "the face of the killer," maybe now we've got Minority Report-style concerns...)