Idea: a trigger system and a tripartite system. In transitive sentences there is one trigger for patient and agent, so: "hug-trig I-dir you-pat" OR "hug-trig I-ag you-dir" BUT intransitive sentences use only agent and patient case which are "presumed" to be the direct case and show volition. A single verb has different meanings depending on transitivity so transitively "hug-trig I-dir" is I hug BUT in "hug-trig dir-I" (I am hugged), the direct case is prefix. What are your thoughts? Possible?
Anything is possible. You just did it right there: It’s possible. Is it natural? Probably not. That only matters if you’re creating a naturalistic language. I personally don’t find it particularly interesting, because (a) trigger systems don’t exist, and (b) the conlanging community went through a trigger language phase way back when. That’s just me, though.
It really is important, though, in creating a language to distinguish between what’s natural and what’s possible, because let me tell you, one of those is a subset of the other.
Think of it this way. How many 1-20 character strings can you type on a qwerty keyboard? Some gigantic number. How many of those are licit English words? Some very large but still necessarily smaller number. So if you look at the following:
All of those fall under the category “1-20 character strings you can type on a qwerty keyboard”. Only one of them falls under the category of “English words”.
Your ask basically sounds like, “Can I type 83hf38 on a qwerty keyboard?” I mean, the answer is obviously yes. Were the ask, “Is 83hf38 a plausible English word?” the answer is obviously no. It feels like, though, you asked the first question, but meant the second—yet if the answer were no, you would’ve understood that to mean “not possible ever”.
So just to be really, really clear: In a constructed language, what is natural is different from what is possible. Everything that is natural is possible; NOT everything that’s possible is natural.
Furthermore, naturalistic conlangs are one type of conlang. They are not the best type of conlang, anymore than, say, naturalist painting is the best type of art. There is no best type of conlang. It’s just a matter of what the conlanger wants to do, and how well they do what they’re trying to do.
Also, this response here isn’t just in response to your ask. I get these types of questions a lot, where I feel like an answer of “yes” is bad and an answer of “no” is bad, and it’s because of how the question was asked, and because of the assumptions that underlie the question. I really do not want to give the impression that naturalistic conlangs are the best type of conlang. There are lots of other types of conlangs that can be made, and you can tackle a lot more interesting things if you’re not boxed in by naturalism. A naturalistic conlang is never going to push the boundaries of what humans could potentially do with language, simply because natural languages never do. They’re far, far too limited for the potential our brains possess.