X-bar theory is prescriptivist in a certain sense. It prescribes certain things about the structure of syntax trees: that all branching is binary, for example, and that every XP level dominates an X' level dominates an X level.
However, these are all properties of the syntax trees, theoretical constructs postulated to explain the data. This generally isn't what linguists mean by "prescriptivism", which is about declaring utterances correct or incorrect against native speakers' intuition.
The goal of X-bar theory (and any other theory of syntax) is to explain the data: which utterances native speakers consider grammatical versus ungrammatical. A theory that claimed sentences can't start with conjunctions would simply be incorrect (for most dialects), because native English-speakers do that all the time. It might claim that sentences can't start with conjunctions in the underlying form, and then some sort of movement, deletion, or other transformation happens. But the data indicates pretty clearly that surface structures can start with a conjunction, and these theories are descriptivist in the sense that they're meant to explain the data rather than prescribing it.
