The consistent histories approach to quantum mechanics doesn't require the word "we" or any similar philosophical word to be defined. Instead, it defines what a history is (a sequence of projection operators at different times) and what it means for pairs of histories to be consistent (an orthogonality condition of a sort). That's enough. The dynamical maths of quantum mechanics can then be used to calculate the probabilities of different histories.
People's formulations only used the word "we" associated with a set of consistent histories for a particular set of consistent histories that (incidentally) "we" or one of us could find relevant or helpful to do planning of anything etc. However, the rules of the consistent histories approach do not require any consistent histories to be "relevant" or "important" (for anyone or anything, whether it's "we" or "them" or anything else), so there's no need to define "us".
The consistent histories approach, as well as any meaningful Copenhagen-like interpretation of quantum mechanics, rules out the possibility that we're a "classical simulation" of the quantum system (and I guess you meant a "classical" simulation). Our world is genuinely and fundamentally quantum and this fact is indeed very important for the consistent histories formalism to make sense. A classical simulation always envisions just one set of observables, one sets of questions that fundamentally make sense. But it's a genuine and true feature of quantum mechanics that is made plain obvious in the consistent histories approach that there can be many ways to choose the set of consistent histories and none of them is "objectively better" than others.
The closest question to yours that could be a genuine concern is how the consistent histories approach guarantees the consistency between the conclusions of different observers – different sets of consistent histories. But it does guarantee that. As long as two sets of consistent histories contain some questions that may be answered in both sets, the mathematical formalism guarantees that the predicted probabilities will match independently of which set of consistent histories we choose. That's ultimately guaranteed by the "consistency condition", after all.