I know how that feels -- I've dropped a few projects before over graphics, so I'm going to keep the graphics on a tight leash this time around.
As far as models and skins and so on, yeah, that's the general idea. However, there are at least two things that are going to keep the problems in check:
First of all, Armok hasn't quite gotten to the point where, say, lots of items are being added in at a feverish pace. That comes a little later -- right now, we're laying down the conceptual framework. This applies to the models. At first, I'll make a few crappy models (one for humanoid, quadraped, etc.), and different creatures will probably be differentiated by size and color (neither of these require making a new model). Then, just as I'd add in three different basic item types and ten pieces of furniture further down the road, the models would also become more complicated and numerous. Once the general Armok framework is in, there will be programming sessions devoted simply to bloating the number of items, spells, creatures, and perhaps the graphics bells and whistles as well. Right now I'm just putting in enough to
a) see the basic 3D maps without having to scratch my head and puzzle out where things are
b) get a real interface up.
Secondly, each random creature doesn't need to have a separate model, at least not with its own image file for a skin and vertex lists and so on. Just as the parts are stuck together and altered in the game to make the creature in the first place, vertex lists can be spliced and pieces of a few preset skins can be merged to form the creatures overall skin. Each part or whole model can be scaled to any size, and the stick-figure animation framework can be warped to change the gait of the creature, etc. This can be done by the game on the spot whenever a creature is loaded. I don't have to sit in front of a 3D Modeling Program pulling out my rapidly thinning hair to do it.