This game sounds better and better.
Usually, a tier system (few skill levels that provide clear benefit, vs classical skill that is a sliding scale with unclear payoffs) works best in a game with 1-3 PC, while in a game with dozen(s) of controllable characters, it creates other issues. How do you plan to handle multiple instances of the same skill on different levels?
The survival aspect already provides a good solution, time management: a lowly mechanic is still valuable when you have a top tier expert, because he can handle the plumbing while the expert builds some crazy device. In fact, his presence *allows* the expert to build his stuff because otherwise he would be bogged down in trivial stuff that is essential for survival.
Id like to hear more about how you plan to approach this aspect, especially if you havent given it much thought yet - it ties in perfectly with the human drama you want to focus on, and enhances a "simple" tier system really well. It probably solves the problem of in-depth builds better then adding different perks. Both at the same time would be preferable, of course, but making them balanced with themselves, and the game in general, is easier said then done even for 5-10 new perks.
Some theoretical stuff:
Few broad skills vs many one-dimension ones isnt just practical consideration, its complexity versus depth.
Take
Game of Life. Its turing-complete, meaning that it can represent everything a normal computer (and probably the human brain) can. You could run Crysis on that thing. Now start adding arbitrary rules to make it "more complex", because it is "way to simple and childish". Chances are high it will simply collapse after just a couple rules.
Complexity (amount of choices) is boring without depth (meaningfull choices). Complexity is easy to create, but "balance" / "depth" gets significantly harder past a certain point.
Of course, for real, practical games, costs of creation are the dominant factor. Just keep in mind that "more" is usually worse then "less" unless you pour significant playtesting ressources into it.
inhuman:
Gameplay core being the foundation of the story-world is one is my opinion as well. However, there is no free lunch, and making the story-world you want, then engineering the gameplay core that would create it is probably alot more realistic then creating an incredible system, then trying to hammer your world into it.