I'll use that example to illustrate what I mean.
Combat could be described the same binary way: either you win, or the other guy wins. But most games don't do that: they using sliding scales for health, and track positioning, and give you choices of different moves and counter-moves.
So we could represent lockpicking the same way: model different kinds of locks, have various skills that make you better at manipulating pins vs. tumblers, tracking each one independently, etc, with a failure at any point meaning that you had to start over, or the lock is broken, or it's just beyond your ability, etc. We could build an entire sub-system for just lockpicking.
But that sub-system wouldn't apply to other skills (haggling, tracking, climbing, horseback riding, forging documents, cooking, etc.) for two reasons:
1. The mechanics that would evoke each of those activities, with the risks involved, would vary.
2. The fluff that gives meaning to the abstract mechanics would vary dramatically.
I guess I just believe that any multi-step conflict/task resolution system designed to apply to any scenario is not going to address any one scenario in a very meaningful, engaging way.