If you are not stealing ideas in your game design, I might suggest you are doing it wrong. There are a lot of games out there that have hit upon really good ideas, even if the games themselves have met with a variety of fates.
Margin of Success Tricks
There was a game once called Secrets of Zir'an. It was a clever, interesting game, but it suffered the kind of printing problems that make it a cautionary tale rather than any kind of real success. It's a shame, since it had some clever ideas, and one in particular really caught my eye.
SoZ had a system where you could learn various tricks to use with a skill, things like combat tricks for extra damage, extra spell effects, and so on. There were a lot of these, and they allowed for a lot of differentiation between skills. The trick was that each one had a particular cost, and that cost was paid out of the margin of success of an attack.
To illustrate, imagine a fairly D&D type-attack, where I roll d20, add a number, and hit you to do, say, 2d6 points of damage. I only needed to roll a 19 to hit you, and I rolled 24: that's a margin of Success of 5. I have a few tricks: for 1 point, I can increase damage by 1. For 3 points, I can knock you down. For 5 points, I can perform a disarm. So using those 5 points, I could do +5 damage, +2 damage and knock my enemy down, or do a disarm.
There's a downside to this in that it's a little math-ey, but it's got a big upside (beyond just cool fiddliness), since it clearly calls for description after resolution, so you get to incorporate your awesomeness.
Wilderness of Mirrors is an innovative game in a a number of ways, but one of my favorite elements to it is often overshadowed by the other novel elements, and that is role pricing. WoM doesn't have skills, rather there are 5 roles you buy not representing the roles on the spy team. The danger with such systems historically is that the benefit of having a broad skill at all is usually much greater than the benefit of incrementally improving it, so there's a strong incentive to buy many things at low levels rather than specialize. This is especially true if prices get progressively higher.
WoM turns that on its head by making the first rank cost the most, and each subsequent rank cost lest (so the price is 4:3:2:1, not 1:2:3:4), and I dig the behavior that encourages. It leans naturally towards niche protection and excellence within your niche.
"Wild Card" Skill
Eden's "Buffy" system was a lot of fun, and it's a shame it's vanished down the licensing hole, but I want to see if we can hang onto one particular idea.
Buffy had what I would describe as a medium sized skill list. Long enough that I couldn't rattle it off by memory, but shorter than 15-20. It's a good size, since it allows for skills that are neither too broad nor too narrow, but it's a bit of a crapshoot. See, at that size list, you have the greatest chance of having overlooked something important to one of your players. Buffy's skill list included one blank slot where a player could insert any skill he needed, so if it was really important that his character is an excellent accountant, he could put that in the Wild Card Slot.
There's an obvious benefit to this - it introduces a bit of flexibility into the skill list without demanding that players come up with the skills on their own. But there's also a subtle benefit: whatever skill the player put in that slot is a flag - he is more or less calling out the GM the thing he thinks makes his character stands out, and is inviting plot hooks to be attached to it.
There are, of course, lots and lots of other ideas worth stealing, but you can't steal them all at once. I figure three at a time is probably digestible.
1 - Something similar was done in another brilliant-but-gone game called Fireborn. The difference between them us subtle, but potent. In Fireborn, you effectively paid for actions in the difficulty. So if I want to do extra damage, I increase the difficulty to hit by some amount. There's a compelling logic to this, but it has the downside of making cool behavior (stunting and the like) less likely to succeed - players are punished for doing something risky. It's more 'realistic' (harder tasks are harder), but not necessarily more fun.
2 - With fewer skills, odds are good it falls under the penumbra of one of them. If there are lots of skills, then adding another is pretty trivial.
3 - Going from memory here though, so apologies if I'm misstating.