Forget unused skills -- the highest number of skills on a standard level is 320 but a user-defined level could have up to 792, so even 10 points per skill would unfairly swamp out the lemmings saved and time bonuses, while any fewer would make the scores from best play on a level with fewer skills available not worth having.
Why are you assuming that the score must be such a simple function? What if you got a bonus for the *percentage* of skills that you have leftover. Then you have no such problems for high numbers of skills, since you can set a hard cap. Make it very low, sure.
The point is, if two people save the same number of lemmings, and have the same time remaining, the person who used fewer skills should get a better score!
--------------------
The problem with any such scoring system is that it only makes sense in the context of a single level. In my mind, the best theoretical scoring system would be where each level had an equal "perfect" score (most saved, minimum time, minimum number of skills used), and you *deduct* points from that score for every imperfection.
Of course, this has several implementation problems. For one, who determines the most possible saved, the minimum time possible, and the minimum number of skills used? Surely you might think you have the best solution without actually having it. Two options:
1) A dynamic retro-active scoring system could solve this issue. If you save more than the max thought to be possible, then the max is set to the new max. Etc. for the other possibilities. [This would work best with an online hall of fame.]
2) Just add a small bonus formula, so you can get higher than the "perfect" score. (I would not really recommend using the negative version of the deduction formula, because for instance in a lose 2 level where lose 2 is thought to be the max possible, saving 100% would result in an explosion (or divide by zero, or something of that nature) of that part of the score.) [This would work best with an isolated single player game.]
Another problem, precedence for scoring would have to be set up properly. What is "more perfect", having more time left over, or using fewer skills? Is saving one second better than saving 10 builders? And even within skills, is saving 2 climbers better than saving 1 builder? For the most clarity, the precedence should be clearly defined, so that for instance, saving 1 second IS better than saving 10 builders, OR that saving 1 builder IS better than saving 2 minutes. And probably each skill should be treated equally for simplicity.
So probably deduct the most points for a lemming lost, less points for a skill used (can certainly use a percentage based system here), and single points for each extra second used (or use a percentage based system, but then you are liable to run into ties when scores differ only by 1 second in long levels).
The system probably sounds a bit complicated, but it has the advantage that the maximum score the player is striving for is clearly defined, and that all levels would carry the same weight. This value could be multiplied by the difficulty rating, so that Fun levels are worth at most 5000, Tricky 10000, Taxing 15000, Mayhem 20000... Or whatever (and if you use the bonus system for exceeding expectations, of course you can go over those maximums slightly). Then when you see that you got 18240 score on a Mayhem level, you know that you can do better... and if you get 14997 on a Taxing level, you probably used the intended solution, just slightly slower. (And very high scores above the "perfect" score would instantly flag a backroute!)