So, I was thinking of a way to reduce skill list length, and it hit me. Lets use d20 for example. Instead of all the skills there are, you'd only have Strength, Dexterity, Constitution, Intelligence, Wisdom and Charisma skills. You'd rank them up as normal and whenever you attempted to use a skill based off one of those stats, it'd be at a -10. You'd also get a feat-like skill every level. You'd choose something like "Find Traps - beginner." Now, you only have a -5 when trying to use Dexterity to find traps. You could then get intermediate and master levels, which reduce the penalty to 0. Alternatively, the feats would give you a bonus, and the TN would be higher. It'd be a very short skill list, but still allow for character differentiation and some crunchy bits.
I'd play test it, but my playtesters are overworked as it is. :P What do you think?

Hi!
  That doesn't really reduce the number of skills, just migrates them from skills to feats, no?
  I feel like the key to reducing the number of skills is to create some catch-all skills for stuff that is not central  to your setting. It just takes a couple of broadly named skills to cover you, like: Lore, Athletics, Artisan and so on. Communicate the intent of these broad-based skills as well. This will help prevent Skill creep (where the number of skills increases over time) in the future...
  I think that if you decide to include a skill system in your game, there are a coupe of pitfalls you need to avoid:
1) Not including necessary skills (easy mistake to make)
2) Including unnecessary skills (Takes focus away from what the game is about)
3) Including too few skills (This has two disadvantages, one its very easy to be very good at many things and may prevent players from making the character they want)
4) Including too many skills (this makes it harder to make a well rounded character)
5) Confusing skill overlap (I think you should design the skills to overlap consistently, meaning how many skills you can apply to a specific task)
  I hope this helps.

Are you specifically looking to reduce the number of Skills in your D20 games? I'm just trying to discern your intent... If so, then try Castles and Crusades from Troll Lord Games. Its d20 - same Stats etc, with no 'skills'. Instead, each character class has certain things they are good at doing, if they don't have that thing listed when trying an action (like Stealth) then either they are a at a penalty (or perhaps its a higher TN - I'm a player, not the GM!). You get to chose two stats that are Primary (3 if you're Human) and you get +6 to your d20 roll for actions goverend by that stat.

Personally I like a skill list with fairly broad skills in it, including combat ones like edged/blunt/ranged/thrown/2-h. Like most systems, skills are good at representing certain activities, e.g. investigation where you can learn the steps to pick up clues, where to look, track suspects, etc. But other things that often come under the skills tag don't seem to fit this model, such as bluff.

True 20 has some collapsing of # of d20 skills (e.g. Sneak and Move Silent become one skill, Stealth or something like that).

What about skill lists being derived from attributes? For example, a character with a Dexterity of five could choose up to ten specific simple dexterity based tasks he could perform, five elaborate  dexterity based tasks, or three incredibly elaborate dexterity based tasks. Who needs a skill list in this case? It would allow players to creatively express the capabilities of their character and the GM would have an easier time gauging the capabilities of the PCs. Just scrawl down a list of sample tasks and give it to the PCs. The Dexterity five character could: catch a baseball (simple, 1 choice), Throw a weapon (elaborate, 2 choices), dodge a punch (elaborate, 2 choices), dodge a bullet (incredibly elaborate, 3 choices) , fire a handgun skillfully (elaborate, 2 choices). There might be arguments about which tasks can be accomplished by which attributes, but with GM discretion and mature players, skill selection need no longer be an issue. Saves enough paper and would allow more space in corebooks for import things.