I actually designed a game using that very system you described about three or four years ago, when we did an RPG design contest on these boards. I could probably dig up a copy, though if you did a search for "RPG design contest" on these boards, you might find my entry.
The gist of the game was this (I'm going from memory, I might miss a few numbers):
* all skills started at a Base of 10. This cannot be improved in character creation.
* Characters could have minor skills (+10 bonus), major skills (+20) and expert skills (+30). These applied a bonus to the base skill, but did not actually modify it (this is important).
* Character attributes likewise modified skills scores, from -10% to +30% or something like that. Again, this didn't modify the base score.
* For difficult tasks, you had to roll over the difficulty, but under your score. So, if something had a difficulty of 20, and you had a 53 in the skill, you succeeded on a roll of 21 to 53. For easier tasks, they were always done in increments of 10 to your percentage score.
* You could take two actions per round, with no penalty on either roll (moving is an action). You could take any single action type (shoot a particular gun ,use a particular skill) twice per round max. But you could take additional actions with a penalty (20%, I believe?), or you could take only one action with a commesurate bonus (+20, I believe). So, if you moved and shot your gun twice, it was -20 on both shots. But if you just aimed, it was +20 on the single shot.
* Degrees of success: each 10 points you rolled on the die was one degree of success. So a roll of 03 had no degress of success, while 29 had two degrees. If it was a difficult task, you'd subtract the difficulty from the degrees scored (so if you succeeded on a roll of 52 against a difficulty of 25, you had 5 - 2 = 3 degrees of success). Degrees were traded for additional effects, as decreed by the GM. This wasn't really expanded on in my draft, due to space constraints, unfortunately.
* Opposed Tests: Whoever rolled higher and succeeds, wins. If one guy fails and the other wins, the success is obvious. If both people fail, whoever rolled higher wins. If a degree of success matters, subtract the higher degree of success from the lower if both succeed, or else just count the degrees of the guy who won. If both failed, there is no degree of success.
* The game had no critical success or failures. But 01-05 was an auto success, and 96-00 was an autofailure.
* My favourite part of the system: Whenever you rolled a "0" on a skill check (either the ones or tens column), you had a chance to improve your skill. You did this by roll d100 against your base score in the skill. If you rolled over the base score, you improved the skill by 1 point. This made skills level slower as they improved, but because everyone starts with the same base score in each skill and modifies it through flat modifiers, it meant that the super skilled guys levelled at the same rate as the unskilled guys - but it also allowed for weird quirks wherein the guy who seldom uses a gun found himself levelling it quickly, which I liked.
* I had feats in the game, of a sort - once your skills hit certain threshold levels, you'd unlock additional uses of the skill. In reality, this was a PITA, and if I made the game again, I'd drop it in favour of a traditional talent/feat system. But at the time of designing the game, I couldn't do that, because my main goal was to have character progression occur entirely "in game" and without any awarding of experience points.
Hope this plants an idea. The thing I liked about this system was that there was minimal math, and it encouraged players to use their skills, while still letting people focus on their specialties. It opened up a nice dynamic, too - is it smarter to let your non-focused characters tackle the easier challenges, to "level up" important skills, or should you let the specialists tackle the big tasks, using every chance to improve their most vital skills?
As a downside, because certain skills in the game were more important than others (I find perception, stealth, and knowledge skills get used a lot more in my game, and this game had combat abilities as skills as well), there was the risk of seeing rarely used but still important skills (such as repair or survival skills) not getting the same amount of screen time to improve. This was a design flaw I wasn't able to tackle in my final draft, although it isn't an insurmountable one (and would be easy to fix if the game was allowed to include experience points).