I was thinking it would simplify the skills system, making it easier to manage characters.
It wouldn't simplify the skills system, it would simplify character creation/advancement. That's a very important distinction to keep in mind. Simplifying play always has a direct benefit. Simplifying character creation is mainly good for pickup games, and tends to be outweighed by the downside of having less flexible and customizable characters.
No great rafts of skills to check.
There are two problems with this. First of all, you aren't actually reducing the number of skills by elimenating skill points. You are simply giving everyone a predictable easy to calculate (non-customizable) level of skill. The skill system remains (or not) a 'great raft'.
Secondly, you are I think treating skills mentally in this design as if they were solely passive abilities used only when the DM asks you to make a skill check and as a means of passing hurdles. I grant you that looking at the stock D20 skill system you can be forgiven for thinking thing, but you have to keep in mind that skills can also be active abilities used to solve problems in the way you use spells and class abilities.
[quote[So everyone would have a skill of Level/2. With the feat you get Level. With Skill Focus, you get Level +4, and with SF+ Epic Skill Focus you get Level+8.[/quote]
I understood your concept for the system. What I was saying is that you can't just give characters extra feats. If you give them just extra feats they'll be tempted to min-max and make narrow highly effective characters who - since D20 sadly tends to have weak passive skills and many DMs do low skill usage games - will then simply never pick a 'Trained' or 'Skill Focus' feat at all prefering instead to use the feat to enhance their active problem solving abilities directly (spells and class abilities). What you have to do is do like 4e did and give your characters X free 'Skill Feats' at 1st level which can only be used for picking up the skills you want to distinguish your character by.
I should say that from my experience this tends to often be a moot point. If your prior tables haven't featured balance and appraisal checks pretty much every session, and you've never had players pick up Skill Focus or similar feats (and certainly not after 1st level) then I suggest skills aren't particularly important to your game and if you really want to simplify you might be better off getting rid of them and just using ability checks for any skill situation. Consider what it is you really want and design to effect that goal. If skills aren't important to your table and seem like a drag on the game, just get rid of them. If your players think Balance, Appraisal, and Diplomacy are useless skills because they either never come up or have no effect on the game, and if pretty much everyone feels that skills become irrelevant once you start getting 'good' spells then skills are sufficiently a minor part of your game that you should either a) not worry about the system because you don't use it anyway or b) get rid of it because it's in the way.
Really? I thought it would simplify matters and prevent extremes.
Ok, so suppose I'm a Fighter. I have a +2 bonus on Fort saves. Then at 2nd level I take a level of Rogue. Do I now have a +2 bonus on both Fort and Reflex saves, or a +1 bonus in both, or a +0 bonus in both? If the answer is "+2 bonus in both", then you are heavily rewarding people for multiclassing gaining essentially a free valuable feat in addition to whatever other benefits you gain for dipping in another class. What happens if at 3rd level I now pick up a level of Ranger (which is actually a build I've used before)? Do I know have a +4 bonus on Fort and a +4 bonus on Reflex? Isn't that a huge advantage compared to not multiclassing?
4e gets away with the flat bonuses because it essentially kills multiclassing entirely. However, that is IMO a much less than satisfactory solution. The simple solution of course is to say that it doesn't matter how you multiclass you always keep your bonuses from your first class and they don't change, but if I want to build a character that is equal parts fighter and thief it will stand to reason that I want a character whose abilities are pretty much the average of both. The great thing about 3rd edition is that it pretty much let you do that, and under your system to get the same effect you'll need rules that say in effect 'If you are equal parts fighter and rogue you get +1 bonus to fort and +1 bonus to reflex'. But trying to write that rule out for the general case is not at all easy.
Are you going the 'everything stacks' route, or the 'take only the best bonus' route?
Either one would be fine. The important thing is that both are more simple that having named bonuses. Named bonuses produce a huge bookkeeping burden because you have to remember not only the bonus, but its name and check against every other bonus with the same name to make sure you don't accidently stack them. This is fairly easy for a computer to do but very hard for a person, especially when the bonuses are changing from round to round.
That isn't to say that I would necessarily go either way - I outlined my thoughts on named bonuses - but I'm trying to explain why giving everything a named bonus isn't in and of itself a good idea. It's a good idea to cut down drastically on the number of different types of named bonuses, and therefore 'giving everything a named bonus' depending on the implementation could be a very bad idea.
Wis is already the dump stat in 3e. You just take the Force of Personality feat to use Cha instead anyway.
The example was arbitarily chosen. The point is that you'd move more to the 4e model of having at most 3 stats that actually count, and 3 stats you dump as useless. If you are going to go to '3 stats actually count' just go to three stats - Body, Mind, and Heart (or something like that) - and be done with it.