We've already established, however, that skill systems are in D&D from the beginning, even before thieves show up in (I think, from what I've read) Greyhawk. So we're going to have to modify this objection some. I think that this objection has more to do with thieves necessitating, or at least encouraging, skill systems that do what I described in this post, that is, that the existence of thief classes necessitates the existence of a skill system that encourages players to solve problems with abstract dice throwing and not concrete creativity.
The question, then, becomes how we can create a skill system that encourages concrete creativity and not abstract dice throwing. I think there are at least two ways to do this.
One is to go the route of Rob Conley in the Majestic Wilderlands: abstract, general skills. He creates a skill system with a bit more granularity than mere attribute checks; that is, each attribute is divided up into a few broad skills, each of which may receive bonuses based on class or some other factor. For example, Dexterity is divided into Climbing, Legerdemain (manual dexterity- pickpocketing, sleight-of-hand, disarming traps), and Stealth. (I assume that any Dexterity-based actions that don't fall under these broad categories would just get straight attribute checks.) These abstract skills encourage concrete descriptions in the same way that just using skill checks does: the skills are so abstract that the player isn't exactly sure how the ref will adjudicate what he wants to do, so he describes his actions concretely and lets the ref figure out what dice to roll against what number. This is all well and good, but it's not to my personal taste right now; I'm enjoying how well attribute checks work for me in this way at the moment and don't really see the need to split them up like this in my own game.
The other way of designing a skill system that encourages concrete description is to use skills that are for concrete, exact tasks, but to ensure that these tasks are tasks that can't (or won't) be described easily at the table.
For example, a skill system like this might include a "Lockpick" skill because (unless you want to add a mini-game or have a lock at the table or are familiar enough with picking locks that you can easily have a conversation about picking an imaginary lock) that's not something that can easily be modeled at the table: it's not the focus of the game and, like combat, it's much easier to abstract the details away and move on to further exploration.
By contrast, this kind of skill system would probably not include a "Find Traps" or "Disable Traps" skill. These are things that are easily described in concrete ways by your average player: "Grom prods the floor all around with his 10 foot pole," "Bill shines his bullseye lantern around the ceiling, looking for anything unusual," "Robert looks the chest over carefully and scrapes off the grime you mentioned from around the lock to check for any holes for poison needles," "Grom cuts the tripwire and carefully ties it to a stake he's hammered into the ground but out of the way," "Bill sets the end of his rope on fire and catches it on the rope of the pendulum trap to catch that rope on fire and destroy the trap," "Robert takes some large wooden splinters and stuffs them into the poison needle hole he's found," etc., etc.
This is the way I want to go with the Scoundrel: Each Scoundrel will have a collection of abilities and skills that the player has chosen from many options. This will allow for both the ability to shape Scoundrels to each player or ref's concept (a ref can disallow certain abilities or skills to fit his vision for the campaign or for the Scoundrel), and the fact that the archetype I'm going for here is one that picks up, opportunistically, whatever skills and abilities he can as he goes along; by the very nature of the Scoundrel, there is no defined path to stay on by discipline and will power because Scoundrels are opportunistic and almost always undisciplined. Abilities will, like with all classes, be exclusive to the Scoundrel, while skills will, by definition, be available to all classes; in fact, I would argue that each character, regardless of class, should start with a base chance of success in each skill. Just as fighters have better chances of hitting opponents in fighting but everyone has some chance, though, Scoundrels will be allowed to progress is skills more quickly than other classes if they so choose.
The main work, then, of creating the Scoundrel will be figuring out the mechanics of the skill system and figuring out what abilities and skills should be available. I've got some help on both of these tasks from around the OSR blogosphere that will prove invaluable as I go about doing this. Next time, we'll see if we can hammer out a good skill system mechanic.