Even if I'm planning to divide abilities into levels of expertise (none, basic, intermediate, expert), the easiest way to implement these would be to simply open new options to characters who possess these ability levels.
This has precedent in a lot of systems I've encountered. Here are some specific ideas applicable to this game (also with precedence in a lot of Boffer Systems)...
None: You may common use weapons up to 30cm/1ft in length.
Basic: You may use common weapons up to 60cm/2ft in length.
Intermediate: You may use common weapons up to 90cm/3ft in length.
Expert: You may use any common weapons.
This may seem a bit strange, but among reasonably equally skilled opposing players (such as most LARPers), weapon length is a great way to differentiate character skill.
None: You may not use shields.
Basic: You may use a shield up to 900sq cm (30x30cm/1sq ft).
Intermediate: You may use a shield up to 3600sq cm (60x60cm/2ft x 2ft).
Expert: You may use any shield.
None: You don't know what to do when a ritual is conducted.
Basic: You may act as a participant/follower in common rituals.
Intermediate: You may act as a participant/follower in any rituals, or may learn and lead common rituals.
Expert: You may learn and lead any ritual.
(The exact wording may need to be changed to avoid ambiguity or abuse from rule lawyers).
I'm not so sure how these skills really balance against one another, but it's got the right kind of feel and roughly matches a few of the other systems I've seen. It doesn't require to much thought during the course of play, it just opens options.
Beyond these abilities that simply open up options for characters, we can provide others that modify the numbers used in the game.
None: You suffer injuries like everybody else.
Basic: When making a test to see if you suffer permanent injury or death, reduce your effective number of injuries by 2.
Intermediate: You gain an extra Hit Point.
Expert: Any healing effect restores an extra Hit Point when it is applied to you.
None: You deal injuries like everybody else.
Basic: On the first hit on an opponent during each conflict, you may declare "Strike!". This hit now does an extra point of damage.
Intermediate: The first time an opponent's armour is declared against one of your strikes, you may ignore it.
Expert: You may declare "Strike!" to deal extra damage as many times as you want during a conflict, but if you do so more than once on a specific opponent, you risk damaging your weapon. (Check with the GM at end of conflict).
None: You live according to what's in your pockets.
Basic: You have a modest small business or farm that generates 1 Gold per month.
Intermediate: You have an interest in a few modest businesses/farms (or a very successful one), generating an income of 2 Gold per month.
Expert: You have a finger in many pies, your diverse investments generate 1 Gold per week.
(I'm not sure of how the monetary economy will function in the game, so this ability is certainly subject to change).
These get even harder to balance against one another. I'm not going to include things like "Athletics" or abilities that really do rely on the physical capacities of the player (unless someone can come up with a good way to simulate this).
I'm a bit torn when it comes to abilities with random chances of success or failure. Things like medicine (where injuries might be worse than expected, or where medicines just don't take), repairs (the same general ideas apply), research (where the information might be present, but you just glance over some sections while concentrating on others), etc. I could simply reduce it to a 50/50 chance because the game is focused on other areas of narrative and simulation, but that just feels wrong.
I don't really want to bring too many numbers into the game, because that also works to break apart the narrative elements and the game mechanisms, leading down the path to broken immersion (which then leads to players breaking from the flow of the game, and other issues). So, I'm ruling out attribute numbers.
The other option that might work uses a trait system. Perhaps this would be akin to FUBAR (where players would draw a minimum of two cards, choosing the best or worst two among them based on the traits appropriate to the situation then allocating the results to success and failure categories) or Tooth and Claw (where players would draw a number of cards based on how much risk they take and other traits appropriate to the situation, then keep a certain number wher every card applies a success or failure condition).