I have grouped several suggestions together after reading the playtest packet and from experience with D&D (all editions)/Pathfinder and other games. Sorry if this should have been split into several topics.
1. New Skills
I have read that there is no option for learning new skills outside of Jack of All Trades speciality and perhaps new specialties. I come from a gaming group where the ability to buy new skills is always considered important and interesting and I think it needs to be there to reflect growth of characters in new areas (I like to buy new skills in relevant areas). Otherwise the Noble Fighter will never be able to buy the Spot skill even after years (and multiple levels) of keeping watch in the game etc. Most characters have just 3 to 4 skills from background/class skills out of 25+ skills while the rogue has 6 skills.
The simplest solution could simply be to modify description of Skill Training acquired at even levels to "increase an existing skill by 1 or (and here there are two possible solutions) purchase a new skill (at usual rating of 3) OR purchase a new skill at 1 (Rogue Skill Mastery would need to be modified to give bonus for skills of 3+ only). GM should approve of any new skills purchased."
While players are buying new skills the character is not improving existing skills so I don't think either option would be overpowering (with a maximum of 10 new skills by Level 20 on top of 3-4 starting skills and only if they never improved existing skills). I know there are ability improvements but it is not the same as actually having the skill itself
2. Choice in Class Abilities
I have read a lot about growth of characters being in options rather than more powerful options. This is presumably also true to Class abilities.
If that is the case then what about making class abilities a selection from a list to improve choice/customisation without adding complexity? E.g. Fighter choosing combat maneuvers from within their fighting style list or Rogues class features list from within their Scheme. If some abilities are more powerful than make a list for Levels 1 to 10 and then a list that can be selected from for Levels 11-20. I know this is similar to what Pathfinder does for some classes but it is a good idea that fits in with your concept for modularity and choices (obviously I am not advocating Pathfinder as a whole with this suggestion as I find that system too fiddly for my personal tastes and am very interested in D&D Next for many reasons) and makes it easier to expand on existing class abilities (in addition to offering new combat maneuvers/class features). This wouldn't be relevant for all classes obviously.
Some specific Rogue class comments
3. Skill Mastery - I think this would prove problematic in play (and I have read some playtester comments to suggest that has been the case). The Rogue already gains more skills than other classes AND +3 ability modier even if actual ability is lower AND Knack ability to get Advantage on rolls a certain number of times a day. It seems unnecessary to also give them an option to take 10 on their skills. Taking 10 actually makes Rogue tasks quite routine and safe which should not be the case - they should have the same chance of failure to keep system consistent across characters. Knack gives Rogues option to improve odds on keep tasks and stratetic decision making on when to use without guaranteeing a reasult and thus is more interesting.
One problem with 3rd Ed and Pathfinder is that Rogues tend to have such high skills leading to DCs for typical rogue tasks like picking locks/disarming trapshaving to be higher than typical for other skills to present a challenge. I can see the same problem arising here with 'take 10' (starting characters will always succeed with DC's up to 16 and as Skill Mastery improves they can clearly succeed automatically on DC 19 later on).
My suggestion is that Skill Mastery should drop the' take 10' component. If you do drop this then you could reflect higher ability modifiers at higher levels by saying that at say 8th or 10th level the ability modifer default increases from +3 to +4 (though I don't think that is actually needed).
4. Sneak Attack
Thug: Thug Tactics - This assumes a group of 3 characters or a friendly NPC. I am gaming with a group where we often only have 2 players and it is not always convenient/possible to have an NPC along for the ride. I am just suggesting that abilities should be described in a way that can be useful regardless of the number of players/NPCs present.
More generally I like Sneak Attack as an ability (it helps Rogues to be useful in combat) but find it can be hard to use in practice in smaller groups since it has often involved flanking an opponent in the past (harder with only one other player) or on first round of combat with surprise/initiative. I hope consideration is being given in D&D Next to other ways Rogues can gain Advantage throughout a fight to continue using Sneak Attack. There is a comment in the description about outwitting foe (and perhaps using tricks/fighting dirty as an alternative) which would be good to see expanded on as these presumably give Advantage, and could be one of the 'combined actions' that I read on the website?. Maybe a Rogue class ability?
The way you are going with resistances may also overcome anonther longstanding problem. Typically demons/undead/elementals/barbarians etc are immune to critical hits or flanking negating the poor Rogues sneak attack which can make them very poor cousins to the other classes abilities/spells and be very frustrating later on. Again hopefully something that is being considered.