So, here's my approach:
A custom skill can be bought at the cost of any 2 other skills. This includes skills a character gets as part of his class.
The custom skill must be some sort of role (sailor, soldier, blacksmith, whatever) but it must also have a setting specific context (i.e. a sailor must have sailed with a particular navy or merchant fleet or something similar). This doesn't need to be written down in the skill name (though it can be, if it's brief), it's just something the GM and player need to be aware of, both to tie the custom skill to the setting and to help answer questions of context when they come up.
This custom skill can now be used as any skill within its specific context. That is to say, that when performing shipboard tasks, the 'Sailor' custom skill can be rolled in place of Endurance, Athletics or anything else. When it comes to knowing strange lore of the sea it may be used in place of Arcana. In short, it is a superskill within that specific context1.
Now, this is in part why it's important to keep the context very clear. Without boundaries, it is entirely possible to make custom skills overwhelming, especially if you treat it as just geography. It is not that a sailor can use his Sailor skill for EVERY activity on a ship, just the ones that sailor's do. Thus, he might be able to use sailor as a perception base to spot what's wrong with a ship, but not to spot ninjas sneaking aboard a ship.
Still, this is easily addressed with clear communication, and because this system mostly works within the context of existing skills, it's minimally disruptive while still expanding the scope of what can be done beyond the existing list.
That said, here are some optional rules:
If a custom skill seems too broad but not broken, it might be purchasable for 3 skill slots.
A less potent version of this approach swaps in custom skills for a single skill. In this system, you roll the new skill when no other skill is appropriate, but if you roll a real skill instead, you gain a +2 to the roll2. If you use this rule, you can replace the racial skill bonuses with a racial lore skill which implicitly covers those bonuses and which also allows for knowledge of what Tieflings enjoy for breakfast (and, implicitly, make those skills available to people outside the race)
The least potent version is simply "Works like a skill when no actual skill exists". In this case, I would not charge for it, but instead give each character one for free as part of the background system.
Hope that helps.
1. Structurally, the skill provides the capability to perform actions, knowledge about the topic, and a certain amount of awareness about relevant data.
2 - I forget the bonus type, but it doesn't self-stack.
Use the suggested moderate DCs from the most recent skill update
We have three classes of characters who grab: The Normal Character, The Weaponmaster Brawler, and the Red Scales Assassin.
The normal character grabs to satisfy rare tactical objectives of capturing an enemy without knocking them unconscious. It would be reasonable to assume that the rest of the party would pile on and that this would represent the conclusion of the combat if the grab can be held for some duration.
The Brawler grabs to lock a single enemy down. She will use this every fight and will serve as second defender. This, therefore, should be a good tactic for her to use. In the best of all possible worlds, a brawler should (by default) start her turn with an enemy grabbed roughly to-hit % of the time.
The Assassin can attack through a sustain standard on his grab. It would be nice if that had approximately equal odds to his attack hitting.
Powers of especial note are: Garrote Strangle (Executioner Assassin), Grappling Strike (Weaponmaster Fighter), Grab (Default. STR v. REF, no modifiers. Immobilizes), Net Snare (Foamgather Warrior Feat)
Rough target DC: to-hit chance.
Fighter:
- Heroic: Assume 18 base STR (+4) , +2 brawler style, +2 class, +10 base = 18, 23
- Paragon: no special +fort PP. Str 21 at 11, 22 at 16, +3, +4 items. = 28, 32
- Epic: +2 to str from ED, Str 26 at 22, 28 at 30, +5, +6 items = 37, 44
Assassin
- Heroic: Assume 14 base STR (as per guide, +2), +1 class, +10 base = 13, 18
- Paragon: guild exectuioner, Str 17 at 11, 18 at 16 = 22, 27
- Epic: +2 Str from ED, 22 at 22, 24 at 30 = 33, 39
For purposes of this analysis, we'll assume the average grabber follows the assassin fort progression and that no-one ever tries to escape via reflex (fighter has a feat that prevents it, assassin just laughs.)
Average of first 5 standard monsters at each level
Monster untrained athletics at 1,6,12,16,22,30:
- 1: +1, -1, +3, +3, +0 = +1.2
- 6: +6, +6, +8, +2, +7 = +5.8
- 12: +12, +11, +11, +12, +11 = +11.4
- 16: +15, +15, +15, +9 = +13.5
- 22: +20, +18, +13 = +17
- 30: +24 (MM3) = +24 (not enough data)
Brawler Fort at 1,6,12,16,22,30: 18, 23, 28, 32, 37, 44
Assassin Fort @ 1,6,12,16,22,30: 13, 18, 22, 27, 33, 39
Mob Athletics @ 1,6,12,16,22,30: +1, +6, 11, 14, 17, 24
%Odds of escaping fighter: .... 20, 20, 20, 15, 05, 05
%Odds of escaping assassin: ... 45, 45, 50, 40, 25, 30
Moderate DC @ . 1,6,12,16,22,30: 12, 15, 20, 22, 27, 32
%odds of escaping moderate dc: . 50, 55, 55, 60, 50, 60
While moderate DC isn't perfect, it does represent roughly even odds on both sides. Feats and items that improve defences for purposes of sustaining a grab modify the DC instead. It keeps the odds roughly consistent across levels, and means that a absurd fortitude pump isn't necessary to spec as a grabber. So long as the grabber carries a way to prevent teleportation, the modification to grab rules means that targets will have even odds of being grabbed and will generally waste their move actions trying to escape, maintaining the efficacy of the grab as action-denial.
Best Answer
Being trained gives you access to options for certain skills. Aside from these there is no difference in what you can do untrained compared to trained, assuming you can make the required DC skill check (clearly easier to do if you are trained).
Trained only abilities: