So I have a skill system hack I am working on and I want to share it with you.
Start with Lamentations of the Flame Princess, the rules set I am using for my Megadungeon Madness game. It's basically B/X but with some tweaks (ascending values of armor class, a silver piece standard, and no codified bestiary, for starters). One of the nice things about it is that skills are based on d6 rolls. Virtually everything is a 1-in-6 chance, and specialists (read: thieves, but a less pigeonholed concept) get to invest points into expanding these skills. E.g., put two points in Sleight of Hand and your chance to execute such a task goes from 1-in-6 to 3-in-6.
I am hacking that d6 system a bit. Mostly because I love dice pools, but there is also a logic behind it.
First off, rather than expanding the range of success, I expand the number of dice you roll, while keeping 6 the target number. Note that statistically this is more difficult. For example, it is harder to roll three dice and get a 6 than to get a 4, 5, or 6 on a single die (the 3-in-6 case):
(A) Dice pool: 3-dice probablility = 1-(5/6)^3 = 42.13%
(B) One-die probability = (3/6) = 50%
A solution to the increased difficulty is that the GM should be more generous with the bonus dice. We already add dice based on invested Specialist skill points, but let's also add Ability Score bonuses. A +1 to Strength is easily added to your pool if you want to Open Doors, for example. Have a crowbar? Add another die. And so on. This borrows heavily (if it isn't identical) from the "negotiated skill system" that I heard used on the Dwimmermars game of +Adam Muszkiewicz (of such notable endeavors as Drink Spin Run, Metal Gods of Ur-Hadad, and whose name I can apparently spell without having to Google it now).
Another nice feature is that the bonuses can easily go to six or higher. Whereas a 6-in-6 chance of the basic one-die system is a guaranteed success, a pool of 6 dice is only going to yield a 6 at a rate of 66.51%. So don't be afraid to be liberal with your bonuses.
The problem that we have already come across thus far in my Megadungon game is that opening doors is still fairly likely to result in failure. Even with crowbars, people helping, and strength bonuses, you are pretty likely to not open the door.
This is where I got the idea to go straight into story game territory and offer the players a narrative choice. In the first option, the players could choose to let the door be. The door is swollen shut, just like St. Gary said it probably would be. The other option is to note your margin of failure. (E.g., was your highest result a 4? Then your margin of failure is 6-4 = 2.) I'll let your character(s) persist at opening the door until they succeed, but I get to roll the margin of failure in Wandering Monster checks. (A dice pool of 2 in this example.) Again, each of these is a simple 1-in-6 chance. The logical basis is that the worse your initial check result, the more noisy your success is going to be and the more likely you are to attract attention to yourself.
I think this is a pretty elegant mechanic for doors if you're going to go Full Dungeon Crawl. I'd like to think that my game ups the tension by emphasizing slow, careful mapping, and keeping track of resources, etc. Wandering monsters (or more accurately, random encounters) are another high-stakes element, and it's nice to put the devil's bargain in the players' court. The alternative would be to have the players reroll the check while I check for the random encounter with each failure. This has the potential to be tediously drawn out, especially if the player has a small dice pool.
I've already used this mechanical tweak in one full game session, and the resulting random encounter was deadly. But more importantly, the decision of whether to push forward at the door had a lot of gravitas behind it. In all, it's an idea I'm quite proud of, given that I'm only starting to break out of the habit of sticking to rules as written. I'd probably have broken out of that habit ages ago, but it really requires sitting down to run some real games.