In an earlier incarnation of the project, I was looking at the idea of maintaining the existing character sheets and stats inherent in Mage and trying to work a simpler system through them. This could be feasible, but I haven't worked out a clean way to do it.
I'm thinking it might be easier to strip things back to an approximation of the Minds Eye Theater rules. Instead of nine attributes in three categories (Physical = Strength, Dexterity, Stamina, Social = Charisma, Manipulation, Appearance, Mental = Perception, Intelligence, Wits), we just use the three categories (Physical, Social, Mental). Instead of adding attribute+ability for a dice pool to determine degree of success, you simply test with the attribute and the possession of an appropriate ability allows a retest.
The attribute categories start at 7 (for the best attributes), 5 (for average attributes) and 3 (for the characters worst attribute category). An attribute category total of 10 is the highest a non-supernatural human can go. Willpower still starts as 5, and functions as a generic pool of retests that can be applied to anything, Arete starts at 1.
For simplicity, I'm thinking of just adding a randomly drawn card's rank to the attribute, then comparing it to a static difficulty added to a random card.
Botch if Difficulty + Card is more than five points above the Character's Attribute + Card
Fail if Difficulty + Card is below Character's Attribute + Card
Success if Character's Attribute + Card is equal to or greater than Difficulty + Card
Bonus Success if Character's Attribute + Card is five or more points greater than Difficulty + Card
Since we're using a tarot deck for this, and there are four suits in the Mage Tarot deck that match the four essences of Mages, a character might earn an additional degree of success if they are attempting an action in line with their essence and draw a card of a suit matching the essence. This means the character won't also get the additional degree of success, but it does give them a motivation to act according to their essence for the potential bonus it might deliver.
I'm similarly thinking of using the tarot deck to drive a scene generation system akin to the Hold 'Em Scene Generator I came up with a few years ago. This scene generator will create a type of conflict inherent in the scene, a twist applied to it, and a general difficulty in order to succeed in the scene. But that's where we come back to Tony's idea. One of the options that the generator produces is an NPC antagonist, and through the course of the story, we can add information to the NPCs produced in this manner. This might be done through successful (or maybe unsuccessful) actions. Here, I'm considering the idea that a success might allow a player to reveal through their character a bonus that is applied to an ally, or a weakness applied to an antagonist...conversely, a failure might allow one of the GM (or one of the other players) to define a weakness to an ally, or an advantage to an antagonist.
Every time a new advantage or weakness is defined, any attempts to further defined them are at a higher difficulty.
There's a lot more thought to go into this.