For instance, if you rip off an idea from D&D and imagine an order of paladins grown soft and imagining that the legends of their former powers -- serial numbers neatly filed off, but holy powers -- were just legends, you can have a young, newly knighted paladin reveal that, actualy, no, they have just lost them from their lack of virtue. (And, as he puts it, it could not be a high standard, since he lived up to it.) But to do that, you need a situation where they are put into play.
An attack no doubt. They lost them during a time of peace. But then you need something to attack that needs holy powers. And the attack has to occur where he would be seen, preferably by lots of people, unambiguously.
You notice that gives a lot of leeway. Too much to easily narrow down the choices. Even adding that I wanted the end of the story to be when the young paladin saves the day from the monsters and cuts free from the order -- the politics of becoming accepted again is not what I want to write -- still only adds the requirement of their being containable.
So I take one stab, and then another, and then another. I thought I had introduced it nicely to the idea of how the problem got out in the first place, but the outline is lying on the page, incomplete, and trailing off. Perhaps I should lop off the idea of how the problem got out, and keep the notion of the living shadows that were the second idea of a monster. Perhaps I should brainstorm up a new monster, but then I need a way to contain them all.
One character has emerged from the darkness, talking nicely to my young hero, and inspiring the hero to retort that he's trying to taint him and lose him his powers. I don't even know whether he's going in, though he may make a nice target.