Benjamin Gauronskas put up a post about "meta" challenges in games, kind of thinking them through for his own games.
I've written about these before, but I wanted to address some of the things he brought up and my own experience.
I prefer rolling in general
That is, when possible, I prefer that the character's skill matters. Even if the character's skill is just determining how well you spoke or how well you solved the problem, I want the character's skill to matter. I like that about game systems - you can have playing pieces of different ability, so even a skilled player has to leverage what they have and don't have on the sheet itself.
I'll skip out if player actions are so on-target that failure isn't possible, or if success isn't possible. For example, if the players offer something in negotiation that's so perfect it can't go wrong, the NPCs will just accept it. Or if the players make some terrible error, like, "I tell them I'm allied with their friends, the orcs!" when the orcs are their mortal foes, well, even a 3 on Diplomacy isn't changing their minds. (Although, if the players try to backpedal and then roll that 3, well, sure.)
But I do like player challenges
After all, there are meta-elements to games. Decisions about where to step on the battle map. Who to attack. What spell to cast. Left or right or straight. Whether to search that given room or not. They automatically determine some success or failure.
Puzzles, riddles, obstacles - these are often solvable just by the players showing some skill and ability. There really isn't anything to roll. There potentially could be - if you have to pick between the Red Handle, the Yellow Handle, and the Blue Handle and you can't figure out the riddle, the Intuition advantage might point you to one of them. But just like I don't let you just roll IQ to see if left or right is better, I don't like to allow knowledge skills to bypass riddles and puzzles.
For example, I'll make you roll Search to see if you find stuff, but if you say, "I lift the book and look under it" I'll tell you what's under it without a roll. Occultism might give a clue to a mystical puzzle. And so on. But I can and will put things in your way you need to decide how to deal with, or wholly deal with, through your own thinking. That's part of the game I run. Character abilities can't wholly replace player skills, just like player skills can't wholly replace character abilities.
Don't tart it up
Probably not the best wording for that, but there you go. Tell me what you are doing, how you are doing it, and then let's get it done. I don't need prose explanations of your character's actions. Unless, of course, they add value to it. "I step forward and make a mighty swing of my broadsword at my hated foe, so that we may strike down the evil before us!" = Step and Attack, Broadsword swing, roll. So just say that last bit. That's all I need. But "I yank my sword out of that jerk and say, 'That's for crossing me in Swampsedge'!" or "I quietly shiv the guy next to me while I keep my eyes forward." = yes, that makes it better and helps us visualize the situation in a useful way.
I'm all for roleplaying, but don't complicate the elements of the game not subject to roleplaying. If the dice are telling us what happens, then all I need from you is telling me what the dice don't. Where you stand, what you say, how you react. Not how mighty your swing is or why you so desperately need to open this lock or find those tracks. I'm not handing out a bonus because that's the kind of extra I don't want to have.
Avoid Dead Ends
As a GM, generally, you want to avoid total adventure dead-ends with any challenge if it's feasible. You don't want a door lock that must be opened and the game stops if it can't. You don't want a puzzle that dead-ends an adventure unless you solve it. This is just because sessions are less fun when you spend 90 minutes with everyone trying to roll better or figure out the answer that allows them to actually get on with things.
I know there is a philosophical argument against this, but my experience says, try not to put in a game-stopper with a narrow solution. It's better if a puzzle or an especially difficult challenge is stopping a bonus, not stopping all action.
That's just the series of thoughts occasioned by Benjamin's post.