Our standard DM (not me–I find DMing really interesting but I’ve never been quite as good at improvising as I need to be, I mostly get caught up in the cartography and the social stuff and forget about the little things that actually make an adventure) does a lot of this, not so much in terms of actually creating the skills but simply improvising the rolls, often basing things on int or str checks with some ad hoc modifier.

I want to pick up on a point you make in your last paragraph:

Sometimes I like to throw in a check, opposed if applicable, to see if the player can determine the amount of truth, exaggeration or manipulation they are getting from an NPC…

I will usually throw in some hint of this (or even an outright statement) if the NPC fails his attempt. It depends on how badly he fails, really, and how blatant he’s being about it. And I assume that if a PC fails a bluff/intimidate/whatever check, that the attempted target might also cotton on, and react accordingly.

In many situations, I like to see players (when I have them) determine on their own what skill checks they intend to make, and if they pick the wrong ones, or only some of the right ones, penalties apply, particularly relating to knowledge and sensory skills. For instance, if I decide that a given bit of info might be garnered from a guard with a lie/bluff, then a good intimidation or bribe might get half the information or slight misinformation if I’m thinking quick enough.

Sometimes I like to throw in a check, opposed if applicable, to see if the player can determine the amount of truth, exaggeration or manipulation they are getting from an NPC, or the amount of confidence they might have in their own skill before attempting a certain task. And that check can be modified by recent successes or failures or pertinent information the character might already have. For instance, in the case of pachristian’s waterfall, I’d probably tell the player that it looks “not so tough” or something similar, since he did a bit of climbing and jumping earlier at lower DCs (better light, dryer surface, fewer injuries, etc.).

Your suggestion is excellent, and exactly the sort of thing that the blog was intended to suggest – provided that the GM doesn’t simply announce (in the case of your first example), “There’s a strange symbol on the wall. Make a knowledge (arcane) roll to analyze it.” The first part of the announcement is fine, the second part should be left to the players to work out.

I often use “Obstacles” in my games. A bit of backtracking.
Trap resolution in D&D 3/3.5 is a series of skill checks: Usually Search (to find the trap) and Disable Device (to disarm the trap). The DC is set by the challenge of the trap. If the rogue fails to disarm the trap, there are consequences – usually somebody takes damage.

For an obstacle, simply replace the skills needed, retain “consequences”.

Example 1:
An arcane glylph blocking a stairway requires Knowledge (Arcane) to analyze and Spellcraft to ‘disarm’. DC for both rolls is 22. If the magician fails the second roll, the glyph explodes like a weak fireball (4d6 damage, effectively a L2 spell). In short, this is a trap to be disarmed by the party magician, not the rogue. CR = 2

Example 2:
A cave has a 20′ high waterfall, splashing onto jagged rocks (3d6 damage to fall…). If somebody climps to the top, and then jumps over to a nearby ledge, they can tie off a rope, or rope ladder, for the rest of the party. Skills required: Climb (DC 20), and Jump (DC 20). Failure means a fall onto rocks for 3d6 damage. Game effect, a CR = 1 (maybe 2) “trap” that the fighter can resolve.

It should have been added that when you have no more forward swimming speed to sacrifice, that’s when the drowning starts – or being swept towards rocks by the current, or whatever…

In fact, the notion of “taking 20″ and getting a result in a fixed amount of time irks me so much that in some of my campaigns the whole thing has been removed and replaced with a system of getting an ‘extra time required’ value based on the margin of success or failure. If you know the DC and the character’s skill total and die roll, determining how much time to spend to achieve a 20 (ie best success) is easy – but if you’re ignorant of one of those, you just have to spend as much time on the task as you think it warrants and live with the results. But it’s (unfortunately) a little fiddly in play, taking extra GM time as well, so I’ve never been completely happy with that solution, either – which is why I didn’t detail it in the blog post.

There’s another thing: time. Taking 20 takes time — enough time to try everything and decide what works best. Not so good with situations where you are in any kind of hurry.

I really, really like that. So many tasks are just impossible in D&D because you’re gonna roll a 1 eventually.