Do caster out of combat options matter?

[Skrum]

I'm toying with the 'every skill DC is 15', if you fail by 5 or less, you succeed with a drawback. If you succeed by 5 or more, you have a critical success.

I've thought of this too, but every time I've considered it in practice it becomes a massive headache. What the drawback is, what super-success is, like that's specific to not only the skill being used but even the particular context. The social skills are relatively easy to imagine in this regard, but what does it mean to "crit success" jumping over a ravine? Or sneaking? And then there's knowledge skills: now the DM has to come up with 2, 3, 4 layers of info to correspond to varying levels of success. TBC, I've done this before (esp. for knowledge checks), but it's time consuming and only easily works for some skills.

But I'm also thinking of using 3rd Ed style Skill Points, maximizing 1 per character level. If your Skill Points equal your current level, you don't have to roll for that skill, you just succeed as normal. But you can roll if you'd like, for the chance at a critical success. Skill points probably wouldn't provide a bonus to the roll, just determine your ability to even make the attempt, and if you can auto succeed or not. Rolling will always just be d20+Mod+PB.

Some things I've toyed around with (but never play tested)

1) Untrained skills have a DC cap of 8 + proficiency bonus. DC's above that cannot be attempted by someone without proficiency in that skill
2) Expertise grants a "take 10" feature; instead of rolling, the character may count the check like the rolled a 10. This cannot be used in combat or other hostile situation
3) Not really a mechanical change but a table of DC's to use as a reference is really needed. My general opinion is skills tend to be too high of a DC; someone who is trained in something should probably be succeeding at that thing like 75% of the time. The norm DC should be in the 10-12 range; higher is notably hard things to do
4) Yeah, something like total fail, partial fail, success, and total success would be nice (but see above comments)

But - none of this addresses what might be my core problem. There's no agency in skill checks. It's just use the obviously best, highest check you possibly can. I guess something like the Varying Success system could address some of this? Have poorly matched skills have terrible downside risk? But again, that's just a lot of DM time to make it feel right.