IMO the crux of understanding skilled play in OSR terms is understanding how early D&D play worked and then understanding the techniques players could use to more efficiently navigate the obstacles. But all that ultimately entails is applying the regular term 'skilled play' to an early D&D or OSR game.
To add to this:
The key technique in classic D&D play is
the players engaging with the fiction, especially the immediate geography and architecture, and having the GM extrapolate from that engagement.
This is where mechanics then come into play. If the engagement with the fiction is
I poke it with a 10' pole, normally no mechanical mediation is required - the GM just extrapolates and described what happens next. If the engagement is
I push the statue over, then it is very common for a mechanic (say, a bend bars roll if it's a big statute) to be invoked, and the result of that check will determine how the GM extrapolates. In some cases, like
I force open the stuck door or
I pick the lock, mechanical intermediation is almost mandatory.
As I and some others posted in another thread, there is no deep logic to the way classic D&D sometimes does and sometimes doesn't require mechanical mediation between
player engagement and
GM extrapolation. It reflects the way that Gygax et al did it, and not much more than that. There's no deeper system logic (cf, say, the structure for skill challenges in 4e D&D). And it's very easy to come up with
player engagements with the fiction where it's completely unclear whether mechanical mediation is a necessary precursor to
GM extrapolation. For instance (and coming out of another recent thread), suppose a player wants his/her PC to shoot a flaming arrow at a static, taut rope. Does the player just declare and the the GM extrapolate what happens when the arrow hits the rope? Or is some sort of to hit check required? Is this in the same territory as poking things with poles? Or in the same territory as forcing open locked doors? You can read and reread Gygax's rulebooks from cover to cover, but they won't give you an answer, and they won't even give you a principle out of which you can come up with an answer. You'll just have to make a decision of your own.
What will be fundamental, though, is that the
purpose of any mechanical mediation that is called for will simply be to establish whether or not the PC's engagement with the fiction actually took place as the player hoped (
did I knock the statue over? did I burst the door open? did I pick the lock? did I hit the rope with my flaming arrow?) Determining what follows from that engagement with the fiction is not a mechanical process at all in this sort of play. The GM decides, by extrapolation.
This is also the great limit on this sort of play. Because I think it's obvious that with a certain degree of complexity the idea that the GM is
extrapolating rather than just
making stuff up becomes hard to sustain. My own view is that some wilderness adventures, and basically all urban adventuring, fall on the unhappy side of this line. I regard White Plume Mountain, Tomb of Horror and Hidden Shrine of Tamoachan as exemplars of what the happy side of the line looks like.