Is there a best practice around which model + setting to use for /skill-creator?

[wilfredom]

Trying to make sure I use /skill-creator well, I'm curious if there's a consensus around what settings produce the best skills? I.e., do I use Opus 4.7 (with or without adaptive thinking?), Opus 4.6 (w/ or w/o Extended Thinking), Sonnet (ditto), etc. I know I can test all of these and test each resulting skill, but that feels like it'd be very token-intensive (as in, designing the same skill with all 3 models, testing them, etc.). I'd rather save the tokens if there is a consensus / trend that I can lean on.

Thanks!