Here is my attempt at a complete list of skills a course needs to test.
But first I think the test needs to be dialed in to a specific throwing distance. For me, a 400' hole is good to have since it proves that I cannot throw 400' (and allows others to outscore me who have this skill). But if the majority of the course forces throws that require all sorts of shot shapes that also require 400' D to score well on.....then that course fails at really testing my other skills.
In my opinion holes should vary in effective length approximately +/-20% of the target D. So, for Blue level players, average golf D is around 340'.....so the majority of holes should vary between 275' and 400'.
For Blue the course should force these throws:
375-400' straight open-ish (2 of these per course)
325' straight open-ish
325' straight open-ish finishing R (rewards (RHBH or LHFH)
325' straight open-ish finishing L (rewards LHBH or RHFH)
300' R to L fairway curve finishing L
300' L to R fairway curve finishing R
300' R to L fairway curve finishing R (S-curve)
300' L to R fairway curve finishing L (S-curve)
275' tunnel straight (pinch points at over 12-16' wide at mind point)
275' tunnel curving R finishing right (pinch points at over 12-16' wide at mind point)
275' tunnel curving L finishing L (pinch points at over 12-16' wide at mind point)
275' tunnel curving R finishing L (pinch points at over 12-16' wide at mind point)
275' tunnel curving L finishing R (pinch points at over 12-16' wide at mind point)
Flat, uphill, and down hill versions of each.
Green types and other "course management" could/would be varied throughout.
As you see, there are around 14 variations, so to factor in terrain adequately, you need well over 24 holes (throw segments) to even consider calling a course a complete test of skills.
One other thought: IMO, rollers, thumbers, tomahawks, scoobies, etc are specialty shots that do not have to be tested per se. They come in handy and help out in scoring but there is such a huge variety of of shot shapes they add, that the number of permutations of holes one would need to test all of them is staggeringly huge (let alone extremely challenging to construct).