It is interesting to repeat some mechanics using video as an example.
A text assignment can always be understood in different ways.
But so long as it follows the letter, its all good. Usually though, you define the inputs that need to be handled and then define acceptable solutions. So if I give you a problem with an input and define acceptable outputs, so long as the output meets the criteria, is all good.
The toolkit should be complete, what you write without using behaviors turns the challenge into a programming competition. Competitive olympiad solutions are very rarely used in practice.
I mean, challenges are about, the challenge... so yeah... I agree, some challenges should use the whole toolkit, but the challenge still needs to be quantifiable.
Of the second point, Okay, I disagree - knowing how to recreate the wheel is important and applicable... what of when the provided tools fail. Pathfinding doesn't allow you to iterate through the mesh yourself. The platformer behavior solves collisions in a very particular way that doesn't work with "soft ejection" found in most retro games. To build a clone of mario3, using the the built in platformer behavior won't help you. It solves the collisions all wrong and you won't be able to truly recreate how Mario interacts with solids. Every built in behavior will fail if you have specific needs it doesn't cover.
I read your example assignment and realized that I am not interested in this kind of work at all.
What types of challenges would you find enjoyable if you were participating?