You haven't been keepign track of the advances in robotics. I'll give you OB Gyns, as most women would probably be reluctant to trust a robot. (Note that this isn't capability based, but trust based.)
You ask "Why are the robots making this stuff?" The answer is "To trade with other people who also own robots that make other stuff." Each robot factory will have multiple possible products. This will come at the cost of being less efficient at the manufacture of each one, but it will mean that there's no reason to just crank through as much as you can. It will generally be "bespoke trade", i.e., you get the order before you make the stuff. (Sort of like JIT carried to the next level.)
What skilled jobs do you think will be available, that robots won't be able to do better? In what year? (Part of the problem is that it's a moving target. It can take 10 years to turn a high school graduate into a skilled professional, and at the start of those 10 years, you don't know which jobs will be available at the end. Well, one can generally make a decent mid-course correction for the first 5-6 years, so if your general area of prediction is correct, you can adopt the most promising remaining specialty. But things become a lot less flexible as you get closer to the end of training. And if your career becomes obsolete two years after you finish training, you get to start back several years, if you can get a school to take you, only now you've already got a huge debt...or, at minimum, a huge sunk cost that you need to write off.
Currently I'm really reluctant to advise kids to go to college. If they don't want to badly enough that they would go against my advice, it looks like a bad bet. Being a welder might actually be a much better choice. The training is both shorter and less expensive. People tend to have really silly ideas about what skills a robot will have 10 years from now. They will range from nearly human down to insectile, and be priced inversely. But a huge lot of what we do isn't based on intelligence, it's based on organized memory, and while it might take intelligence to organize the memory, it doesn't take much to use memories that are already organized. And robots can copy memories easily. So you need one really intelligent robot to train to do something, and the contexts in which which action is appropriate. Then you copy it to thousands of robots that are a lot less intelligent, but a lot lighter and cheaper. And that's only 10 years from now. 20 years from now is a whole different ball game. And in between then and now things aren't going to be standing still, even if I can't predict just how they will be changing.