I think anything that takes years to learn properly (and of which the knowledge is not neatly packed into a book); communication skills, using vertical-specific language/terms correctly, integrating complex systems etc. Things you cannot take a book, read and do without countless failure, retry, mentors etc. Anything you can pick up quickly, probably anyone can pick up quickly... You cannot learn production firmware coding fast; sure you can pick up ARM c/assembly, you can read books, watch videos, but when you get a totally strange device that has 20kb (kilobytes!) of usable memory, you won't be able to get a factory ready implementation without years of experience. The same with a lot of other niche areas; sure you can learn it, sure you can probably get 80% there (wasting a lot of time on learning) but delivery is not reachable without the place burning down after.
Things that are not hard necessarily, but just a lot of work to learn to do well. Again, and again, i'm not critiquing the tech here, just the niche-ness of it; I hate frontend work and yet it took me 2 days to learn react native well enough to make a production app. If I can do that, too many other people can do that as well and I do not believe that's a viable longterm strategy. But if it works for someone then glory to them; I am older and know it will likely be gone in 10 years; most of everyone (by far) won't be millionaires by then even though they aspire to it now, so you need something for after 10 years. Go into management is an option, but I like remote work, coding and tech, so I had to find another way.
I don't think it's one skill; it's having your antennae up at all times and sensing what the next big thing might be, as well as actually liking that thing.
For me it was AWS. About 7 years ago I was sitting in "Architecting on AWS" and I had an epiphany that this wasn't a game changer, this was a game redefiner. I was jazzed about it and have been pursuing AWS jobs and expertise every since. AWS isn't the end-all and be-all, and it's abused more often than not, but it has provided an extremely lucrative career for me. Also, as much shit as Amazon gets, AWS is a very, very good product, or suite of products. It only gets more complicated every year and my idea of it evolves every year as well.
It's still going strong, but I do have my eye on other technologies and am making sure I don't become a dinosaur.
#1 Security: no one knows how to secure AWS. But I also really, really hate how the Cybersecurity organizations operate in every job I've ever had so that would feel like selling my soul to the devil.
#2 "Big Data"
#3 Containerization (working with it on AWS specifically)
