This is like arguing that surely we can get rid of all our formers as soon as we have a widespread enough caste of priests.
There is no such thing as "prompt engineering", because there is no formal logic to be understood or engineered into submission. That's just not how it works.
Being good at debugging a system is based more on experience and gut feelings than following some kind of formal logic. LLMs are quite useful debugging assistants. Using an LLM to assist with such tasks takes tacit knowledge itself.
The internal statistical models generated during training are capable of applying higher-ordered pattern matching that while informal are still quite useful. Learning how to use these tools is a skill.
Higher-level languages still operate according to defined rules and logic, even if we can sometimes disagree with those rules, and it still takes time to learn the implications of those rules.
AI prompts.. do not. It's fundamentally just not how the technology works.
There is no such thing as "prompt engineering", because there is no formal logic to be understood or engineered into submission. That's just not how it works.