And if learned how to articulate assignments for humans right in a clear way, you likely also can do a prompt right.

None of that knowlege will get useless, only working around current limitations of agents will.