As far as I can tell, yes, the workflow is "assume the LLM can fix it."

I've certainly seen people on this site defend that idea vociferously.

I won't tell you it's impossible - Claude does do legitimately amazing things.

It does not, however, seem to me to actually have deep understanding of what it's doing. "Amazing" does not mean "admirable" or "trustworthy."