An AI agent only checks what it thought of: the stage order that found 9 blocking issues before the first line of code

An agent that reviews its own work checks it against the same failure model that wrote it. Putting a fresh-context reviewer on the plan before any code existed returned 9 blocking issues on one stage; skipping that step on an earlier stage left 2 serious defects for the code review, one of them a silent merge of unrelated records.