This builds directly on the distinction Module 5 drew between rephrasing a prompt and reviewing a Project's configuration (§2.4 there): fixing one bad response by rewording your ask is a one-off, in-conversation correction. It doesn't change anything about the next conversation.
The recurrence itself is the signal. If you find yourself making the same correction across multiple conversations, that repetition is diagnostic — it means the fix belongs at a persistent layer (Project instructions, a Skill, a knowledge source), not in your hands every single time.
"Adjusting your approach" is really two separate steps. First: fix the immediate output, in the conversation you're in. Second, and separately: ask whether this was a one-time thing or a recurring pattern worth promoting. Skipping the second question is exactly how teams end up making the same manual correction indefinitely — they got good at step one and never asked step two.
This is Module 5's core reframe again, now applied as an ongoing habit rather than a one-time setup decision: a correction made in a prompt benefits the current exchange; the same correction promoted to configuration benefits every future conversation.