Outcome: every line gets tagged — capability zone (checkable target: a number, a format word, an audience/grade level, a yes/no rule), limitation zone (a vague quality directive with nothing to verify against), or neutral (context, not an instruction). Tick which lines to keep, get a real prompt to fix the flagged ones, paste the fixes back, and the tool assembles your refined prompt.
Useful for: turning "did I write a good prompt?" into a concrete line-by-line pass — catching instructions the model will honor literally but where your actual intent has nowhere to land.
This scan is a heuristic word-pattern check, not a judgment of your prompt's quality — it looks for numbers, format keywords, audience/grade-level phrasing, and known vague-quality words. It will miss nuance and occasionally mislabel a line; that's why every line has a checkbox and nothing is decided for you.