requirement-ambiguity-huntlisted

# Requirement Ambiguity Hunt > **Standalone adaptation.** This is a self-contained version of the `need-understanding` skill > from [QAIA](https://github.com/QAIA-Project/QAIA) (MIT), packaged as a single file for > directories that expect one. The canonical version, its `references/`, and the evaluation > evidence behind the claims below live in that repository. QAIA is pre-alpha and says so. ## When to use this Before designing tests for any story, and whenever a specification looks incomplete, contradictory, or open to interpretation. Also when someone asks *"what should I clarify with the product owner?"*. ## The failure mode this exists to prevent Almost any generator will produce a test for *"above €500 requires finance approval"*. The question is what it does with **exactly €500**. Pick silently and you ship a suite that looks complete and encodes a guess — at the boundary, which is where the defects are. The test then passes, looks green, and proves nothing about the actual rule. Worse, the guess is invisible: nothing in the output distinguishes an asserted fact from an asserted assumption. **Never silently resolve an ambiguity.** A surfaced question costs one line. A resolved one costs the credibility of every green run in the suite. ## Steps ### 0. Nothing-to-understand check If the source has no capability or behaviour to test — a design doc, an RFC process, an empty template, a title with no acceptance criteria — **do not fabricate requirements**. Say wh