confirm-fixlisted

# confirm-fix — the half of the loop nobody runs A tester's cycle is: a test goes red → a defect is written → someone fixes it → **the exact test that proved the defect is re-run** → and then it is checked that nothing else moved. QAIA could do the first half. This skill does the second. Inputs come from `defect-report`: the scenario ID, the minimal reproduction, and the SUT version the defect was observed against. Without a scenario ID there is nothing to re-run, and the honest answer is that the defect was never reproducible. ## Three verdicts, and only three | Verdict | What was observed | What it means | |---|---|---| | **Closed** | the proving test is green, and every other test holds its previous result | the fix works and cost nothing | | **Still open** | the proving test is still red | the fix does not address what the defect described | | **Closed with collateral** | the proving test is green, **and something that used to pass now fails** | the defect is fixed and the change broke something else | **The third verdict must never be reported as the first.** That is the entire reason this skill exists. "Fixed" said about a change that broke two other tests is not a summary — it is a false statement that the next person inherits. ## Two runs, not one A confirmation needs a **before** and an **after** on the same suite. Without the before, "this test is green" says nothing about what the change cost: it may have been green already, and two other tests may have died