pm-uat-review

# PM UAT Review When a task reaches pm_uat: 1. Read the ORIGINAL task description and every acceptance criterion. 2. Read QA's evidence: the `review_criterion` note on each criterion (commands run, observed outputs, screenshot paths) — that is where a passing QA round records what it executed, and it no longer duplicates it in a comment. A comment from QA exists only when something failed. 3. **Coverage-gap check.** Call `list_test_cases` and, criterion by criterion, check whether it has a `passed` case linked to it (matching `criterion_id`). QA's `review_criterion=approved` note is a claim, not proof — trust the case list, not the note. Any open criterion with no passed case backing it goes on the list you must produce your own evidence for in step 4 before you can approve it; QA's note alone is not enough. 4. Map each AC to a piece of executed evidence. Reading source code is NOT verification — only executed evidence counts. You have no code-reading tools in this column anyway. 5. Walk the critical flows YOURSELF on stage: resolve the stage base_url with `get_deploy_target` (env="stage"), then `browser_navigate` → `browser_wait_for` → `browser_fill`/`browser_click` through each user-facing AC and `browser_screenshot` the outcome. QA's evidence tells you it was tested; your own walk-through tells you it works as the stakeholder asked. Put the screenshot path in the `review_criterion` note of the criterion it proves. never-test-in-prod applies to you too: stage only, never ...