qa-planlisted

# qa-plan This does **not** write the plan. `npx playwright init-agents` ships a planner that explores the live app and writes a better one than a blind LLM pass can. This is the gate that plan must clear. ## Input A plan from `playwright-test-planner` plus the ticket's acceptance criteria. ## Checklist 1. **Every AC is accounted for.** Each one is either covered by a scenario or listed under `## Untestable` with a reason. An AC that is simply absent is the single most common way an agentic QA run reports success while proving nothing. 2. **Every scenario carries its AC tag** (`@AC1`, `@AC-P2-3`) in the title that will become the test title. `tools/report.ts` computes coverage from these; an untagged test is invisible to the gate. 3. **File paths point where the test runner actually looks.** Planners have been observed writing spec paths into the plans directory, which produces files Playwright never runs and a report of zero tests with no error. 4. **Assertions are observable outcomes**, not intentions. "Total shows $97.17", not "totals are correct". 5. **No ambiguity is left for the generator to resolve.** If the plan offers two ways to assert something, pick one now. The generator will otherwise pick the easier one. 6. **Values that must not rot are not hardcoded.** Prefer comparing the app against itself (does the pre-filled value equal what the template menu inserts?) over pinning a test to today's copy. 7. **Shared state is identified.