qa-uilisted

## Resolve the verification path Resolve the current plan from an explicit path, then the current non-delegated session. When `## 実装準備 > 手動QA手順` exists, its QA-ID methods are authoritative: `manual` means human handoff and `auto` means the planned shell command. Ignore caller-supplied alternate execution strategies, ledger paths, browser execution, or `ui-evaluator` wording. If that QA-ID source is absent, warn and derive browser targets in this order: 1. explicit UI requirements in the invocation; 2. UI-related AC from `<plan>.analysis.md`; 3. `差分`/`未実装` rows in `## 正本抽出結果` (`一致` is excluded); 4. UI files in the branch diff, with minimum render and console-error checks. Assign `QA-F-01...` in source order. This browser fallback uses [references/browser-fallback.md](references/browser-fallback.md), not a ledger. If no UI target exists, report that QA is unnecessary. For any handoff or browser target, resolve the URL in order: complete argument URL; non-placeholder preflight base URL plus argument path; concrete plan URL. If only a path or no base remains, ask once; a noninteractive run returns `不足入力: ベ��ス URL` without later steps. Never hardcode or auto-login. Use preflight login guidance and ask once for unresolved role-specific accounts. Prepare only test data commands explicitly documented in preflight, plan, README, or tests. Do not invent commands. Ask once for remaining QA-ID-specific states/data; unavailable items become unverifiable. ## QA-ID ledger Use `<plan