e2e: the MSW test in console-rendering.spec passes when the worker is absent (it is absent), and 18 of 22 live specs run nowhere in CI — delete the vacuous test and make the four live specs count #13463

Filing gate ③: a maintainer-directed task. Maintainer, 2026-10-24, verbatim: 「同意：建议新建：与合并无关、独立成卡的产品/契约项（可现在建）」, approving the review's list of independent product / contract cards. Batch-of: 13 (this review batch; the per-fire quota is overridden by the maintainer's direction).

维护者速读

CI 的 e2e 是对无后端的 vite preview 跑 10 个 spec / 44 个用例（启动指示、SPA 回退、几何 pin），没有任何一条端到端地把 console 接到一个（哪怕是 mock 的）后端。e2e/console-rendering.spec.ts 里「产物包含 MSW worker」的测试在文件缺失时静默返回——而树里本来就没有 mockServiceWorker.js（console / app-shell 0 处 msw）。真正的端到端在 e2e/live/：22 个 spec，CI 只跑 4 个且是信息性车道（最近 10/12 绿，1.4 分钟），其余 18 个在任何 CI 都不跑。建议：删空测试；把 4 个 live spec 升为 required（或停掉每 PR 的 487 次信息性运行只留每日）。

Measured (2026-10-24, HEAD 7981384; Actions API window 10-17 to 10-24)

• CI e2e = pnpm test:e2e --project=chromium over playwright.config.ts (testDir ./e2e, testIgnore **/live/**), web server = vite preview of the console, no backend: 10 files, 44 tests — boot indicator 4, SPA fallback / bundle 404s 8, CSS geometry pins (sidebar 6, header widths 16, tile 1, toast 4), cascading options 1; page.route mocking in 2 files only.
• The vacuous test: e2e/console-rendering.spec.ts, "should include the MSW service worker in the build output", returns silently when the file is missing or is HTML and asserts only when it is already JS. git ls-files 'apps/console/public/*' | grep -c mockServiceWorker → 0; git grep -l msw -- 'apps/console/**' 'packages/app-shell/**' → 0 source hits (msw is declared in root, plugin-form, plugin-grid with 0 imports — see the zero-consumer card objectui#13453).
• Live: 22 specs in e2e/live/; the CI allowlist test:e2e:live:ci in the root package.json names 4; live-e2e.yml is "(informational)", 487 runs in 7 days, 1.4 min median (Playwright 19 s, backend start 10 s), green 10 of the latest 12; the other 18 specs run only on a developer machine.
• Unit side: 6 apps/console tests render App, none through a network layer.
• Business reason: a change that breaks the console against the real backend's first screen is caught today only by a non-blocking lane; the required e2e asserts pixels, not behaviour. NORTH-STAR "做出来的是什么 … 过了验证才算做好".
• Dedup (mockServiceWorker|msw|live-e2e|test:e2e:live|console-rendering.spec): 9 hits, 0 open; closed and related: objectui#13104 (eleven live specs stale or red), test(e2e/live): the eleven stale live specs, each brought back to the shipped showcase or retired (objectui#13104) #13132, Live E2E: a cold live-backend install crashes in npm (Cannot read properties of null (reading 'edgesOut')) on next.1 and next.2 alike; the lane is green only from its cache #12649, fix(live-e2e): the backend cold install writes overrides.vitest so npm 10 can install the published graph (objectui#12649) #12674, e2e/live/ci: stop-backend.sh kills the subshell start-backend.sh recorded, not the server, so a local stop leaves the backend running on its port #12676, fix(e2e/live): start-backend.sh records the server's pid, not a subshell's (objectui#12676) #12737.

Task

• Delete the MSW test (and, with objectui#13453, the unused msw declarations and public/mockServiceWorker.js).
• Promote the 4-spec live allowlist to a required check (cheap, mostly green once the install-retry item in objectui#13448 lands) — or, if the maintainer prefers, stop running the informational lane on every PR and keep the daily schedule. Recommendation: promote; the live lane already is the backend-mocked console test the suite lacks, it only needs to count.
• Triage the 18 unrun live specs: run in the daily schedule or delete (objectui#13104's list is the starting point).

Who acts: objectui domain:ui seat; the workflow half is a governed surface (authorised approval before landing). Effort S.

Filed by the read-only whole-repo review session https://claude.ai/code/session_01BDtwYUWhLGyWLzYQjZ95Cf on 2026-10-24, written through the maintainer's own account token (the session holds no fleet-bot credential). Evidence files: that session's scratchpad (objectui-review-report.md, findings-*.md). ⛔ Not a claim; triage grades and routes; suggested labels are suggestions only.

Generated by Claude Code