Repair Playwright visual CI gate - #1908

Summary

• quarantine known-red Playwright visual audit specs in an explicit tracked list
• select the non-quarantined visual audit subset deterministically
• make the Playwright visual job blocking again by removing continue-on-error and the warning-only fake failure step
• add quality tests that fail if the selector is bypassed or continue-on-error returns

Validation

• pnpm exec vitest run --config scripts/quality/vitest.config.ts scripts/quality/playwright-visual-audits.test.ts scripts/quality/ci-worker-suite.test.ts
• pnpm exec tsx scripts/quality/select-playwright-visual-audits.ts > /tmp/playwright-visual-audits.txt && cd apps/web && CI=true xargs npx playwright test --project="iPhone 14 (390x844)" --reporter=list < /tmp/playwright-visual-audits.txt — 29 passed in 36.3s
• pnpm lint — passed with existing warnings
• pnpm typecheck — passed; existing Astro template baseline reported
• pnpm test — passed: 21/21 tasks, API 8248 tests, web 3510 tests
• pnpm build — passed
• pnpm format:check — passed

Staging

Skipped: CI/test-tooling-only change; no deployed app/runtime behavior to validate on staging.

Review

• test-engineer: PASS — selector tests, workflow regression tests, exact Playwright gate all passed
• constitution-validator: PASS — no new hardcoded internal URLs/timeouts/limits/deployment identifiers
• task-completion-validator: PASS — research findings, checklist, acceptance criteria, diff, and tests align

Notes

This keeps the existing PR + web-ui path gate. The quarantined specs remain explicit and reviewable, with stale-spec cleanup tracked separately in tasks/backlog/2026-07-17-stale-playwright-audit-specs.md.

Agent Preflight

• Preflight completed before code changes

Classification

• external-api-change
• cross-component-change
• business-logic-change
• public-surface-change
• docs-sync-change
• security-sensitive-change
• ui-change
• infra-change

External References

N/A: repository-local CI workflow, task history, and Playwright audit files only.

Codebase Impact Analysis

Touched .github/workflows/ci.yml, apps/web/tests/playwright/visual-audit-quarantine.txt, scripts/quality/select-playwright-visual-audits.ts, and scripts/quality/*playwright*/CI quality tests. No runtime API, UI, database, or deployment app behavior changes.

Documentation & Specs

N/A: no public docs/spec surfaces changed; task tracking was archived in tasks/archive/2026-08-25-repair-playwright-visual-ci.md.

Constitution & Risk Check

Checked Principle XI for new tooling/config code: no hardcoded internal URLs, business timeouts, runtime limits, or deployment identifiers were introduced. Main risk is CI false-negative/false-positive behavior, covered by selector tests, workflow wiring tests, and the exact Playwright CI command.