fix(visual-proof): outputPath escape + networkidle timeout (W8-14)

[Ghenghis]

Summary

Surgical fixes to W6-6's Playwright visual-proof harness so it produces real match/diff data against Images-GUI/. Stacks on PR #188 (W8-12 Vite Fast-Refresh). Builds on PR #189 (W6-6 first-run JSON summary).

• outputPath is not allowed outside of the parent directory — snapshotPathTemplate: \"{arg}{ext}\" plus spec segments that escaped the test root tripped Playwright's safety check. Fix: new tests/visual/global-setup.ts mirrors Images-GUI/ into tests/visual/__refs__/ (idempotent, size+mtime check, one-way only). Spec now emits forward-only segments into __refs__/. The mirror is gitignored; Images-GUI/ remains the single source of truth.
• 30s networkidle timeouts — the Hermes3D SPA's long-poll + SSE channels never go idle. Per Playwright's docs, networkidle is DISCOURAGED for this shape. Fix: drop networkidle from waitForStable(); keep domcontentloaded + 500ms quiet + the per-target wait_test_id toBeVisible(15s) assertion that already proves the route mounted.

Re-run Delta

03_implementation/docs/evidence/visual_proof_2026-05-09/summary.json:

	Before (PR #189)	After
match	0	0
diff	0	11
error	11	0
skipped-future	20	20

All 11 live targets now produce real pixel-diff data. The harness end-to-end works.

Persistence — 11 unfixed diffs (single shared cause)

All 11 live targets exceed tolerance with ratio ~0.28-0.31. Single shared root cause: reference PNGs were captured at 1536x1024, harness viewport is 1920x1080. Not 11 independent UI defects. Owned by W6-3 to fix (either re-render references at 1920x1080 or drop viewport to 1536x1024). See handoff doc table for per-target evidence_path / diff_path.

Sources

1. https://playwright.dev/docs/api/class-testconfig#test-config-snapshot-path-template (parent-dir safety check that triggered Bug 1)
2. https://playwright.dev/docs/api/class-page#page-wait-for-load-state (networkidle marked DISCOURAGED)

Test plan

• Single-target run: npx playwright test --config playwright.visual.config.ts -g \"01_dashboard_advanced_a\" — completes in 2.3s with real diff data (diff_pixels=594010 ratio=0.29).
• Full harness run — 31 targets in ~38s; reporter writes summary.json with 11 diff / 20 skipped / 0 error.
• globalSetup idempotent — second run hits (copied=0 skipped=31 pruned=0).
• No commits include mirrored PNGs (verified via git check-ignore against tests/visual/.gitignore).
• W6-6's design preserved — manifest, reporter contract, updateSnapshots: \"none\", fail-on-missing-baseline all unchanged.

Hermes evidence chain: PASS

• Task ID: W8-14-VISUAL-PROOF-HARNESS-FIX-2026-05-09
• Lock owner: claude-w8-14-harness-fix
• hermes_run_gate: not applicable (no policy gate for visual-proof)

🤖 Generated with Claude Code

[gemini-code-assist]

Warning

You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again!

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: 0b839def-81e0-4f15-935b-0a0d0475f71e

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch claude/w8-14-visual-proof-harness-fix

---

_{Comment @coderabbitai help to get the list of available commands and usage tips.}