diff --git a/.planning/phases/06-ux-polish/06-VALIDATION.md b/.planning/phases/06-ux-polish/06-VALIDATION.md
new file mode 100644
index 0000000..b31d561
--- /dev/null
+++ b/.planning/phases/06-ux-polish/06-VALIDATION.md
@@ -0,0 +1,82 @@
+---
+phase: 6
+slug: ux-polish
+status: draft
+nyquist_compliant: false
+wave_0_complete: false
+created: 2026-06-10
+---
+
+# Phase 6 — Validation Strategy
+
+> Per-phase validation contract for feedback sampling during execution.
+> Source: `06-RESEARCH.md` § Validation Architecture. Planner fills the Per-Task Verification Map below as plans are written.
+
+---
+
+## Test Infrastructure
+
+| Property | Value |
+|----------|-------|
+| **Framework** | vitest (both `apps/pwa` and `apps/api`) |
+| **Config file** | `apps/pwa/vitest.config.ts`, `apps/api/vitest.config.ts` |
+| **Quick run command** | `pnpm --filter @familysync/pwa test` (PWA-side fixes) · `pnpm --filter @familysync/api test` (recurrence write/expand) |
+| **Full suite command** | `pnpm -r test` |
+| **Browser verification** | `playwright-cli` skill (desktop Chromium) for visual/behavioral items — per CLAUDE.md convention |
+| **Estimated runtime** | ~30–60 seconds (unit); browser checks additive |
+
+---
+
+## Sampling Rate
+
+- **After every task commit:** Run the quick run command for the touched app
+- **After every plan wave:** Run `pnpm -r test`
+- **Before `/gsd-verify-work`:** Full suite green + playwright-cli checks pass
+- **Max feedback latency:** 60 seconds (unit)
+
+---
+
+## Per-Task Verification Map
+
+> Planner: populate one row per task as plans are authored. Test-type guidance from RESEARCH § Validation Architecture:
+> - **unit** — duration-preservation math (`eventDateTime.ts`), RRULE UNTIL/COUNT serialization (`vevent.ts`), inclusive↔exclusive DTEND round-trip, FREQ-persistence regression (D-07), opaqueredirect/401 detection (`client.ts`), `hasRrule` population in `expandOccurrences()`.
+> - **integration** — PWA→API→expand recurrence round-trip (bounded series renders correct occurrence count, per-occurrence duration = start→end delta).
+> - **browser (playwright-cli, desktop Chromium)** — all-day visual distinctness (999.6), spinner/pulse actually animating (D-13), no calendar/"sign-in required" flash on cold load (999.2), clean session-expiry redirect (999.3), whole-series edit prompt (999.9).
+
+| Task ID | Plan | Wave | Decision | Threat Ref | Secure Behavior | Test Type | Automated Command | File Exists | Status |
+|---------|------|------|----------|------------|-----------------|-----------|-------------------|-------------|--------|
+| TBD | TBD | TBD | D-XX | T-6-XX / — | {expected or N/A} | unit/integration/browser | `{command}` | ✅ / ❌ W0 | ⬜ pending |
+
+*Status: ⬜ pending · ✅ green · ❌ red · ⚠️ flaky*
+
+---
+
+## Wave 0 Requirements
+
+- [ ] Confirm existing vitest infra covers new test files (no framework install needed — vitest already configured in both apps)
+- [ ] Test stubs for: duration-preservation, RRULE UNTIL/COUNT, FREQ round-trip, session-error detection
+
+*Existing infrastructure (vitest) covers all phase test types; Wave 0 is stub creation only.*
+
+---
+
+## Manual-Only Verifications
+
+| Behavior | Decision | Why Manual | Test Instructions |
+|----------|----------|------------|-------------------|
+| iOS-Safari standalone behavior (only if a fix regresses install/standalone) | D-10/D-11 | Cannot be driven by playwright-cli (per CLAUDE.md exception) | Add to Home Screen, cold-load, confirm splash + redirect on real iOS device |
+
+*All other phase behaviors have automated verification (vitest) or desktop-Chromium browser verification (playwright-cli).*
+
+---
+
+## Validation Sign-Off
+
+- [ ] All tasks have `<automated>` verify or Wave 0 dependencies
+- [ ] Sampling continuity: no 3 consecutive tasks without automated verify
+- [ ] Wave 0 covers all MISSING references
+- [ ] No watch-mode flags (`vitest run`, not `vitest --watch`)
+- [ ] Feedback latency < 60s
+- [ ] `nyquist_compliant: true` set in frontmatter
+
+**Approval:** pending