feat(engine): tabbed Settings panel — Phase A of the menu redesign

Replaces the 37-row single-scroll settings modal with five tabs
(Audio, Gameplay, Appearance, Accessibility, Account); only the active
tab's rows spawn, so every tab fits without scroll-hunting.

- SettingsTab + ActiveSettingsTab (session-only) + SettingsTabButton
  chips under the modal header; rebuild-on-switch mirrors the
  leaderboard despawn/respawn pattern via a shared build_panel helper
- Accessibility tab gathers color-blind, high-contrast, reduce-motion,
  touch-input, and tooltip-delay from the old Gameplay/Cosmetic mix;
  Account = Sync + Privacy
- SettingsButton enum, input handlers, and persistence untouched
- Tests: per-tab Focusable/Tooltip sweeps, tab-switch rebuild test,
  picker/thumbnail tests pinned to the Appearance tab

Plan: docs/ui-redesign-2026-07.md (also added, phases A–M)

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

docs/ui-redesign-2026-07.md

@@ -0,0 +1,204 @@
+# Menu UX Redesign — July 2026
+
+Status: PLANNING. Visual identity (Terminal / base16-eighties) is settled and
+out of scope — this is about **structure and interaction**, not colors or type.
+
+## Diagnosis (from code survey, 2026-07-07)
+
+| Surface | Today | Pain |
+| --- | --- | --- |
+| Settings | One scrolling modal, 5 section labels, ~37 control rows, single Done | Scroll-hunting for any toggle; endless on the folded (cover) screen; every new feature (theme store, import) stretches it further |
+| Home | Modal launcher: stats strip, draw-mode row, difficulty rows, 6 equal mode cards, Cancel | No "Continue" primacy; deal options (draw/difficulty/winnable) crowd the mode choice; six cards weighted identically though Classic dominates play |
+| HUD menu | Popover with 7 destinations (Help, Modes, Stats, Achievements, Profile, Settings, Leaderboard) | "Modes" duplicates Home's job; flat list, no grouping; each item opens another modal — modal-on-modal navigation |
+| Global | Everything is a modal over the felt | No hierarchy; near-square unfolded Fold renders one narrow wasted column |
+
+## Phase A — Settings tabs (highest pain, lowest risk)
+
+- Replace the single scroll with **tabs**: `Audio · Gameplay · Appearance ·
+  Accessibility · Account`.
+  - *Appearance* = today's Cosmetic (card back, background, card theme,
+    theme store, import).
+  - *Accessibility* = extracted from Gameplay/Cosmetic: color-blind,
+    high-contrast, reduce-motion, touch input mode, tooltip delay.
+  - *Account* = Sync + Privacy merged.
+- Narrow screens: tab chips in a row under the header. Wide (aspect > ~1.2 —
+  unfolded Fold, desktop): left rail, rows in two columns.
+- Only the active tab's rows are spawned → each tab is ≤ 8 rows and fits the
+  cover screen without scrolling.
+- **Implementation constraint:** pure `ui.rs` re-layout + one `SettingsTab`
+  resource. The `SettingsButton` enum, input handlers, and persistence are
+  untouched, so the 25 settings tests and the ambiguity gate stay green.
+
+## Phase B — Home becomes a real home
+
+- **Hierarchy, top to bottom:**
+  1. `Continue` card (only when a game is in progress) — mode, elapsed, score.
+  2. Hero `New Game` — one tap, reuses last mode + deal options.
+  3. Compact 2×3 mode grid (Classic, Daily, Zen, Challenge, Time Attack,
+     Seed) — smaller cards, glyph + name + one-line description on wide only.
+  4. Stats strip moves to the bottom (or right pane when unfolded).
+- **Deal options** (draw 1/3, difficulty, winnable-only) move off the top
+  level into a disclosure on the Classic card / New Game hero — they are
+  Classic-mode concerns, not global ones.
+- Cancel becomes `Back to table` and only renders when a game exists.
+- Unfolded Fold / desktop: two panes — modes left, Continue + stats +
+  daily/weekly right.
+
+## Phase C — HUD menu consolidation
+
+- Drop `Modes` from the popover (Home owns mode selection; HUD's New-Game
+  path opens Home).
+- Group the remaining six: **Play** (Home) · **You** (Profile, Stats,
+  Achievements) · **Community** (Leaderboard) · **System** (Settings, Help) —
+  section dividers in the existing popover, not a new widget.
+- Dismissal audit: Esc / scrim-tap / Done must behave identically on every
+  modal (most already do via `ScrimDismissible`; sweep the stragglers).
+
+## Phase D — Validation gates
+
+- Optional Stitch mockups (Terminal design-system asset
+  `assets/0f1652274fea460585d0b331a9ddbb06`, project 593811793268308763) for
+  Home + Settings before implementing Phase B. Phase A is safe to build
+  straight in-engine.
+- Per-phase gates: full test suite, clippy `-D warnings`, ambiguity gate at 0,
+  Android checklist (§15.3), on-device pass on the Fold 7 in BOTH postures.
+
+## Sequencing & risk
+
+| Phase | Size | Risk | Depends on |
+| --- | --- | --- | --- |
+| A — Settings tabs | ~1 session | Low (layout-only) | — |
+| B — Home hierarchy | 1–2 sessions | Medium (touches launch flow) | D mockups if wanted |
+| C — Menu grouping | small | Low | B (Modes removal) |
+
+## Phase E — "You" hub (reuses the Phase A tab component)
+
+Profile, Stats, Achievements, and the replay browser are four separate
+modals today, reached through the HUD popover one at a time. Fold them
+into ONE tabbed hub — `Profile · Stats · Achievements · Replays` — using
+the same tab widget Phase A builds for Settings. Cuts the HUD popover to
+four destinations and makes the tab component pay for itself twice.
+Existing per-screen systems keep their markers/handlers; only the outer
+shell changes (same trick as Phase A).
+
+## Phase F — Mobile ergonomics (Fold-first)
+
+- **Bottom action bar** on touch: Undo, Hint, Draw within thumb reach at
+  the bottom safe-area edge (via the existing `SafeAreaAnchoredBottom`),
+  score-only band stays on top. Desktop keeps the current top band.
+  Biggest one-handed-play win available; folded posture is tall and
+  top-heavy today.
+- **Hold-to-repeat Undo** — press-and-hold steps back repeatedly
+  (respecting the existing undo scoring penalty), instead of tap-tap-tap.
+- Both are additive; game stays fully playable with the top band alone
+  (UI-first rule §3.3 satisfied).
+
+## Phase G — Post-win flow
+
+Win summary today is a stats dump with a close. Give it an action
+hierarchy: **Play again** (same mode/options, primary) · **Share replay**
+(the upload already returns a share URL; surface it here with copy
+feedback) · **Watch replay** · quiet stats below. Rematch loop drops from
+4 taps to 1.
+
+## Phase H — Feedback polish (small, bundling candidates)
+
+- Unify queued vs. immediate toast styling (two subtly different styles
+  exist today) and anchor both to one position; consider a 5-item toast
+  history on the stats screen.
+- Theme-store modal: render the preview PNGs the server already serves
+  (v1 is text-only rows).
+- Hint: optional ghost-motion preview of the suggested move instead of a
+  static highlight (auto-disabled under reduce-motion).
+
+## Phase I — Onboarding & discoverability
+
+The first-run onboarding exists, but everything learned after it is
+invisible: the radial menu (long-press/right-click), hint cycling,
+tap-to-toggle HUD chrome, hotkeys. Add **contextual one-time tips** —
+fired by the situation, not a tour (first stall → hint tip; first
+long-press-able stack → radial tip; each shows once, stored in
+Settings like `shown_achievement_onboarding` already is). Plus a
+one-shot **"What's new"** card on first launch after a version bump —
+ObtainX updates are silent today, so shipped features go unnoticed
+(nobody will find the theme store on their own).
+
+## Phase J — Keyboard & focus completeness (desktop + web)
+
+`ui_focus` already gives deterministic Esc order and tab-walk
+(`FocusRow`/`Focusable`), and `KeyboardDragState` exists — but full
+keyboard-only play has never been audited end to end. Deliverables: every
+modal reachable/dismissable without a mouse, a visible focus ring styled
+to the Terminal system (current focus state is subtle), and a hotkey
+cheat-sheet overlay (hold `?`) generated from the actual bindings instead
+of the static Help text. Web build (Rhys) benefits most.
+
+## Phase K — UI scale & touch-target accessibility
+
+All type sizes flow through `TYPE_*` tokens and layout through
+`compute_layout` — which makes a **UI scale setting** (90/100/115/130 %)
+cheap to wire and genuinely useful on the Fold's dense cover screen.
+Pair with: respect Android system font scale on first run (seed the
+setting from it), and a one-time audit that every interactive element
+meets the 44 px logical minimum in BOTH Fold postures (the pill buttons
+and picker swatches are the suspects). Lands in the Accessibility tab
+that Phase A creates.
+
+## Phase L — Empty & loading states, standardized
+
+Async/empty surfaces each improvise today: leaderboard with no entries,
+stats with no games, replay browser with no replays, theme store
+loading/error (just added), sync status text. Define ONE pattern —
+glyph + one-line explanation + a single next-step action — as a
+`spawn_empty_state` helper in `ui_modal`, and sweep all five surfaces
+onto it. Small, mechanical, big perceived-quality win.
+
+## Phase M — Sync transparency & data stewardship
+
+The sync layer returns `ConflictReport`s ("data is never silently
+discarded") — but no UI ever shows them; players can't tell what a merge
+did. Add: a post-sync summary line ("merged, 2 conflicts kept newer
+values" → tap for detail in the Account tab), a visible last-synced
+timestamp (exists as `SyncStatus::LastSynced`, barely surfaced), and
+local **data export/import** (zip of the JSON saves) for device
+migrations without a server. Closes the trust loop the server work
+opened.
+
+## Roadmap tie-ins (already in flight elsewhere)
+
+- **Spider mode** (feat/spider-core, PR #157): Home's mode grid gains a
+  7th card once engine work lands — the 2×3 grid in Phase B should be
+  designed as N-card flow from day one. Spider UI also needs positional
+  card→entity keys (documented in spider.rs module docs).
+- **Show solution** (feat/solution-line, in flight): lands in the pause
+  menu; Phase C's grouping should leave room for it under a "Game"
+  cluster if the pause menu grows.
+
+## Sequencing (updated)
+
+| Phase | Size | Risk | Depends on |
+| --- | --- | --- | --- |
+| A — Settings tabs | ~1 session | Low | — (DECIDED: tabs) |
+| E — "You" hub | ~1 session | Low | A's tab component |
+| C — Menu grouping | small | Low | E (popover shrinks) |
+| B — Home hierarchy | 1–2 sessions | Medium | D mockups if wanted |
+| F — Bottom action bar | ~1 session | Medium (input paths) | — |
+| G — Post-win flow | small | Low | — |
+| H — Feedback polish | small each | Low | — |
+| I — Onboarding & discoverability | ~1 session | Low | ships best after B/F land |
+| J — Keyboard & focus completeness | ~1 session | Low | A/E (fewer modals to audit) |
+| K — UI scale & touch targets | ~1 session | Medium (layout-wide) | A (Accessibility tab) |
+| L — Empty-state standardization | small | Low | — |
+| M — Sync transparency & export | ~1 session | Low–Medium | A (Account tab) |
+
+Suggested order: **A → E → C** (one arc: the tab component and the menu
+slimming), then **B**, then **F/G/H** as independent follow-ups.
+
+## Open decisions
+
+1. ~~Settings: tabs vs. sub-pages~~ — **DECIDED 2026-07-07: tabs.**
+2. Deal options: disclosure on Classic card (proposed) vs. keep global row?
+3. Time Attack + Seed: top-level cards (proposed, grid stays symmetric) vs.
+   tucked under a "More" card?
+4. Stitch mockups for Phase B, or iterate directly in-engine?
+5. Phase F bottom bar: touch-only (proposed) or also desktop?