Capsule-style goggled characters (inspired by the goggled-buddy look, but
NOT Minions): an upright pill body with a bolt antenna and a riveted goggle
strap holding metal lenses — distinct silhouette and non-yellow shells keep
them clearly original.
- gizmoTeal, gizmoCoral, gizmoViolet, gizmoLime (two-eyed) and gizmoCyan
  (one-eyed cyclops), each expression-aware (happy/cheer/surprised/dizzy/ko).
- Registered in MascotKind + MASCOT_KINDS + PARTS, and added to the Math
  Runner and Lucky Crawl hero pickers so they're pullable.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01YJgCstFD5JgXpXb7CbsG5V

- Add 3 original "Racer" mascots inspired by kart-racer art (a cap with a
  short brim + goggles pushed up on the forehead) — no logos or real-character
  likeness. racerRed, racerGreen, racerViolet.
- Extend the Gizmo goggle-buddy family: gizmoSun (three-eyed), gizmoRose
  (spring antenna), gizmoSlate — the factory now supports 1/2/3 lenses and a
  springy antenna.
- Register all in MascotKind + MASCOT_KINDS + PARTS and add them to the Math
  Runner and Lucky Crawl hero pickers so they're pullable.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01YJgCstFD5JgXpXb7CbsG5V

A Mario-style thick-bordered carpentry builder that relates to the 6.G
Geometry unit (answers recorded under '6.G'):
- 7 levels assemble a 2-story cottage: floor (L×W), walls (L×W + 90°), trim &
  siding (SUBTRACT wall−window + 180°), roof (½·b·h triangle + 45°/15°),
  chimney + round window (rect + circle π·r², π≈3), yard/fence/mailbox
  (circle + 270°), then a Big Bad Wolf sturdiness test — survive and cute
  raccoons move in, with a house-warming bonus word problem for coins.
- Doors + square window + round window included; the house SVG builds up
  piece-by-piece with bold outlines.
- Tap answer chips, then SWIPE back-and-forth to saw each board (new
  sfx.saw/sfx.cut + buzzy haptics).
- Drawers for reference & worksheets: a "📐 Guide" drawer (area formulas with
  worked examples + angle guide + scratchpad), plus per-question "📝 How to
  solve" (ProblemAid) and the header Help.
- Registered (tile, route, mascot 'raccoon', how-to). New sfx.saw/sfx.cut in
  arcadeAV.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01YJgCstFD5JgXpXb7CbsG5V

…spector report

- New Level 8 "Set the Trap" phase (6.G): square area (side × side) as a new
  sub-skill, plus a Goldilocks sizing question (wolf's paw ≈ 36 sq units) with
  "too small — won't catch him" / "too big — he'll see it" feedback.
- Dramatic multi-scene WolfTrapCutscene: wide → zoom-in on the trap → SNAP
  (screen shake + boom) → zoom-out, advancing on a Continue button press.
- Building Inspector graded report: A+/A/B/C/D by first-try accuracy, stat tiles
  (rooms built / first-try / re-cuts), and a 6-skill checklist; Collect opens the
  word-problem house-warming bonus quiz.
- House gains a circular trap pad and a caged wolf; Guide adds a Square row.
- howto.ts updated to 8 levels + the trap/capture.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01YJgCstFD5JgXpXb7CbsG5V

Checkmate Lab is now a set of sharp "only-move" positions you play by CHOOSING
your move. Find the one right move (often a mate of your own) and you win; pick a
wrong move and the computer plays on and CHECKMATES YOU, move-by-move on the
board, then you can try again.

- 5 new positions (back-rank, Arabian, smothered, long-diagonal, mate-in-1 race),
  each with move-choice buttons: one key move + tempting blunders.
- Every key line and every blunder's punishment mate is machine-verified with
  chess.js offline (engine-free at runtime).
- Wrong move plays the forced mate with ominous fx to a "Checkmate — the computer
  beat you!" card + Try again.
- Cancelable timers fix the old setTimeout leak during playback.
- howto.ts chess copy updated for the choose-move + punishment flow.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01YJgCstFD5JgXpXb7CbsG5V

…he answer reveal

- Explanation no longer hides the worked steps behind a countdown — the "how & why"
  is readable at any time (removed the ThinkTimeCover gate).
- The configurable think-time (arcadeConfig.answerRevealSeconds) now guards only the
  ANSWER reveal: the Hint's final "Reveal the answer" tier counts down first
  ("Think first… answer in Ns") so kids attempt before peeking. Nudge/guide hint
  tiers and all explanations are instant.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01YJgCstFD5JgXpXb7CbsG5V

- New parent-configurable play-time economy (defaults, adjustable in Settings):
  gameMaxSeconds (180 = 3 min, 0 = off), extendMinutes (3), extendCoinCost (10).
- Store: adds the three ArcadeConfig fields (read via ?? default, no version bump)
  and a spendCoins(n) action.
- Gate (ArcadeWarmup): counts active play seconds; at the cap the game freezes and
  a "Time's up!" overlay offers spend-coins or do-a-lesson to add time, then the
  SAME game resumes. Admin/unlimited bypass. The cap replaces the legacy earnRatio
  budget (earnRatio only applies when the cap is off).
- Settings: new "Play time" group; relabeled the answer-reveal think-time control.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01YJgCstFD5JgXpXb7CbsG5V

- Mobile shows a 3-across grid (4 on larger screens) so many more games fit.
- Each tile is one big game-matching icon + bold name, no subtitle — chunky
  Mario-style look (thick black border, drop-shadowed icon, pressable shadow).
- Played check badge and locked (price) overlay kept; dropped the per-tile mascot.
- Fixed the Sumo Math icon to match the game.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01YJgCstFD5JgXpXb7CbsG5V

Phase 1 (code) of the big content rewrite:
- New LessonSlide format (objective/concept/example/protip/trap/summary) with
  per-domain slide banks under src/data/lessonSlides/, merged onto LESSONS.
  5.F decks fully authored (6 lessons × 12-13 slides) as the exemplar.
- LessonCard rewritten: intro → ONE video (the idea animation; extra videos stay
  in the Video Library only) → story-style slides with kind badges → interactive
  practice → wrap-up. Min-read gate is now per-kind (8s examples, 5s concepts,
  3s short slides), scaled by the existing lessonScreenSeconds admin setting;
  decks still advance only on button press.
- StorySlide: clamp beat→video-segment mapping so stories with more beats than
  rendered chapters freeze on the final frame (prep for expanded stories).
- New MathematicianDeckPlayer (full-screen story-style deck with a giant
  emoji-scene illustration pane) + deck registry; Mathematicians cards now open
  "📖 Their story" decks and keep the animated-story links.

Content (all 51 remaining lesson decks, expanded stories, 8 mathematician
decks) lands in the follow-up commits.

Co-Authored-By: Claude <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01YJgCstFD5JgXpXb7CbsG5V

… tests; rename dress game

- lessonSlides/ns.ts + ee.ts: 20 lessons × 13–15 story-style slides each
  (objective → concepts → worked examples → pro tip → traps → summary).
- mathematicianDecks.ts: full 12–16-slide decks for all 8 mathematicians with
  emoji-scene visuals and unit tie-ins.
- tests/slide-decks.test.ts: guards — every lesson 12–20 slides w/ full arc,
  stories ≥7 beats with meaty bodies, 8 decks of 12–20 slides.
  (6.RP/6.G/6.SP decks + story expansion land in the next commit.)
- Rename "Dress to Impress" → "Cool vs Warm Dress Up" everywhere (registry,
  route title, headers, instructions).

Co-Authored-By: Claude <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01YJgCstFD5JgXpXb7CbsG5V

Co-Authored-By: Claude <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01YJgCstFD5JgXpXb7CbsG5V

Co-Authored-By: Claude <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01YJgCstFD5JgXpXb7CbsG5V

Every existing beat body gains 8–15 words of concept explanation (why the math
works, tie to the unit). Every story gains 3–5 new closing beats: a worked
example with real numbers, a pro tip, a trap naming the classic mistake, and a
real-world tie-in. Player already clamps extra beats to the video's last frame.

Co-Authored-By: Claude <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01YJgCstFD5JgXpXb7CbsG5V