/* ============================================================================
   符咒课 · TALISMAN CRAFT (HANDS-ON v2) — prom/js/drills/talisman2.js
   ============================================================================
   Benchmark: Okami. Almost everything here is drawn on the canvas — the
   talisman sheet, the yaoguai trouble, the wet ink and the seal chop are all
   painted in JS, because ink that dries has to be a material and not a CSS
   class. This sheet is only the FURNITURE around the paper: the case slip Rip
   reads from, the ink shelf, the verdict, and the shell the night's characters
   are carved into.

   Materials are the kit's and nothing else: ink #241335, cream stock #f2e4c2,
   lantern gold #ffc84a, vermillion #e04a3a. One accent (gold). No second hue.
   Loaded after drill-kit.css; the drill claims the sheet itself (ensureCss).

   THEME NOTE, because it caused a shipped blocker. prom.css PINS the night
   tokens inside `.minigame` in BOTH themes (`--card: #2a1c44`, `--paper:
   #f2efe9`, `--haze: #a49fae`) — a drill stage is scenery, and the theme toggle
   only rules reading surfaces. So a `data-theme` rule here would be dead code;
   what matters is picking the right SIDE of the kit's palette per material:
     · on .dk-plate  (cream stock)  → --dk-ink / --dk-ink-soft / --dk-verm-deep
     · on .dk-slab, .dk-scrim (night) → --paper / --haze / --dk-gold / --dk-verm
   Putting an on-cream token on the night slab is what made 墨 measure 1.07:1.

   TYPE — exactly the kit's four steps, and only these four:
     display  the sealed characters on the shell, and the verdict's character
     read     hanzi the player must read: the case slip, the ink label, buttons
     label    latin the player must read: button faces, the verdict's reason
     caption  gloss and pinyin only — never the only home of a fact
   ========================================================================= */

.tk-game {
  --dk-accent: #ffc84a;              /* the drill's ONE accent: lantern gold */
  --tk-lane: 0px;                    /* room reserved for Rip's cutout on wide screens */
  width: 100%; display: flex; flex-direction: column; align-items: center;
}

/* ------------------------------------------------------------------- COURT */
.tk-court {
  display: flex; flex-direction: column; align-items: center; gap: 10px;
  width: min(760px, 94vw); margin-top: 8px; padding-left: var(--tk-lane);
}
@media (min-width: 1080px) { .tk-game { --tk-lane: 150px; } }

/* -------------------------------------------------- THE CASE SLIP (案卷)
   A paper docket, not a banner: what is wrong tonight, in Rip's hand.

   THE SLIP IS CHINESE ONLY WHILE THE ROUND IS LIVE. It used to print the gloss
   from the first frame, which is the standing audit finding — a gloss on screen
   turns comprehension into translation and leaves the Chinese carrying nothing.
   The gloss now arrives at the resolve and nowhere earlier (.tk-told), under a
   torn rule, so it reads as the case being ANSWERED rather than as a caption
   that was always there. Hanzi at reading size, gloss at caption size, and the
   line never runs past the measure. */
.tk-slip {
  width: 100%; padding: 9px 14px 10px; display: flex; align-items: flex-start; gap: 12px;
  position: relative;
}
/* The punched string hole of a real 案卷 — and it used to be drawn as an OUTLINED
   circle, which in this game is the shape of an empty slot or a control. The
   round-12 judge listed it as unidentifiable: "might be a button to hear it read
   aloud, might be a status light." A hole is not a ring: it is a shadowed dent
   in the stock, so it is drawn as one and stops competing. */
.tk-slip::after {
  content: ""; position: absolute; top: 9px; right: 11px; width: 9px; height: 9px;
  border-radius: 50%; background: rgba(36, 19, 53, 0.16);
  box-shadow: inset 0 1px 1px rgba(36, 19, 53, 0.3);
}
/* The case number down the left edge, in INK.
   ROUND-6 JUDGE FINDING: it used to be --dk-verm-deep, on the reasoning that
   the deep twin is the legible vermillion on cream. Legibility was never the
   problem — SEMANTICS was. 案 is a filing mark on a docket. It is not an
   error, it is not "about to lapse", and vermillion in this game is allowed to
   mean one thing. Ink at 0.68 measures 5.4:1 on the cream slip and says
   "quiet label", which is what this mark is. The vermillion that survives in
   this drill is on the PAINTING (the cinnabar rule around the ward), never in
   the chrome.

   ROUND-12 JUDGE: "a single Chinese character in a little box — a label for the
   panel, like a heading. No idea what it says." The mark is a LABEL, and the
   house bilingual rule covers labels, so the English rides under it exactly the
   way it does on every button face. The SENTENCE beside it stays Chinese-only
   until the resolve — that is the drill's one unglossed block and it is the
   puzzle itself; glossing it would answer the question before it is asked, and
   an unglossed heading is what made the unglossed sentence look impenetrable
   rather than deliberate. The vertical 案卷 setting goes: a rotated two-glyph
   column with a latin caption under it is unreadable at this size. */
.tk-slip-mark {
  flex: 0 0 auto; display: flex; flex-direction: column; align-items: center;
  gap: 1px; padding-top: 1px; color: var(--dk-ink-soft);
}
.tk-slip-mark .han {
  font-family: var(--han); font-size: var(--dk-read); font-weight: 900;
  letter-spacing: 0.14em; line-height: var(--dk-lh-tight, 1.15);
}
.tk-slip-mark i {
  font-style: normal; font-size: var(--dk-caption); font-weight: 800;
  letter-spacing: 0.02em; white-space: nowrap; line-height: 1.2;
}
.tk-slip-body { min-width: 0; }
.tk-slip-zh { font-family: var(--han); font-size: var(--dk-read); font-weight: 800;
  line-height: var(--dk-lh-read); color: var(--dk-ink); max-width: var(--dk-measure); }
/* THE BLANK, ALREADY FILLED. talisman2.js's blankHTML() wraps the one target
   character inside the case sentence in this — the fill-in-the-blank slot the
   case is grammatically built around, with the answer sitting in it from the
   first frame (see the file header for why: guessing which character a scene
   wants is no longer the game). Dashed gold rather than a solid box, so it
   still reads as a BLANK — filled — rather than as a stamp or a button; gold
   is the kit's "look here now" and this is exactly the thing to look at. */
.tk-blank {
  display: inline-block; padding: 0 3px; margin: 0 1px;
  border: 2px dashed var(--dk-gold, #ffc84a); border-radius: 5px;
}
/* The gloss: hidden for the whole live round, then unrolled once the character
   has been committed and judged. The dashed rule above it is the slip's own
   fold, so the answer reads as a second entry on the docket. */
.tk-slip-en { font-size: var(--dk-caption); font-weight: 700; line-height: 1.4;
  color: var(--dk-ink-soft); max-width: var(--dk-measure);
  margin-top: 6px; padding-top: 5px; border-top: 1.5px dashed rgba(36, 19, 53, 0.34);
  animation: tk-told-in .34s cubic-bezier(.2, 1, .3, 1) both; }
.tk-slip-en[hidden] { display: none; }
@keyframes tk-told-in { from { opacity: 0; transform: translateY(-4px); } to { opacity: 1; } }
@media (prefers-reduced-motion: reduce) { .tk-slip-en { animation: none; } }

/* -------------------------------------------------------------- THE PAPER
   The canvas is Rip's WRITING DESK seen from above — lacquer, a felt mat, the
   talisman sheet, an inkstone on the left and a paperweight on the right. Every
   object on it is a thing a calligrapher owns.
   The earlier build painted a fake night sky over the real painted courtyard —
   stars, a roofline, and two lantern posts whose lit ellipses read, correctly,
   as "two unexplained gold ovals flanking the paper". A desk cannot have
   unexplained furniture: everything on it is either the sheet or a tool. */
.tk-papercase { width: 100%; position: relative; line-height: 0; }
.tk-canvas {
  width: 100%; height: auto; display: block; border-radius: 14px;
  border: 3px solid var(--dk-ink); box-shadow: 0 5px 0 rgba(14, 6, 24, 0.8);
  touch-action: none;                 /* the brush owns the pointer; no scroll-drag */
  cursor: crosshair;
}
.tk-canvas.dry { cursor: crosshair; }
.tk-canvas.busy { cursor: progress; }

/* ------------------------------------------------------- THE ONE LIVE MARK
   The kit's .dk-live ring answers "which of these do I touch", and on every
   other board it goes straight onto a button. This drill's first verb is a drag
   on a CANVAS, so the ring rides a zero-interaction proxy that talisman2.js
   (paintLive) lays exactly over the start-of-stroke wisp — the pixel the brush
   is meant to land on. It has no face of its own on purpose: the wisp, the
   painted 起笔 plate and the paper underneath are already there, and a second
   filled disc on top of them would hide the target the ring is pointing at.
   Round, so the kit's outline (which follows border-radius) draws a circle
   rather than a box around a circle. pointer-events:none is load-bearing —
   the press has to reach the canvas, or the ringed thing and the pressed thing
   stop being the same place. */
.tk-livespot {
  position: absolute; z-index: 2; pointer-events: none;
  border-radius: 50%; box-sizing: border-box; background: transparent;
}
.tk-livespot[hidden] { display: none; }

/* Rip has the brush. This is the whole reason the demonstration used to feel
   broken: pointerdown was refused while she wrote and nothing said so. */
.tk-watch {
  position: absolute; left: 50%; top: 10px; transform: translateX(-50%);
  display: flex; align-items: center; gap: 10px; line-height: 1.25;
  flex-wrap: wrap; max-width: 94%; text-align: left; pointer-events: none; z-index: 3;
  animation: tk-watch-in .28s cubic-bezier(.2, 1, .3, 1) both;
}
.tk-watch[hidden] { display: none; }
.tk-watch .han { flex: 0 0 auto; white-space: nowrap;
  font-family: var(--han); font-size: var(--dk-read); font-weight: 900; color: var(--dk-gold); }
/* flex WILL squeeze this to its longest word if you let it — at 375px the ask
   sentence rendered one word per line down the whole talisman sheet. Give it a
   basis it can wrap at, and let it drop to its own row rather than compress. */
.tk-watch .say { flex: 1 1 210px; min-width: 0;
  font-size: var(--dk-label); font-weight: 700; color: var(--paper, #f2efe9); }
@keyframes tk-watch-in { from { opacity: 0; transform: translateX(-50%) translateY(-7px); } to { opacity: 1; } }
@media (prefers-reduced-motion: reduce) {
  .tk-watch { animation: none; }
}

/* ------------------------------------------------------- THE INK SHELF
   Ink pots on the left (spend them), the brush rest and 重写 on the right.
   Nothing here states an amount — the pots are a stock, not a stepper.
   墨 is night-side text on a night-side slab: --paper, at reading size. */
.tk-shelf {
  width: 100%; display: flex; align-items: center; gap: 12px; flex-wrap: wrap;
  padding: 8px 12px; justify-content: space-between;
}
.tk-pots { display: flex; align-items: center; gap: 9px; }
.tk-pots-label { display: flex; align-items: baseline; gap: 6px; }
.tk-pots-label .han { font-family: var(--han); font-size: var(--dk-read); font-weight: 900;
  color: var(--paper, #f2efe9); }
/* The gloss says what the stock is FOR, not just what it is called: the judge
   counted seven pots and could not tell whether spending them was good or bad.
   Caption size, because it is a gloss on a readout and must not compete with
   the three button faces beside it. */
.tk-pots-label .en { font-size: var(--dk-caption); font-weight: 700; color: var(--haze, #a49fae); }
.tk-potrow { display: flex; align-items: center; gap: 5px; }
.tk-pot { width: 22px; height: 25px; flex: 0 0 auto; transition: opacity .3s ease, transform .3s ease; }
.tk-pot.spent { opacity: 0.3; transform: translateY(2px) scale(0.9); }
@media (prefers-reduced-motion: reduce) { .tk-pot { transition: none; } }

.tk-tools { display: flex; align-items: center; gap: 8px; flex-wrap: wrap; }
.tk-btn {
  font-size: var(--dk-label); font-weight: 700; font-family: inherit; cursor: pointer;
  display: inline-flex; align-items: center; gap: 5px;
  background: var(--dk-stock-2); color: var(--dk-ink);
  border: 3px solid var(--dk-ink); border-radius: 10px; padding: 6px 13px;
  box-shadow: 0 3px 0 rgba(14, 6, 24, 0.75);
}
.tk-btn .han { font-family: var(--han); font-size: var(--dk-read); font-weight: 900; }
.tk-btn:hover { background: #fff6e0; }
/* 再念一遍 reads as an icon alone — see AGAIN_ICON — so it drops the label. */
.tk-btn-ic { padding: 8px; }
.tk-btn-ic svg { width: 20px; height: 20px; }
.tk-btn:active { transform: translateY(2px); box-shadow: 0 1px 0 rgba(14, 6, 24, 0.75); }
/* SHUT IS THE KIT'S JOB NOW, and this rule used to fight it. drill-kit.css
   paints every shut control in .minigame as TIRED PAPER — filter steps the
   object's own colour down, the ledge flattens, an inset says the front is
   pushed in — and its note is explicit that `opacity` is the one thing never to
   use, because opacity drags the type toward whatever art is behind the rail.
   This drill kept an opacity:0.4 from before that rule existed, and it showed
   the moment 重写 started opening shut on an untouched sheet: the ink label on
   the disabled face measured 2.87:1 against the night slab. Drop the opacity,
   keep the cursor, let the kit carry it — the button still plainly reads shut
   (filter + inset + flattened ledge, all three measured by ui-census) and the
   words on it stay readable, which is what a player needs to know WHICH control
   is the one that is closed. */
.tk-btn:disabled { cursor: default; transform: none; }

/* ---------------------------------------------------------- THE VERDICT
   One line under the paper, and ONLY when there is a verdict — an empty
   scrimmed bar sitting under the shelf for the whole round is a panel that owns
   nothing. It carries the reason, always: a miss that only says "wrong"
   teaches nothing (pedagogy-rubric, dimension 2). */
.tk-verdict {
  width: 100%; min-height: 44px; display: flex; align-items: center; gap: 10px;
  padding: 7px 13px; flex-wrap: wrap;
  animation: tk-verdict-in .3s cubic-bezier(.2, 1, .3, 1) both;
}
.tk-verdict:empty { display: none; }
.tk-verdict .han { font-family: var(--han); font-size: var(--dk-display); font-weight: 900;
  line-height: var(--dk-lh-tight); }
.tk-verdict .why { font-size: var(--dk-label); font-weight: 700; max-width: var(--dk-measure); }
/* the stroke-class sequence — hanzi the player reads, so it keeps the 20px
   floor instead of riding inside the 16px latin sentence */
.tk-verdict .seq { font-family: var(--han); font-size: var(--dk-read); font-weight: 800;
  color: var(--paper, #f2efe9); letter-spacing: 0.02em; white-space: nowrap; }
/* The ONE stroke that missed, marked inside the run. Gold, because gold means
   "the thing to look at now" — never vermillion: every glyph in this run is the
   RIGHT answer, and vermillion in this game means wrong. */
.tk-verdict .seq i { font-style: normal; color: var(--dk-active, #ffc84a);
  text-decoration: underline; text-decoration-thickness: 2px; text-underline-offset: 4px; }
.tk-verdict .gloss { font-size: var(--dk-caption); font-weight: 700; color: var(--haze, #a49fae); }
/* THE CHARACTER ON THIS STRIP IS THE ANSWER, IN BOTH DIRECTIONS — round-6 judge
   finding, and it was the drill's loudest colour lie: on a miss the strip
   printed the target character (土) in --dk-wrong, which tells a player who is
   learning that character that 土 is the wrong answer. It is the RIGHT answer;
   they simply did not write it.
     · sealed  → teal. The character they wrote was correct, so teal is true.
     · missed  → gold. "The thing to look at now", the same token the marked
                 stroke in the order run already carries, so the eye goes
                 character → stroke → order in one colour.
   Right/wrong still reads instantly, because the VERDICT carries it: a rule
   down the leading edge of the strip, teal or vermillion. That is the semantic
   pair used on the thing it actually describes — the judgement — instead of on
   a character the player is supposed to learn. */
.tk-verdict { border-left: 5px solid transparent; }
.tk-verdict.is-right { border-left-color: var(--dk-right); }
.tk-verdict.is-wrong { border-left-color: var(--dk-wrong); }
.tk-verdict.is-right .han { color: var(--dk-right); }
.tk-verdict.is-wrong .han { color: var(--dk-active, #ffc84a); }
@keyframes tk-verdict-in { from { opacity: 0; transform: translateY(-6px); } to { opacity: 1; } }
@media (prefers-reduced-motion: reduce) { .tk-verdict { animation: none; } }

/* ------------------------------------------------------- RIP SAYS IT TWICE
   House rule from playtesting: every direction, warning, correction and notice
   carries English as well as Chinese. The verdict panel has done this since
   round 4; her BUBBLE — where the whole assist ladder lives, including the
   price of a hint and the warning that the shelf is dry — did not. Her hanzi
   leads and the English takes its own line beneath, at --dk-label (16px, the
   readable floor) and NOT --dk-caption: the kit keeps caption size for pinyin
   and glosses that are never load-bearing, and a translated instruction is
   read-to-play by definition. No fifth type size — the sheet already uses
   display / --dk-read / --dk-label / --dk-caption.
   The bubble's stock is pinned to #fdf6e0 in both themes (prom.css), so this
   ink measures the same in each: #4b3a63 on #fdf6e0 is 9.3:1.
   And prom.css sets bubble hanzi to 19px, a hair under the hanzi floor; her
   lines are Chinese the player reads, so they come up to reading size. */
.tk-host .mg-bubble .han { font-size: var(--dk-read); }
.tk-host .mg-bubble .en {
  display: block; margin-top: 4px;
  font-size: var(--dk-label); font-weight: 700; line-height: 1.35;
  letter-spacing: 0.01em; color: #4b3a63;
}

/* ------------------------------------------------------- THE SHELL (甲骨)
   Principle 7, and it now survives the curtain. This drill used to paint five
   dashed slots labelled "sealed tonight" and drop them on teardown, so a week
   of play left no record of anything (round-7 finding: "None of the three
   shells persists … the game has no completion meter").
   The shelf is now the KIT's `.dk-shell` strip (drill-kit.css), backed by
   prom/js/drills/shell2.js: ONE persistent collection of 23 characters shared
   with 拆字课 Radical Forge, exactly as the benchmark's Radical Forge entry
   asks for. Nothing about it is drill-specific any more, so this sheet holds
   no copy of it — a second copy is how one shell becomes two again.
   Two things it also fixed, both in the kit: the sealed slot was ink on
   LANTERN GOLD and stayed lit for the whole session (gold is "your turn, look
   here now", never a permanent record), and the row was a wrapping flex box,
   which a 23-entry collection would have turned into four rows on a phone.

   AND IT NOW SAYS WHAT IT IS — round-12 blind judge, and Lau's ruling: keep it
   prominent and LABEL it. Fifteen judges looked at 甲骨 0/23 and not one could
   say what it counted; the meter that is the game's real completion record read
   as decoration. The caption is the drill's own chrome around the kit strip,
   never a second copy of it, and it lives OUTSIDE .dk-shell on purpose —
   shellStripPaint() rewrites that element's innerHTML on every carve, so a
   caption inside it would vanish the first time a ward sealed. */
.tk-shellcase { width: 100%; display: flex; flex-direction: column; gap: 5px; }
/* Night-side text on the night stage, so it takes the kit's scrim rather than
   sitting on the painting. Hanzi at reading size (it is Chinese the player
   reads), English under it at label size — the house bilingual rule, and this
   is a notice, so it carries both. */
.tk-shellcap {
  margin: 0; padding: 7px 13px 8px; border-radius: 12px;
  background: var(--dk-mat-scrim); box-sizing: border-box;
}
/* NOT capped at --dk-measure. The measure exists to stop prose running to a
   200-character line; this strip is 594px at the widest the court ever gets —
   about 29 hanzi, well inside a comfortable line — and capping it at the latin
   measure cost this two-sentence notice a third and fourth row of type under an
   already tall column. The container is the measure here. */
.tk-shellcap .han {
  display: block; font-family: var(--han); font-size: var(--dk-read); font-weight: 800;
  line-height: var(--dk-lh-read); color: var(--paper, #f2efe9);
}
.tk-shellcap .en {
  display: block; margin-top: 3px;
  font-size: var(--dk-label); font-weight: 700; line-height: 1.35;
  color: var(--haze, #a49fae);
}

/* phones: the host cutout is gone (prom.css hides it under 720px), so the
   court takes the whole width back and the paper stays as large as possible */
@media (max-width: 720px) {
  .tk-game { --tk-lane: 0px; }
  .tk-court { width: 96vw; gap: 8px; margin-top: 4px; }
  .tk-slip { padding: 8px 11px 9px; }
  /* three tools do not fit one phone row beside eight ink pots, and a shrunk
     button face is the one thing the type rules forbid. Give the tools their
     own full-width row and let them share it evenly. */
  .tk-shelf { gap: 8px; }
  .tk-tools { width: 100%; justify-content: space-between; gap: 6px; }
  .tk-btn { padding: 6px 10px; }
  /* the ask banner stacks rather than sharing a row: a 65-character sentence
     beside a five-hanzi question on a 375px sheet is the one-word-per-line
     column that ate the whole talisman in the first pass */
  .tk-watch { flex-direction: column; align-items: flex-start; gap: 2px; max-width: 90%; }
  .tk-watch .say { flex: 0 0 auto; }
}

/* .minigame centres its single column, so once the column is taller than the
   viewport the centring CLIPS THE TOP and the first row lands under the
   absolutely-positioned .mg-head (measured: column top at -15px, header at
   18-60px in light mode). The drill's own root carries this class, so no :has()
   and no edit to prom.css: top-align, and leave room for the header.

   The header's HEIGHT is not knowable from here: it wraps to two rows on a
   phone in dark and to four in light (where prom.css gives it a cream plate
   with its own border and padding), which is how the round counter ended up
   printed on the case slip at 1.10:1. So this is only the floor — talisman2.js
   measures .mg-head and sets the real padding-top (fitHead). */
.minigame.tk-host { justify-content: flex-start; padding-top: 66px; }

/* NO TEXT SITS ON BARE ART. prom.css re-materialises .mg-head as a cream plate
   in LIGHT theme but leaves it bare on the painting in DARK — where on a phone
   the round counter lands over a lit patch of the courtyard and measures
   3.17:1. Give it the kit's scrim. Scoped to this drill's own root, and the
   light-theme plate rule in prom.css still outranks it, so light is untouched. */
.minigame.tk-host > .mg-head {
  background: var(--dk-mat-scrim); border-radius: 12px;
  padding: 6px 13px; box-sizing: border-box;
}

/* TWO NUMBERS READING ZERO, EQUALLY LOUD, IS WHY NOBODY COULD NAME THE GOAL.
   "There are two counters that both sit at zero … I cannot tell which one is the
   actual goal." The goal chip (封印 0/5 sealed) is the answer to "how do I win"
   and keeps its full weight; the round counter is a position, not a target, so
   it steps back behind it. Nothing is hidden — 4.9:1 on the head's scrim, still
   well over the floor — it simply stops arguing. */
.minigame.tk-host .mg-score .tk-roundtag { font-style: normal; opacity: 0.66; }

/* ------------------------------------------- THE WORDS HAVE TO FEEL A POINTER
   `.mg-host` carries an inline `pointer-events: none` (minigames.js stageHTML)
   — correctly, because Rip's cutout is a 44%-tall painting pinned over the
   court and it must never swallow a press meant for the paper, which on THIS
   board is the whole gesture. `pointer-events` inherits, and her bubble is a
   child of that cutout, so every word she says went deaf with her.

   Measured on the cold board: all 15 words of
   「笔在这儿。笔尖在祸事上停一停，偏旁自会浮上来。」 hit-tested to `div.tk-game`,
   never to the span. That line is the board's biggest block of Chinese AND its
   standing instruction — the sentence that tells a beginner where to rest the
   tip — and the two-line deal deleted a printed gloss from the labels on the
   promise that the card would give it back.

   The exemption is one selector wide, matching the shape the kit's other
   drills already use (drill-cloze2.css, drill-haggle2.css, drill-tone2.css):
   the WORD SPANS take the pointer, her paper and her whole cutout stay
   transparent. Verified by hit-test at four points over her art — all four
   still fall through to `div.tk-game`, so the canvas keeps every press.
   `.tk-watch` is deliberately left alone: talisman2.js never wraps it (see the
   note over showWatch), because it sits across the drill's one live gesture.
   Nothing new becomes pressable — a `.w` carries no handler of its own, the
   desktop path is hover, and game.js refuses any `.w` inside a control. */
.minigame.tk-host .mg-bubble .w { pointer-events: auto; }

/* Packet 09: ink and writing controls wrap without squeezing the paper. */
.tk-host #tkGuide { min-height: 44px; font: 600 1rem/1.2 system-ui; padding: 6px 10px; }
.tk-host .tk-pots-label { max-width: none; }
.tk-host .tk-pots-label .en { font-size: 1rem; }

.tk-host .tk-next { display: block; min-height: 44px; margin: 12px auto 0; padding: 8px 12px; color: var(--dk-ink); background: var(--dk-stock); font: 600 1rem/1.3 system-ui; }

/* Counters must wrap inside their own paper, including enlarged text. */
.minigame.tk-host .mg-score { min-width: 0; max-width: 100%; white-space: normal; overflow-wrap: anywhere; }
.minigame.tk-host .mg-score .dk-goal { min-width: 0; max-width: 100%; flex-wrap: wrap; white-space: normal; }
.minigame.tk-host .mg-score .dk-goal-en { white-space: normal; overflow-wrap: anywhere; }
@media (max-width: 430px) {
  body .minigame.tk-host.has-manual > .mg-head { padding: 56px 12px 10px !important; }
}
