/* Vault1 component stylesheet. Hand-authored (not generated - see
 * static/tokens.css for the one file that is).
 *
 * v1/design-system commit 2: this file currently holds exactly the rules
 * base.html's own <style> block used to define (:root, *, body, header,
 * header a, main, .card, .muted, button, .link-button, .error) - moved
 * here verbatim, with only the four hardcoded colours DESIGN_SPEC.md S1
 * calls out (#172033, #f5f7fb, #e3e7ef, #657087) replaced by their token
 * equivalents. Nothing else changed: same selectors, same other property
 * values, same structure - the token swap is the only appearance change
 * in that commit, reviewable on its own before any more templates land.
 *
 * Components are added as each first gets a real consumer, same
 * precedent as _score_scale_control.html - building unused classes ahead
 * of a caller means reviewing CSS with nothing to see it against:
 *   - eyebrow: static/type.css, commit 3
 *   - row, badge (neutral + success only), empty-state: commit 4,
 *     parent_dashboard.html - see each block below for what's still
 *     missing (action/pending/attention badge families) and why.
 *
 * font-family lives in static/type.css, not here - that file owns
 * everything type-related (the self-hosted @font-face rules and the base
 * font-family both need to load together).
 */

:root {
  color: var(--vault-color-ink);
  background: var(--vault-color-page);
}

* {
  box-sizing: border-box;
}

/* Colin's own walkthrough finding, 2026-09-30 (docs/briefs/
 * INSTRUCTION_PILLAR_DESIGN.md §16): "Autosave is rendered as a
 * greyed-out button. Autosave is a state, not a control." Traced, not
 * guessed - confirmed live (Playwright, a bare `<button hidden>` against
 * this stylesheet) that it was never meant to be visible at all:
 * instruction/compose.html's own #autosave-button already carries the
 * `hidden` attribute and does nothing but submit a hidden frame's worth
 * of autosave data; the real, visible UI is #autosave-status, a plain
 * <span>, exactly the "state, not a control" shape this finding asks
 * for. The button rule below (`button, .button--primary,
 * .button--secondary { display: inline-flex; ... }`) is an AUTHOR
 * stylesheet declaration, which overrides the browser's own `[hidden] {
 * display: none }` UA-stylesheet rule outright regardless of
 * specificity - origin beats specificity in the cascade, and
 * user-agent-normal is the lowest origin there is. Confirmed live:
 * `getComputedStyle(button).display` was "inline-flex", not "none",
 * with a real 102x44px box on screen. instruction/_picker.html's own
 * #picker-clear ("Show everything", also `hidden` by default) has the
 * identical latent bug - found while tracing this one, fixed here too
 * rather than left for someone to find the same way. This is the
 * general fix, not a one-off inline style on either button: nothing
 * with the `hidden` attribute should ever be visible, and now nothing
 * else that reaches for it next can be bitten by this again either. */
[hidden] {
  display: none !important;
}

/* Design coherence audit (docs/briefs/DESIGN_COHERENCE_AUDIT.md), part 2
 * step 1: no :focus rule existed anywhere in this file before this one.
 * Nothing removed the browser default either, so focus was never invisible -
 * only unowned: whatever a given browser/OS supplied, unverified against
 * either theme. One rule, using the existing action-blue rather than
 * inventing a colour, reachable everywhere at once. :focus-visible (not
 * :focus) so a mouse click doesn't ring every button - keyboard and other
 * non-pointer focus only, which is the actual accessibility need. */
:focus-visible {
  outline: 2px solid var(--vault-color-action);
  outline-offset: 2px;
}

body {
  margin: 0;
}

/* 2.278 (Colin, 2026-10-29): "no dark top bar", on every page - and that
 * includes the pages with no sidebar, which is where this element is still
 * rendered. The boards draw the same small wordmark there as the
 * sidebar carries, on the page's own background with a hairline under it.
 *
 * What stays is the right-hand side. A page with no sidebar has no sidebar
 * foot either, and Colin's walkthrough of 2026-10-02 is explicit that a way
 * to reach support has to be visible and persistent - so Account, Contact
 * support and Sign out keep their place here rather than disappearing with
 * the dark background. */
header {
  background: var(--vault-color-page);
  color: var(--vault-color-ink);
  border-bottom: 1px solid var(--vault-color-line);
  /* Design coherence audit, part 2 step 3: converged onto the existing
   * spacing scale, not a new value - 18px/28px was this app's own de
   * facto "header gap," found 166/41 times respectively across the
   * templates as the dominant off-token value for "space between
   * sections." 24px (not 32px, though 28 sits equidistant between them)
   * matches `main`'s own default horizontal padding (24px) - the header
   * and the page content below it now share the same left edge on every
   * page without a sidebar, confirmed live, which they did not before. */
  padding: var(--vault-space-16) var(--vault-space-24);
  display: flex;
  justify-content: space-between;
  align-items: center;
}

header a {
  color: var(--vault-color-ink);
  text-decoration: none;
}
/* The wordmark, the same one the sidebar draws, without the sidebar's own
 * block padding. **Jamie's compact lockup since 2026-11-13** (`C111`): the
 * type rules that set the text mark went with it, and what is left is the
 * image's own box. `display: block` on the image, so the anchor is the
 * lockup's height exactly rather than a line box with a descender's gap
 * under it. */
header .header__wordmark img,
.sidebar__wordmark img {
  display: block;
  height: 32px;
  width: auto;
}

/* PARTNER_BRIEF (2026-09-23) item 7: "'Signed in as' and 'Sign out'
 * crowd each other at narrow widths." The <768px rule further down
 * (docs/FINDINGS.md, 2026-09-20) already drops .header__identity
 * entirely on phones - this is the general-width safety net that fix
 * didn't cover: a long full name at a width just above that cutoff (a
 * small tablet, a phone in landscape) had no constraint at all and
 * could still wrap and push into "Sign out". Truncates instead of
 * wrapping, at every width where the element is shown - a full name is
 * never the thing this app needs someone to read in the header; /account
 * (still linked, right next to it) is. */
.header__identity {
  white-space: nowrap;
  overflow: hidden;
  text-overflow: ellipsis;
  max-width: 220px;
}

main {
  max-width: 1100px;
  margin: 0 auto;
  padding: 40px 24px;
}

/* Two-column app shell (S9's "first real layout work", 5.12). Only
 * `body.has-sidebar` is affected - every other page keeps the plain
 * `header` then `main` stack above untouched, since the bare `body`
 * selector is never changed. Grid, not flex, so `header` can span both
 * columns without the flex-basis:100% wrapping trick, and so `.sidebar`
 * and `main` stretch to match each other's height for free (grid's
 * default align-items) without a fragile `calc(100vh - Npx)`. */
body.has-sidebar {
  display: grid;
  /* 300px, from the console redesign's commit 2 (CONSOLE_REDESIGN_SPEC.md
   * §2, which took it from 220px).
   *
   * **The boards draw 216px** - 20 of them, across the director, c99,
   * c100d1 and c100d2 canvases, all `flex:0 0 216px` - and 2.283 says to
   * take every value from the boards. Tried, measured and reverted, so
   * nobody spends the afternoon repeating it: `scripts/compare-board`
   * scored the director console 0.199 at 300px and 0.198 at 216px, and at
   * 216 the sidebar crowds - "Guardians" and its "2 new" no longer fit on
   * one line and the count wraps underneath. The boards' sidebar carries
   * no icons; this one does (the §4 console redesign), and that is the
   * 84px. Which of the two gives way is a layout decision neither one
   * covers; raised in docs/FINDINGS.md rather than settled here. */
  grid-template-columns: 300px 1fr;
}
body.has-sidebar > header {
  grid-column: 1 / -1;
}
body.has-sidebar > .sidebar {
  grid-column: 1;
  display: flex;
  flex-direction: column;
  border-right: 1px solid var(--vault-color-line);
}
/* **2.319 (j)** (Colin, 2026-11-13, Jamie's second review): Contact support
 * and Sign out stay in view however long the page.
 *
 * They were already pinned *inside* the sidebar - `.sidebar .nav` scrolls and
 * `.sidebar__foot` sits under it - but the sidebar is a grid item that
 * stretches to the content beside it, so on a 4,000px roster the foot sat
 * 4,000px down. Sticking the column itself to the viewport is what makes the
 * foot's own pinning mean anything: the nav scrolls within one screen, and the
 * two links are always on it.
 *
 * `align-self: start` because a stretched grid item cannot be sticky - without
 * it the item is already as tall as the row and never has anywhere to stick.
 * `100dvh` after `100vh` for the phone browsers whose toolbars make the two
 * differ; desktop only, since below 768px the sidebar is a disclosure above
 * the content and pinning it would eat the screen. */
@media (min-width: 768px) {
  body.has-sidebar > .sidebar {
    position: sticky;
    top: 0;
    align-self: start;
    height: 100vh;
    height: 100dvh;
  }
}
/* Console redesign commit 2: logo, org line, the nav list (its own
 * scrolling flex column, unchanged gap/padding from before this round,
 * just moved from .sidebar itself onto this wrapper) and the footer are
 * four stacked sections, each owning its own padding - matching the
 * reference files' own CSS exactly rather than one padding value spread
 * across every child, which is what let the footer and the logo fight
 * over the same spacing before either existed. */
/* 2.278 (Colin, 2026-10-29): the wordmark the boards draw, one line, in
 * place of the 300x200 raster logo that took the top sixth of every staff
 * page. .sidebar__logo stays defined because nothing says every template
 * stopped using it in the same commit; it is the rule that is now unused,
 * not a rule anything still needs. */
.sidebar__logo {
  padding: 16px 16px 0;
}
.sidebar__logo img {
  display: block;
  width: 100%;
  max-width: 300px;
  height: auto;
}
.sidebar__wordmark {
  display: block;
  padding: 16px 16px 6px;
  text-decoration: none;
}
.sidebar__foot-links {
  display: flex;
  flex-wrap: wrap;
  gap: 4px 12px;
  align-items: baseline;
}
/* B1 (2.307 (7), Colin 2026-11-05, docs/boards/staff/B1-Account-block.dc.html):
 * the account block at the top of the sidebar - name, role and club in one
 * tinted block under the wordmark, the whole block being the way to Account.
 *
 * It replaces `.sidebar__org`, the plain line that sat here saying "Director
 * · SouthParkland Lacrosse", and takes the avatar and name off `.sidebar__foot`
 * as well: both said who the reader is, in two places. The tint is the same
 * action-tint fill `.card--tint-action` uses, bordered in action-chip-strong
 * so the block has an edge against a white sidebar - the board's own
 * `#F3F7FF` on `#D8E6FB`, from the two tokens nearest it. */
.sidebar__account {
  display: flex;
  align-items: center;
  gap: 11px;
  margin: 12px 12px 0;
  padding: 10px 12px;
  border-radius: var(--vault-radius-panel);
  border: 1px solid var(--vault-color-action-chip-strong);
  background: var(--vault-color-action-tint);
  text-decoration: none;
  color: inherit;
}
.sidebar__account-who {
  flex: 1 1 auto;
  min-width: 0;
}
/* Both lines ellipsis rather than wrap: the sidebar is 300px and a long club
 * name wrapping to three lines pushes every nav row down, which is the same
 * defect 2.278 removed the raster logo for. */
.sidebar__account-name,
.sidebar__account-role {
  display: block;
  white-space: nowrap;
  overflow: hidden;
  text-overflow: ellipsis;
}
.sidebar__account-name {
  font-size: 14px;
  font-weight: 600;
  color: var(--vault-color-ink);
}
.sidebar__account-role {
  font-size: 13px;
  color: var(--vault-color-ink-muted);
}
.sidebar__account-go {
  flex: 0 0 auto;
  color: var(--vault-color-action);
}
.sidebar .nav {
  flex: 1;
  min-height: 0;
  overflow-y: auto;
  padding: 10px 12px;
  display: flex;
  flex-direction: column;
  gap: 2px;
}
.sidebar__foot {
  border-top: 1px solid var(--vault-color-line);
  padding: 14px 16px;
  display: flex;
  align-items: center;
  gap: 12px;
}
.avatar {
  width: 38px;
  height: 38px;
  border-radius: var(--vault-radius-pill);
  background: var(--vault-color-ink);
  color: var(--vault-color-surface);
  display: flex;
  align-items: center;
  justify-content: center;
  font-size: 13px;
  font-weight: 700;
  flex: none;
}

/* The pale form of the same disc (2.305 (3), docs/boards/c98/Main.dc.html's
 * "Who this covers"): a tint fill with the initials in the action colour,
 * 34px rather than 38. A modifier, not a second avatar: the shape, the
 * centring and the weight are the same component, and the board's own
 * difference from the sidebar's dark disc is the fill and the size. */
.avatar--soft {
  width: 34px;
  height: 34px;
  background: var(--vault-color-action-tint);
  color: var(--vault-color-action);
}
/* B1's own avatar: solid action blue, white initials, 36px - the opposite
 * pair from `.avatar--soft` above, and not the default ink circle, because
 * this one sits inside the action-tinted block and the board draws it in the
 * same family as the block it is in. */
.avatar--action {
  width: 36px;
  height: 36px;
  background: var(--vault-color-action);
  color: var(--vault-color-surface);
}
.sidebar__foot-link {
  display: block;
  margin-top: 2px;
  font-size: 13px;
  font-weight: 600;
  color: var(--vault-color-action);
}
body.has-sidebar > main {
  grid-column: 2;
  /* 960px, not the sidebar-less 1100px: enough room for a card or a row
   * of stat tiles without letting roster/attention rows run to whatever
   * the viewport happens to be - a max-width is what "page padding"
   * actually means once the plain `main{margin:0 auto}` centering (which
   * supplied the visual breathing room before) no longer applies with a
   * sidebar to its left. Left-aligned (margin:0), not centered - content
   * sits against the sidebar the way a dashboard reads, not centered in
   * whatever's left over the way a marketing page would. Horizontal
   * padding uses the existing 32px spacing token (up from the plain
   * `main` rule's 24px) so the column doesn't start flush against the
   * sidebar's border; vertical padding matches every other page's 40px. */
  /* 1020px, measured (2.294): at the 1320 canvas every board is drawn on,
   * a 300px sidebar (2.284 (3), "wrap is bad news") leaves exactly 1020 for
   * this column, and 960 was leaving 60 of them on the floor. The cap's own
   * reason below is unchanged - it still stops a roster running to whatever
   * a 2560 viewport happens to be - it just no longer bites inside the
   * canvas the pages were drawn for. Content lands at 956 (1020 less the
   * 32px gutters), where it was 896.
   *
   * **What this cannot fix, and nobody should read it as fixing**: the
   * boards that draw chrome give the sidebar 217px and the content 1103
   * (support/Main, measured). 1103 + 300 is wider than 1320, so the boards'
   * proportions are unreachable with the sidebar Colin picked, and the
   * boards that draw no chrome at all (evaluations/Main, coaching/B-Item,
   * surveys/Results-A, all 1256) are not comparing like with like. The gap
   * that is left after this is the sidebar and the missing chrome, not the
   * cap. */
  max-width: 1020px;
  margin: 0;
  padding: 40px var(--vault-space-32);
  min-width: 0;
}
.sidebar__divider {
  height: 1px;
  background: var(--vault-color-line);
  margin: 12px 8px;
}
/* Nav-list item (S5, new; recoloured and resized by the console redesign,
 * commit 2). Active state used to reuse `action`/`action-tint` - the
 * spec's own active-row values (#EAF2FF bg, #073B98 ink) are close to
 * but not equal to either token (action-tint is #F3F7FF, action-chip is
 * #E8F0FE), so they stay literal here rather than aliasing to a token
 * that would visibly drift from the approved value - single consumer,
 * same reasoning components.css already uses for the tile/badge
 * borders. Idle text (#10223C) is the same case, close to but not equal
 * to `ink`. min-height 44px is the spec's own row-height floor.
 *
 * flex-wrap: wrap kept (not removed with the sidebar's own hint text,
 * 2026-10-15): parent_player_detail.html reuses .nav-item/.nav-item__hint
 * as its own compact link list, independent of base.html's sidebar and
 * NAV_HINTS - its own comment says so explicitly ("reusing .nav-item/
 * .nav-item__hint as-is"). Removing either would break that page's hint
 * line wrapping, a page this round's dispatch did not name. The sidebar
 * itself needs no CSS change at all for "icon plus label, no hint" - it
 * simply stops emitting a .nav-item__hint span (base.html), and a
 * .nav-item with no hint child never wraps regardless of this rule. */
/* A nav group's section heading (punchlist 2.146). Type comes from
 * .eyebrow (type.css) - the same small, faint, uppercase label the
 * product already uses above a page's title, so a group heading and a
 * page eyebrow read as the same kind of thing rather than two
 * inventions. Only the box is here: the space that makes a heading read
 * as separating two groups, which is what it replaced a 1px divider to
 * do. No background and no border - "plain section headings".
 *
 * Padding matches .nav-item's own 12px so the heading's first letter
 * lines up with the labels under it, not with their icons. */
.nav-section {
  margin: 16px 0 4px;
  padding: 0 12px;
}
.nav-section:first-child {
  margin-top: 0;
}

.nav-item {
  display: flex;
  flex-wrap: wrap;
  align-items: center;
  gap: 2px 8px;
  /* 44px is the phone's touch target and stays; the boards draw a desktop
     sidebar, where their own 9px/12px padding gives about 36px. Keeping
     the minimum means a row is never smaller than a thumb, and the board's
     radius and size below are what the comparison was actually scoring. */
  min-height: 44px;
  padding: 0 12px;
  /* 8px, the boards' own (2.283) - 12px read as a pill at this height. */
  border-radius: 8px;
  text-decoration: none;
  color: #10223C;
  font-weight: 600;
  font-size: 13.5px;
}
/* Guardian "My Children" collapse (punchlist 2.29). A <details> whose
 * <summary> is itself a .nav-item, so the collapsed row is the same
 * height, radius and weight as every other nav row and the sidebar does
 * not acquire a second kind of entry.
 *
 * The marker triangle is removed in both spellings browsers use -
 * ::-webkit-details-marker for WebKit and `list-style: none` on the
 * summary for Firefox - because the row already ends in an icon and a
 * count, and a third glyph competing on the same line is the "two
 * competing numbers" shape one level over. Keyboard and screen-reader
 * behaviour is <details>'s own and nothing here replaces it; there is
 * no JS for this at all, which is why it works before CSS loads and
 * with JS off.
 *
 * Children indent by the icon's own width plus its gap (20 + 8), so a
 * child's label lines up with every top-level label rather than with
 * its parent's icon - the nesting reads from alignment, not from a
 * second border. */
.nav-group {
  display: contents;
}
.nav-group__summary {
  cursor: pointer;
  list-style: none;
}
.nav-group__summary::-webkit-details-marker {
  display: none;
}
.nav-group__child {
  padding-left: 40px;
  font-weight: 500;
}
.nav-item__icon {
  display: flex;
  flex: none;
  color: inherit;
  /* Last in the DOM (base.html's own comment on why), first visually -
   * every existing "class=\"nav-item...\">LABEL" test assertion across
   * the suite stays literally true with the icon appended after, not
   * inserted before. */
  order: -1;
}
.nav-item__icon svg {
  display: block;
}
/* PARTNER_BRIEF (2026-09-23) item 7: "short explanations for the main
 * navigation items, reachable on a phone and by keyboard - not
 * hover-only" - always-visible text, not a tooltip, so it's reachable on
 * every input method by construction. flex-basis:100% forces it onto its
 * own line under the label+count row without any extra wrapper markup.
 * No longer used by base.html's own sidebar (Colin, 2026-10-15 - the
 * hint line repeated the label once rows had icons and a consolidated
 * menu), but kept: parent_player_detail.html's own compact link list
 * still reuses this class for its own, unrelated hint text (see
 * .nav-item's own comment above). */
.nav-item__hint {
  flex-basis: 100%;
  font-size: 11.5px;
  font-weight: 400;
  color: var(--vault-color-ink-faint);
  line-height: 1.3;
}
/* .nav-item--active .nav-item__hint was removed here 2026-10-18 (Colin's
 * own instruction, the C73 descendant-rule measurement): measured, not
 * grepped - .nav-item is used by exactly two templates; base.html no
 * longer emits .nav-item__hint at all (the comment above already said
 * so), and parent_player_detail.html emits it only on .nav-item--compact,
 * never --active. The two classes never co-occurred in any real page. */
.nav-item--active {
  background: #EAF2FF;
  color: #073B98;
  font-weight: 600;
}
.nav-item__count {
  margin-left: auto;
  font-size: 12px;
  font-weight: 600;
  /* 2.242a (Colin, 2026-10-28): orange, as every console board draws it,
   * not grey. A badge on this nav counts the rows in that pillar's
   * "Needs you" and nothing else (2.242's approved behaviour), so it is
   * always work waiting on the reader - which is what `attention` means.
   * Grey said "here is a number"; orange says "this is yours".
   *
   * One class, so this reaches every sidebar that carries a badge, not
   * only the coach's. That is deliberate and not a side effect: the
   * badge means the same thing on each of them. */
  color: var(--vault-color-attention);
  background: var(--vault-color-attention-tint);
  border-radius: var(--vault-radius-pill);
  padding: 1px 7px;
}
/* 2026-10-09, the guardian-nav "something new" signal (docs/FINDINGS.md,
 * "a guardian's Needs Attention sentence names no child on a phone"):
 * deliberately a dot, not a count - .nav-item__count already owns the
 * right-aligned "number she can act on" position, and a second number
 * next to it is exactly the competing-badges shape the attention-sentence
 * fix also avoided. Sits inline right after the label instead. */
.nav-item__dot {
  display: inline-block;
  width: 6px;
  height: 6px;
  border-radius: 50%;
  background: var(--vault-color-action);
}
/* Screen-reader-only text: visually hidden, still in the accessibility
 * tree - the standard clip-based pattern, not display:none (which removes
 * it from both). First caller: .nav-item__dot's own text equivalent,
 * since a bare dot says nothing to anyone not looking at it. */
.u-sr-only {
  position: absolute;
  width: 1px;
  height: 1px;
  padding: 0;
  margin: -1px;
  overflow: hidden;
  clip: rect(0, 0, 0, 0);
  white-space: nowrap;
  border: 0;
}
.nav-item--active .nav-item__count {
  color: #073B98;
  background: var(--vault-color-surface);
}

/* Below 768px: "not broken" is the bar (S7 - the staff console is a
 * desktop surface, mobile is out of scope). The sidebar used to become
 * the pre-5.12 top link-row (wrapped, always open) - correct for "not
 * clipped, no horizontal scroll" but not for a phone specifically: a
 * director's ~10-item nav wraps to roughly 270px before any page
 * content starts, pushing every staff page below the fold on a real
 * phone (confirmed with a Playwright render at the design handoff's own
 * 390x844 mobile reference). Collapsed-behind-a-toggle by default now -
 * survivable, still not a redesign: no drawer, no icon rail, no bottom
 * tab bar, same .nav-item markup and desktop behaviour untouched. */
.sidebar-toggle {
  display: none;
}
@media (max-width: 767px) {
  body.has-sidebar {
    grid-template-columns: 1fr;
  }
  body.has-sidebar > .sidebar {
    grid-column: 1;
    flex-direction: column;
    border-right: none;
    border-bottom: 1px solid var(--vault-color-line);
  }
  body.has-sidebar > .sidebar .sidebar__divider {
    display: none;
  }
  .sidebar-toggle {
    display: block;
    width: 100%;
    min-height: 44px;
    padding: 12px;
    border: none;
    background: none;
    text-align: left;
    font: inherit;
    font-weight: 600;
    color: var(--vault-color-ink);
    cursor: pointer;
  }
  /* Collapsed by default - only the toggle shows until it's clicked
   * (sidebar-toggle.js adds this class, no state on initial load).
   * Console redesign commit 2 added three more direct children of
   * .sidebar above the nav list (.sidebar__logo, and what is now
   * .sidebar__account - B1, 2.307 (7)) and one
   * below it (.sidebar__foot) - collapsed still means collapsed, not
   * "the nav rows hide but the 200px-tall logo doesn't," so all three
   * join the nav-item rule below rather than only updating its own now-
   * one-level-deeper selector (.sidebar > .nav > .nav-item, since nav
   * rows moved inside their own .nav wrapper this same round). */
  body.has-sidebar > .sidebar:not(.sidebar--expanded) > .sidebar__logo,
  /* B1's account block joins this list: the rule enumerates its children
   * rather than hiding everything in the sidebar, so a block added at the
   * top stays visible above a collapsed menu until it is named here - which
   * is exactly the hole 2.146 found for the section headings. */
  body.has-sidebar > .sidebar:not(.sidebar--expanded) > .sidebar__account,
  body.has-sidebar > .sidebar:not(.sidebar--expanded) > .nav > .nav-item,
  body.has-sidebar > .sidebar:not(.sidebar--expanded) > .nav > .nav-section,
  body.has-sidebar > .sidebar:not(.sidebar--expanded) > .nav > .nav-group,
  body.has-sidebar > .sidebar:not(.sidebar--expanded) > .sidebar__foot {
    display: none;
  }
  /* 2.295, measured against c100d1/Phone-Accept.dc.html: the 16px gutter
   * below was given to `body.has-sidebar > main` only, so every page a
   * reader opens *before* they have a sign-in - the athlete accept page,
   * login, a claim, a reset - fell through to the plain `main` rule's
   * 24px. Those are web pages at phone width (§7), and §9's gutter is
   * 16px for all of them; the board puts its cards at x=16 and the built
   * accept page had them at 24. Same value, same reasoning as the rule
   * below, applied to the case that has no sidebar to key off. */
  main {
    padding: 24px var(--vault-space-16);
  }
  body.has-sidebar > main {
    grid-column: 1;
    /* max-width/margin from the desktop rule carry over harmlessly (the
     * viewport is already narrower than 960px), but padding does not -
     * commit 23's 32px/40px bump was sized for the desktop cramped-
     * against-the-sidebar problem and was never re-checked against a
     * 360px phone, where it would otherwise take ~18% of the width
     * before any content starts. Back down to the 16/24px tokens here,
     * mobile-appropriate on their own terms rather than inherited. */
    padding: 24px var(--vault-space-16);
  }
  /* Phone-first walkthrough (docs/FINDINGS.md, 2026-09-20): header never
   * had a mobile rule at all, unlike .sidebar/main above - "Signed in as
   * {full name}" wrapped to two lines crowding directly against the
   * "VAULT" wordmark, and "Sign out" wrapped to "Sign"/"out", on every
   * single authenticated page. header itself is nowrap by design (VAULT
   * and the identity group stay on one row rather than the whole header
   * growing an extra row) - the actual fix is dropping the one variable-
   * length, non-essential piece of content: the guardian or athlete
   * already knows who they are, and /account (still shown) is where
   * identity is actually managed. Everything remaining ("VAULT",
   * "Account", "Sign out") is short and fixed-length enough to fit one
   * row at 390px without wrapping. Desktop is completely unchanged - this
   * rule only exists below 768px. */
  header {
    padding: 14px var(--vault-space-16);
  }
  .header__identity {
    display: none;
  }
  /* Phone-first walkthrough (docs/FINDINGS.md, 2026-09-20) Finding 4: the
   * real "becomes cards at 768px" half of .table (S5's own original
   * definition) - built here as an opt-in (then named .table--responsive), applied
   * to nine tables one at a time over the next several days as someone
   * happened to phone-test that specific page (_timeline_groups.html,
   * then seasons.html, then three more on 2026-09-25's own phone-first
   * pass). Sixteen more <table class="table"> instances across twelve
   * templates were still exposed to the identical STATUS-and-actions-
   * scrolled-off-screen shape the whole time, discoverable only the same
   * way each of the nine already-fixed ones was - surveyed and reported
   * 2026-09-25 (docs/FINDINGS.md), inverted here per Colin's agreement:
   * every <table class="table"> gets this treatment by default now.
   * The .table--responsive class this rule used to key on is gone - the
   * guard that found it (2026-10-16, tests/test_template_class_
   * references.py) found it still sitting in 26 templates' own class
   * attributes weeks after the behavior it once opted into became the
   * default, doing nothing - removed from all 26, not kept as a
   * harmless-looking no-op a future reader could mistake for live.
   * .table--dense is the explicit opt-out, for a table with a stated
   * reason to stay tabular below 768px - none exists today. colspan/
   * rowspan swept across every template before inverting (one hit):
   * outbound_mail.html's two-row-per-message structure (a header row,
   * then a colspan="4" row holding the message body) looked like a
   * candidate for the opt-out at first read, on the theory that each
   * row becoming its own separately-bordered card would lose the visual
   * pairing between them - tried .table--dense there, then rendered
   * both options side by side rather than trusting the theory: the
   * opt-out only brought back the exact hidden-Subject-column bug this
   * whole change exists to fix, and the two-card grouping reads clearly
   * in practice (the body card directly follows its header card, no
   * label supplied for its own colspan cell, no ambiguity live). Kept
   * responsive; data-label added there instead (docs/FINDINGS.md has
   * the full comparison). Each <td> becomes its own labelled block instead
   * of a table cell, data-label supplying the column header the
   * collapsed <thead> no longer shows - absent on the sixteen newly-
   * covered tables today, which still stack correctly (nothing hidden)
   * but without a label above each value until data-label is added to
   * them too, a real but non-blocking follow-up, not silently claimed
   * as done here. */
  /* The phone card form is gone (2.298): it turned every row into a
   * bordered card with a label per cell, which is the stack of cards §9
   * rejects - "a table at 390px is worse than the cards it replaced" was
   * about the *desktop* table surviving, not about bringing the cards
   * back. Every table now takes the dense row form far below. */
}

/* Walkthrough, 2026-10-21 (Colin): "one value can't serve both - 20px
   is 18.9% of a 106px tile and 5.8% of a 342px card." --vault-radius-
   card stays 20px (unchanged value, mobile's own Card component shares
   it for tile corners - ConsoleBuckets.tsx's own comment: "radius.card
   is not among them: the tile's corner comes from Card" - so changing
   its value here would shrink mobile's tiles with no mobile-side
   change to catch it). --vault-radius-panel is the new, second token
   for this group instead: cards, rows, dialogs and panels, 12px. */
.card {
  /* 2.283: the boards draw a flat card - no shadow, 12px radius, 16px 18px.
   * The shadow was the one thing on every page that no board has. */
  background: var(--vault-color-surface);
  border: 1px solid var(--vault-color-line);
  border-radius: var(--vault-radius-panel);
  padding: 16px 18px;
}

/* Colin's own walkthrough finding, 2026-09-30 (docs/briefs/
 * INSTRUCTION_PILLAR_DESIGN.md §16): "'Use' appears to do nothing
 * because it jumps silently to a form far below." The reload was
 * always real; what a coach clicking "Use" needs is confirmation it
 * worked, not just an #message URL fragment quietly scrolling them
 * there (instruction/compose.html's own prefill condition sets this
 * class). A brief border/shadow flash, no JS - respects
 * prefers-reduced-motion by simply not animating, rather than a
 * reduced-but-still-present version, since the flash carries no
 * information a reduced-motion viewer needs beyond the scroll itself
 * already landing them in the right place. */
@keyframes card-highlight-flash {
  0% { border-color: var(--vault-color-action); box-shadow: 0 0 0 3px var(--vault-color-action-tint); }
  100% { border-color: var(--vault-color-line); box-shadow: 0 2px 8px rgba(20, 30, 50, .04); }
}
@media (prefers-reduced-motion: no-preference) {
  .card--highlight {
    animation: card-highlight-flash 1.6s ease-out;
  }
}

/* C64's second caller (scope_grants.html, the "which highlights the new
 * row" requirement) applies .card--highlight to a <tr>, not a <div>, and
 * border-color/box-shadow do not paint on a table-row display box in any
 * engine this was checked in - confirmed by rendering the page (CONF-43):
 * the animation ran (animationName was set) but getComputedStyle reported
 * `border: 0px none` and `boxShadow: none` throughout, and the screenshot
 * showed no visible change at all. background-color does paint on <tr>
 * (the everyday zebra-striping property), so this variant reuses the same
 * class, trigger and timing on a property that actually renders there,
 * rather than a border/shadow flash that silently did nothing. */
@keyframes row-highlight-flash {
  0% { background-color: var(--vault-color-action-tint); }
  100% { background-color: transparent; }
}
@media (prefers-reduced-motion: no-preference) {
  tr.card--highlight {
    animation: row-highlight-flash 1.6s ease-out;
  }
}


/* Design coherence audit, part 2 step 2: the type scale has been fully
 * specified in tokens.css since it landed - seven roles, family/weight/
 * mobile-and-desktop size/line-height each - and only two of seven ever
 * got a consumer (body, via :root{font-family}, and eyebrow). type.css's
 * own commit-3 comment named the reason plainly: "each gets added when
 * the template pass reaches the page whose heading needs it" - a pass
 * that never happened, leaving 77 templates writing a bare <h1> with only
 * an inline margin reset and no font-size at all, and every "card
 * heading"/caption sized ad hoc per template instead.
 *
 * Bound to the element, not shipped as three more utility classes nobody
 * would remember to add to the next new page: this is the same shape
 * body's own role already uses to reach everywhere for free. Every
 * existing `<h1 style="margin:0">` keeps working unchanged - that inline
 * style only ever touched margin, never size, so this is additive; the
 * explicit margin:0 below only changes headings that had no inline
 * override, matching how this app already handles spacing everywhere
 * else (explicit, via a wrapping div/gap - never a relied-on browser
 * default). h4-h6 are not bound - unused anywhere in the templates
 * (confirmed by grep), so there is no fourth role to give them yet. */
h1 {
  font-family: var(--vault-type-page-title-family);
  font-weight: var(--vault-type-page-title-weight);
  font-size: var(--vault-type-page-title-size-mobile);
  line-height: var(--vault-type-page-title-line-height);
  margin: 0;
}
h2 {
  font-family: var(--vault-type-section-heading-family);
  font-weight: var(--vault-type-section-heading-weight);
  font-size: var(--vault-type-section-heading-size-mobile);
  line-height: var(--vault-type-section-heading-line-height);
  margin: 0;
}
h3 {
  font-family: var(--vault-type-card-title-family);
  font-weight: var(--vault-type-card-title-weight);
  font-size: var(--vault-type-card-title-size-mobile);
  line-height: var(--vault-type-card-title-line-height);
  margin: 0;
}
@media (min-width: 768px) {
  h1 { font-size: var(--vault-type-page-title-size-desktop); }
  h2 { font-size: var(--vault-type-section-heading-size-desktop); }
  h3 { font-size: var(--vault-type-card-title-size-desktop); }
}

/* `.secondary` (12.75px, the caption/meta-text role) was added here by
 * this same pass as "a real class - it had zero consumers before this" -
 * without checking that type.css already owned every sibling in this
 * family (page-title, section-heading, card-title, button-label) and
 * had already built this one too. Removed 2026-10-19 (Colin's own
 * instruction, the self-review this found): collapsed to type.css's
 * copy, the canonical one - byte-identical to this one in every
 * property including its own desktop media variant, so this was a
 * live duplicate, not yet a drifted one. `button-label`'s own weight
 * (600, already pulled forward as a literal when .button--primary/
 * secondary were first built) is unrelated to this class and stays
 * wherever the button rules below document it. */

/* Design coherence audit, part 2 step 2: no input/select/textarea rule
 * existed anywhere in this file - 97 form controls across the templates
 * each hand-repeated `border:1px solid #cfd6e4;border-radius:Xpx;
 * padding:Ypx;font:inherit` inline, with #cfd6e4 (32 raw occurrences) as
 * the de facto border colour - never a declared token, and a visibly
 * different, cooler grey than the real one (`line`, #E3EAF3). border-
 * radius drifted 8px (dominant) against a handful of 6px call sites for
 * the identical concept; padding drifted 9px (dominant) against 10px and
 * a few one-off values. This rule is the dominant values, tied to the
 * real token. Scoped to the text-like input types actually in use
 * (confirmed by grep) - never checkbox/radio (native, small, a different
 * shape entirely), hidden, submit or button (those are buttons, styled by
 * the button rule instead, not a text field one). Width is deliberately
 * not set here - every caller's own layout (a full-width stacked field vs.
 * an inline search box beside a button) decides that, not this rule. */
input[type="text"], input[type="email"], input[type="password"], input[type="number"],
input[type="date"], input[type="tel"], input[type="search"], input:not([type]),
select, textarea {
  border: 1px solid var(--vault-color-line-strong);
  border-radius: 8px;
  padding: 9px;
  font: inherit;
}

.muted {
  color: var(--vault-color-ink-soft);
}

/* 5.13: .button--primary/.button--secondary's own rules (below) only ever
   set colour/weight, relying on this block for shape - fine while every
   caller was a real <button>, but organization_survey_detail.html's
   header and organization_surveys.html's "+ New survey" now use the same
   classes on an <a> (a link that reads as an action, not a form submit).
   Matched on class as well as element, once, here - not two copies of
   the geometry inline on each <a>, which would drift the next time this
   block changes. */
/* Phone-first walkthrough (docs/FINDINGS.md, 2026-09-20) Finding 5: "Sign
 * out" (.link-button, padding overridden to 0 below) measured 49x38px live
 * at 390x844, and "Submit survey" (.button--primary) measured 129x37px -
 * both short of the ~44px tap-target minimum. min-height here reaches every
 * button/.button--primary/.button--secondary/.link-button caller from one
 * place, same reasoning as this block's own comment above; inline-flex
 * centering keeps label text centered now that height can exceed the
 * padding-driven box it used to. */
button, .button--primary, .button--secondary, .button--danger, .button--danger-outline {
  /* 2.283: the board's pill - docs/boards/director/Main.dc.html draws every
   * action as border-radius:999px, padding:11px 20px, 14px/600. */
  border: 0;
  border-radius: var(--vault-radius-pill);
  padding: 11px 20px;
  font-size: 14px;
  font-weight: 600;
  min-height: 44px;
  cursor: pointer;
  font: inherit;
  display: inline-flex;
  align-items: center;
  justify-content: center;
  text-decoration: none;
}

/* `inherit`, not white (2.283): this was born in the dark top bar, where
 * white was right, and 2.278 then took the dark bar away - leaving a
 * default that is invisible everywhere it is now used. Inheriting means a
 * link-button is the colour of whatever it sits in, which is what the two
 * surfaces that carry one (the header, the sidebar foot) both want. */
.link-button {
  background: transparent;
  color: inherit;
  padding: 0;
}
/* 2.278 moved Sign out into the sidebar foot, where the surface colour it
 * inherits from .link-button is white on white - invisible, and found in
 * this page's own picture rather than by reading. In the foot it is a link
 * like the two beside it. */
.sidebar__foot .link-button {
  color: var(--vault-color-action);
  min-height: 0;
  font-size: 13px;
  font-weight: 600;
}
/* The third surface, and the one that needs saying out loud: a
 * link-button standing beside a filled button, where inheriting the
 * card's own ink makes it read as a label rather than a control (§9: "a
 * control that disappears into the background is not a style, it is a
 * defect"). docs/boards/c98/Main.dc.html draws exactly that pair - "I
 * agree" filled, "I can't agree to these" as a muted underlined link -
 * so this is the board's own treatment, named. */
.link-button--inline {
  color: var(--vault-color-ink-muted);
  text-decoration: underline;
  font-size: 14px;
}

/* password-toggle.js's own markup contract: a flex wrapper around an
 * existing password <input> plus a sibling toggle button, added without
 * touching the input's own inline style attribute (several callers carry
 * an inline width/padding/box-sizing pattern this does not disturb -
 * flexbox renegotiates width through the input's own flex-basis rather
 * than needing a fixed width here). min-width 44px is this file's own
 * phone-first tap-target minimum (see .score-scale-option above); the
 * button's height is left unset so it stretches to match whichever
 * input it sits beside, since callers use at least two different
 * padding values for the input itself. */
.password-field {
  display: flex;
  align-items: stretch;
  gap: 6px;
}
.password-field > input {
  flex: 1 1 auto;
  min-width: 0;
}
.password-field__toggle {
  flex: none;
  min-width: 44px;
  padding: 0 10px;
  background: transparent;
  border: 1px solid var(--vault-color-line);
  border-radius: 6px;
  cursor: pointer;
  font-size: 0.85em;
  color: inherit;
}

/* Phone-first pass (docs/FINDINGS.md, 2026-09-25): _score_scale_control.html's
 * short-range (span <= 10) SCORE option was a bare <input type="radio"> in a
 * label sized to its content only - measured 13x13px (the input itself) and
 * 21x37px (the label, its real click target since the input is nested
 * inside it) live at 390x844, well under this same file's own 44px minimum
 * for every button/link. Scoring a player means repeatedly tapping the
 * smallest interactive element on the page, immediately beside four others
 * just like it. min-width/min-height here enlarge the label's own click
 * target (a native label wraps its input - clicking anywhere inside it
 * toggles the radio, no `for` attribute needed) without changing the visual
 * size of the circle itself, the same "hit target bigger than the glyph"
 * shape browsers use for native checkboxes on touch platforms. Matches this
 * macro's own existing assumption that each option occupies 44px - see its
 * `scale_span * 44` max-width on the min/max label row just below it. */
.score-scale-option {
  min-width: 44px;
  min-height: 44px;
  justify-content: center;
}

/* Design coherence audit, part 2 step 4 (per-template polish, family-
 * facing first): the same shape as .score-scale-option above, generalised
 * - a native <input type="checkbox"> in a bare label renders at ~13-16px,
 * its real click target since the input is nested inside it, well under
 * this file's own 44px minimum. The audit found nineteen such instances;
 * the only family-facing ones are my_survey_detail.html's two multi-
 * select/nomination option lists (a parent or player answering a real
 * survey) - fixed first, per "family-facing first." The other
 * seventeen are staff consoles (guardian-claims, scope-grants, action-
 * library, survey recipient pickers) and are unchanged in this pass. Only
 * enlarges the label's own hit area, same as the radio precedent - the
 * checkbox glyph itself is unchanged. */
.checkbox-option {
  display: flex;
  align-items: center;
  min-height: 44px;
}

/* Primary / secondary button (S5, new - v1/design-system). Before this,
 * no button anywhere in the app was styled at all - every one rendered
 * in browser-default grey, including login.html's Sign In, which had its
 * own one-off `background:#172033` (ink, not the `action` fill S5 always
 * specified) with nothing else in the app to check it against. Building
 * the real component and applying it everywhere, rather than colouring
 * login alone, is the point - one blue button and grey everywhere else
 * would have read as more inconsistent than the plain grey it replaced.
 * font-weight uses S2's "Button label" role (600) - the one property
 * from that still-unbuilt type-scale role worth pulling forward now,
 * since every button needs a weight decision to look intentional; the
 * responsive font-size in that role is not applied here, same reasoning
 * as the eyebrow rollout not touching body copy it wasn't built for. */
.button--primary {
  background: var(--vault-color-action);
  color: var(--vault-color-surface);
  font-weight: 600;
}
/* Walkthrough, 2026-10-21 (Colin): "the invisible buttons have to go...
   it needs to have a darker color so it's visible." border was
   --vault-color-line (#E3EAF3), barely distinguishable from the page
   background it usually sits on - the button's own outline, not its
   text, was what failed to read as a control. --vault-color-line-strong
   is the token already used wherever a control's own boundary has to
   read as interactive (every text input, select and textarea) - this
   just extends that same rule to a button, rather than inventing a
   second "more visible" border for the same job. */
.button--secondary {
  background: transparent;
  color: var(--vault-color-ink-muted);
  border: 1px solid var(--vault-color-line-strong);
  font-weight: 600;
}

/* Destructive action (component vocabulary proposal, approved after two
 * revision rounds - docs/briefs/COMPONENT_VOCABULARY_PROPOSAL.md).
 * A button that reads as different from Save before the click, not only
 * inside the confirmation dialog that follows it - every Retire, Remove
 * and Revoke in the product before this used button--secondary,
 * identical to the Cancel beside it. Same real error token the confirm
 * dialog's own accent and the field error state already use, not a new
 * red. Restored here after being found missing from this file (a prior
 * attempt was silently dropped - a comment elsewhere in this file that
 * named a wildcard class pattern ending the same two characters a CSS
 * comment closes with, swallowed by the parser's own recovery, caught by
 * getComputedStyle returning the browser default instead of this rule -
 * reading the file would not have shown it, and does not prove this one
 * landed either; verified the same way below). */
.button--danger {
  background: var(--vault-color-error);
  color: var(--vault-color-surface);
  font-weight: 600;
}
/* An outline variant for a destructive action that should not outweigh
 * a primary action on the same screen - the same relationship button--
 * secondary already has to button--primary. */
.button--danger-outline {
  background: transparent;
  color: var(--vault-color-error);
  border: 1px solid var(--vault-color-error);
  font-weight: 600;
}

/* Destructive confirmation (component vocabulary proposal, approved
 * after two revision rounds - docs/briefs/COMPONENT_VOCABULARY_
 * PROPOSAL.md), Stage 3. A native <dialog>, not window.confirm() -
 * every destructive action in the product before this used the
 * browser's own unstyled confirm() popup, outside the design system
 * entirely, identical in weight to a JavaScript alert(). <dialog> is a
 * platform element, not a library - matches this app's own
 * "deliberately boring" infrastructure posture (CLAUDE.md S7) the same
 * way sidebar-toggle.js's own plain DOMContentLoaded listener does.
 * showModal() provides the real modal behaviour (focus trap, Escape to
 * close, ::backdrop) for free - nothing here reimplements that.
 *
 * Markup contract (see confirm-dialog.js): a trigger carries
 * data-confirm-dialog="<dialog id>"; the <dialog> itself holds a title,
 * a body sentence and two actions, the second a real <form
 * method="post"> so the destructive request is a real submission, not a
 * JS side effect - a user with JavaScript disabled still reaches the
 * destructive route directly, the same "disabled, not omitted" honesty
 * rule _empty_state.html's own Subscribe button already follows. */
.confirm-dialog {
  border: none;
  border-radius: var(--vault-radius-panel);
  padding: 0;
  max-width: 420px;
  width: calc(100% - 32px);
  box-shadow: 0 12px 40px rgba(11, 31, 58, 0.18);
}
.confirm-dialog::backdrop {
  background: rgba(11, 31, 58, 0.45);
}
.confirm-dialog__body {
  padding: 24px;
}
.confirm-dialog__title {
  margin: 0 0 8px;
}
.confirm-dialog__message {
  margin: 0;
  color: var(--vault-color-ink-muted);
}
.confirm-dialog__actions {
  display: flex;
  gap: 8px;
  justify-content: flex-end;
  margin-top: 20px;
}
/* The one visual escalation beyond the button colour itself: a left
 * accent on the dialog body, the same border-left-accent idiom
 * .banner--success/--error/--notice already use elsewhere in this file -
 * extending a pattern that already means "this message carries a tone",
 * not inventing a second one for dialogs specifically. */
.confirm-dialog--danger .confirm-dialog__body {
  border-left: 4px solid var(--vault-color-error);
}

/* Design coherence audit, part 2 step 2: fg/bg promoted to real tokens
 * (error/error-tint, new) rather than the literal hex this class shipped
 * with originally - "replace rather than alias" (S1's own rule for a
 * near-identical literal), zero visual change here since the token values
 * are exactly this class's own pre-existing colours. The border stays a
 * literal - no other consumer needs this exact shade, and S4 already
 * established that a chip/box border can stay literal without being a
 * second, undeclared palette (see .badge's own borders). */
.error {
  background: var(--vault-color-error-tint);
  border: 1px solid #f0caca;
  color: var(--vault-color-error);
  padding: 12px;
  border-radius: 8px;
  margin-bottom: 16px;
}

/* Row / list item (S5). Decided (v1/design-system, post-commit-4): both
 * .card and .row track the single 12px token rather than splitting into
 * two - accept the small visible change (.card 14px->12, .row 10px->12)
 * rather than carrying three radii for one token's worth of meaning.
 * 2026-10-21: that token is now --vault-radius-panel, not --vault-
 * radius-card (walkthrough, Colin) - the value here is unchanged (12px),
 * only the name, now that --vault-radius-card means the tile/chip/
 * button group's 20px instead (see .card's own comment). Base shell
 * only: a linked row (my_surveys.html and others, not converted yet)
 * will need its own flex/chevron modifier when it lands. */
.row {
  padding: 16px;
  border: 1px solid var(--vault-color-line);
  border-radius: var(--vault-radius-panel);
}

/* Row modifier: the whole row is a link (S5's "chevron affordance" -
 * my_surveys.html, commit 5, is the first caller; base .row above stays
 * the plain non-link shell parent_dashboard.html already uses). */
.row--link {
  display: flex;
  justify-content: space-between;
  align-items: center;
  text-decoration: none;
  color: inherit;
}

/* In-card row variant (Colin, 2026-10-15, approved - supersedes part of
 * the console-redesign previews): radius.card went 12px->20px in commit
 * 1, and a .row nested inside a .card reaches a corner radius equal to
 * half a short row's own height, reading as a pill - Jamie's own
 * prototypes never nest one rounded box inside another; his inner rows
 * are flat, separated by a hairline, inside one rounded card. This is an
 * in-card variant, not a change to .row/.row--link everywhere - a
 * standalone, page-level row (not a descendant of .card) keeps its own
 * border and radius unchanged, matched by the bare `.row`/`.row--link`
 * rules above. The adjacent-sibling selector puts a hairline between
 * consecutive rows and never adds one after the last - no :last-child
 * exception needed. Also resolves the identical nested-corner problem
 * mobile found on 14 screens; apply this same principle to any other
 * radius.card element found nested inside a .card, not only `.row`. */
.card .row {
  border: none;
  border-radius: 0;
  /* The boards' own in-card row: `padding:10px 0` with a hairline above
     every row, including the first - so a row's text lines up with the
     card's heading instead of sitting indented past it, and the heading
     has a rule under it. --vault-color-line-soft is #EEF1F6, the board's
     own hairline. This replaces a `+ .row` between-rows rule and the base
     row's 16px inset, both of which the comparison showed as a visibly
     taller, visibly indented card (2.283). */
  border-top: 1px solid var(--vault-color-line-soft);
  padding: 10px 0;
}

/* Status badge (S5/S4). Only the two families this commit's caller
 * (parent_dashboard.html's survey status text) needs - action, pending
 * and attention are not here yet, added when a template first needs
 * them. Border colours are DESIGN_SPEC.md S4's own literal per-chip
 * values, not aliased to a core token - decided (v1/design-system,
 * post-commit-4) to stay that way: S4's table already fully specifies
 * bg/fg/border per chip as its own small palette, distinct from S1's
 * general-purpose 16, and that palette should not collapse into a core
 * token. */
/* A withdrawn recognition's row (C109, `CONF-23`: nothing is hard-deleted, so
 * a mistaken one is a state change and stays visible with its reason).
 *
 * Quieted, not hidden - a coach looking at the season should see what they
 * took back, and the row carries a "Withdrawn" pill and the reason in words,
 * so the tint is never the only thing saying so. */
.row--withdrawn {
  background: var(--vault-color-page);
}
.row--withdrawn td {
  color: var(--vault-color-ink-soft);
}

/* C109 D8's kind picker: the nine recognitions, each a tappable option with
 * its badge, its label and whether it has a limit.
 *
 * A real `<label>` wrapping a visually-hidden radio, so the whole option is
 * the target and the keyboard reaches it - not a div with a click handler,
 * which Tab skips. `--on` is the picked one, `--taken` the one D6's cap has
 * already used this season: stated and uncliackable, with the holder's name
 * beside it, which is §9's "disabled is for a destination that does not
 * exist" rather than a control that refuses on submit.
 *
 * The grid is three across and falls to two then one, because nine options
 * in one column is a scroll and in four is unreadable at 1320. */
.kind-grid {
  display: grid;
  grid-template-columns: repeat(3, minmax(0, 1fr));
  gap: var(--vault-space-12);
}
.kind-option {
  display: flex;
  flex-direction: column;
  gap: 4px;
  padding: 11px 12px;
  min-height: 44px;
  box-sizing: border-box;
  border: 1px solid var(--vault-color-line-strong);
  border-radius: var(--vault-radius-panel);
  background: var(--vault-color-surface);
  cursor: pointer;
}
.kind-option:hover {
  border-color: var(--vault-color-ink-soft);
}
.kind-option--on {
  border: 2px solid var(--vault-color-action);
  background: var(--vault-color-action-chip);
}
.kind-option--taken {
  border-style: dashed;
  background: var(--vault-color-page);
  color: var(--vault-color-ink-soft);
  cursor: default;
}
@media (max-width: 1100px) {
  .kind-grid { grid-template-columns: repeat(2, minmax(0, 1fr)); }
}
@media (max-width: 767px) {
  .kind-grid { grid-template-columns: 1fr; }
}

/* C109's recognition badges (Colin, 2026-11-08, D3 and D9: "each kind carries
 * a badge that renders in the athlete's Vault, so a recognition reads as an
 * award and not as a line of text").
 *
 * **One medallion family, ten keys, three cues.** The ground says how singular
 * a kind is - filled for the three that name one athlete, outlined for the ones
 * a coach may give widely - the glyph says which kind, and the label beside it
 * says it in words. Nothing rests on colour: at 20px the glyph carries it and
 * the fill barely reads, which is why `recognition_badge` in `_kit.html` never
 * renders without a label beside it.
 *
 * **Season Participant has no medallion at all** (D9: "visibly plainer"). It is
 * the one kind nobody earns by standing out, so it is a plain ring in the quiet
 * grey rather than a shield - a difference in silhouette, not just in tone,
 * which is the only version of "plainer" that survives being seen next to the
 * other nine.
 *
 * The art lives in `_kit.html`'s macro; these rules carry its colours, so a
 * badge recolours with the product and no page mixes a hex at a call site. The
 * keys are `services/recognitions.py::RecognitionKind.badge` - that module is
 * the authority and `tests/test_recognition_badges.py` compares the two sets. */
.badge-art {
  display: inline-block;
  vertical-align: middle;
  flex: 0 0 auto;
}
/* 2.330 (f): the ribbon's two tails, under the medal. They take the ground's
 * own outline so a filled badge's ribbon fills with it, and they are drawn
 * before the disc so the disc covers where they meet. */
.badge-art .ba-ribbon {
  fill: var(--vault-color-surface);
  stroke: var(--vault-color-ink);
  stroke-width: 2.6;
  stroke-linejoin: round;
}
.badge-art--filled .ba-ribbon {
  fill: var(--vault-color-ink);
}
.badge-art .ba-ground {
  fill: var(--vault-color-surface);
  stroke: var(--vault-color-ink);
  stroke-width: 2.6;
}
.badge-art--filled .ba-ground {
  fill: var(--vault-color-ink);
  stroke: none;
}
.badge-art .ba-glyph {
  fill: none;
  stroke-width: 3.4;
  stroke-linecap: round;
  stroke-linejoin: round;
}
/* The three singular kinds: a gold glyph on the filled ground. */
.badge-art--mvp .ba-glyph,
.badge-art--player-of-the-game .ba-glyph,
.badge-art--coachs-award .ba-glyph {
  stroke: var(--vault-color-award);
}
.badge-art--mvp .ba-glyph--solid,
.badge-art--player-of-the-game .ba-glyph--solid {
  fill: var(--vault-color-award);
  stroke: none;
}
/* The six a coach may give widely, each in a tone already in the palette. */
.badge-art--most-improved .ba-glyph { stroke: var(--vault-color-action); }
.badge-art--leadership .ba-glyph { stroke: var(--vault-color-pending); }
.badge-art--leadership .ba-glyph--solid { fill: var(--vault-color-pending); stroke: none; }
.badge-art--hustle .ba-glyph { stroke: var(--vault-color-attention); }
.badge-art--best-teammate .ba-glyph { stroke: var(--vault-color-success); }
.badge-art--top-performer .ba-glyph--solid { fill: var(--vault-color-action); stroke: none; }
.badge-art--peer-nomination .ba-glyph { stroke: var(--vault-color-success); }
/* 2.330 (g): the coach's own recognition. Ink, where MVP's star is gold on a
 * filled ground - the one kind whose glyph says nothing about the kind,
 * because the coach's words do. Without a rule here `.ba-glyph` carries no
 * stroke of its own and the star would not draw at all. */
.badge-art--custom .ba-glyph { stroke: var(--vault-color-ink); }
/* D9: no medallion, the quiet grey, a ring rather than a shield - and since
 * 2.330 (f) a ring on a ribbon, with the ribbon kept as quiet as the ring.
 * Without this line the tails took the default ink stroke and the one badge
 * meant to recede was the loudest thing in the row. */
.badge-art--season-participant .ba-ground,
.badge-art--season-participant .ba-ribbon {
  fill: none;
  stroke: var(--vault-color-line-strong);
  stroke-width: 2.6;
}
.badge-art--season-participant .ba-glyph {
  stroke: var(--vault-color-ink-soft);
}
.badge-art--season-participant .ba-glyph--solid {
  fill: var(--vault-color-ink-soft);
  stroke: none;
}

/* 2.313's filter chip: the control that narrows a list in place, as
 * distinct from `.badge` above, which reports a state and does not click.
 * Values from `docs/boards/redraw-childs-vault/Main.dc.html` - the picked
 * 2.312 board - where the chips read "Everything 9 · Evaluations 4 ·
 * Recognitions 2 · Uploads 3": 7px/15px, 13px, 600, a pill, outlined when
 * off and filled with the page's own ink when on.
 *
 * **Added because three places were drawing it without a class** (measured
 * 2026-11-08, which is the condition §9 sets for reopening the inline-style
 * consolidation): `my_vault.html` carried the fill inline over `.badge`,
 * `.sl-pill` sits in `static/survey-library.css` and `.co-chip` in
 * `static/coaching.css`. Only `my_vault.html` is converted here - see
 * docs/FINDINGS.md, because **the two boards disagree about the fill**: the
 * survey library's draws the picked pill in action blue (`surveys/Library-B`,
 * `Library-C`) and the Vault's in ink. One `.chip--on` cannot be both, and
 * which it is belongs to whoever owns that answer, not to this commit.
 *
 * `aria-current="true"` is the state a reader hears; `.chip--on` is the one
 * they see. Both, always - the colour alone says nothing to a screen reader,
 * which is the same reason the sidebar's current row carries it. */
.chip {
  display: inline-flex;
  align-items: center;
  padding: 7px 15px;
  border-radius: var(--vault-radius-pill);
  font-size: 13px;
  font-weight: 600;
  text-decoration: none;
  white-space: nowrap;
  border: 1px solid var(--vault-color-line-strong);
  background: var(--vault-color-surface);
  color: var(--vault-color-ink);
}
.chip:hover {
  border-color: var(--vault-color-ink-soft);
}
.chip--on {
  background: var(--vault-color-ink);
  border-color: var(--vault-color-ink);
  color: var(--vault-color-surface);
}

/* 2.354 (Colin, 2026-11-19 08:18): the divider in a filter row whose chips
 * narrow on two axes - "a Favorites filter past a divider from the kind
 * chips". A chip past it can be on at the same time as a chip before it, and
 * without the divider it reads as one more kind. Drawn by `_kit.html`'s
 * `chip_row` for any chip carrying `split`, and sized to the chips beside it
 * rather than to the row, so it holds when they wrap. */
.chip-row__split {
  flex: 0 0 1px;
  align-self: stretch;
  min-height: 22px;
  background: var(--vault-color-line-strong);
  margin: 0 4px;
}

/* 2.354's own chip tone: the mark is not a state and not a kind, so it takes
 * neither the kind chips' ink nor a status colour. `--vault-color-warning` is
 * the family the kit already uses for "noticed, not urgent". */
.chip--star {
  border-color: var(--vault-color-warning);
  color: var(--vault-color-warning);
}
.chip--star.chip--on {
  background: var(--vault-color-warning-tint);
  border-color: var(--vault-color-warning);
  color: var(--vault-color-warning);
}
.badge {
  /* 2.283: the board's status pill - a tint and its own ink, no border
   * (docs/boards/director/A-Rosters.dc.html and A-Teams.dc.html). */
  display: inline-flex;
  align-items: center;
  padding: 3px 10px;
  border-radius: var(--vault-radius-pill);
  font-size: 12.5px;
  font-weight: 600;
  border: 0;
  white-space: nowrap;
}
/* **2.330 (d)** (Colin, 2026-11-16 14:02, Jamie's coach review): an
 * evaluation reads Published green, Returned to you red, **Not started
 * yellow**. This badge was the page's own grey, which said "nothing here"
 * where the decision wants "nothing here yet, and it is on somebody".
 *
 * The yellow is a token of its own, `warning`, rather than either of the two
 * it might have borrowed: `award` is the gold a recognition's medallion is
 * drawn in (2.330 (f) is about those), and `pending` is the purple that means
 * out with somebody else. A third meaning on either would be one colour
 * saying two things.
 *
 * The same three carry Rosters' Vault account column (2.330 (a)), which is
 * why the tone is a class of its own below and this state reads it. */
.badge--warning,
.badge--not-started {
  background: var(--vault-color-warning-tint);
  color: var(--vault-color-warning);
  border-color: #E9D79A;
}
.badge--success {
  background: var(--vault-color-success-tint);
  color: var(--vault-color-success);
  border-color: #BFE3D2;
}
/* "In progress" chip (S4). organization_surveys.html, commit 11: the first
 * chip whose S4 background doesn't happen to equal an existing S1 token -
 * success/not-started's bg and fg values coincide exactly with success-tint/
 * success and page/ink-soft, which is why those two could alias straight
 * to core tokens with only the border staying literal. #E8F0FE used to
 * have no S1 equivalent (action-tint is #F3F7FF, a different colour) -
 * it is now its own token, action-chip (design/tokens.json, Colin's own
 * instruction: this value had two live consumers before a third,
 * mobile's tokens.json import, made the literal a drift risk). fg does
 * equal `action` exactly and uses that token; border stays literal
 * (single consumer - see .tile--new below). */
.badge--action {
  background: var(--vault-color-action-chip);
  color: var(--vault-color-action);
  border-color: var(--vault-color-action);
}
/* "Pending verification" chip (S4). #ECEAFA used to have no S1 equivalent
 * (there was no `pending-tint` token - unlike action/success/attention,
 * `pending` was a foreground-only colour in S1's table) - it is now
 * pending-tint (design/tokens.json), filling that gap the same way
 * action-chip fills action's. fg equals `pending` exactly and uses the
 * token; border stays literal (single consumer - see .tile--pending
 * below). */
.badge--pending {
  background: var(--vault-color-pending-tint);
  color: var(--vault-color-pending);
  border-color: var(--vault-color-pending);
}
/* "Another attempt needed" / "Overdue" chip (S4) - same family, S4 gives
 * them identical bg/fg/border. Unlike --pending, both bg and fg happen to
 * equal S1 tokens exactly (attention-tint, attention), so only the border
 * stays literal, same pattern as --success and --not-started. */
.badge--attention {
  background: var(--vault-color-attention-tint);
  color: var(--vault-color-attention);
  border-color: var(--vault-color-attention);
}
/* "Not activated" chip (S4) - the genuine `neutral` token
 * (#F1F2F4/#6B7280/#E1E3E8), a different grey from --not-started's
 * page/ink-soft (that rename note lived here until this class existed;
 * see git history). 6.6, action derived-label surface: also the chip
 * for CANCELLED, which S4 doesn't cover - reconsidered from an initial
 * --not-started reuse on review, since "not started" (work ahead) and
 * "cancelled" (work withdrawn) sharing a treatment would hide the
 * distinction a badge exists to show. Both CANCELLED and NOT_ACTIVATED
 * represent "not live work" for an administrative/external reason
 * (withdrawn by a coach; blocked on a membership never claimed) - a
 * real pairing, not a second problem. Recorded in docs/FINDINGS.md for
 * the design session to land in DESIGN_SPEC.md S4 as an eighth chip. */
.badge--neutral {
  background: #F1F2F4;
  color: var(--vault-color-neutral);
  border-color: #E1E3E8;
}
/* Nine chips specified in DESIGN_SPEC.md S4, one ("Self-reported") retired
 * by 6.6's D2a resolution - SELF_REPORTED only ever means "awaiting a
 * coach" (see docs/FINDINGS.md), so it maps to --pending, never a chip of
 * its own ("Self-reported" used to reuse --success's colours in an
 * earlier draft of this comment; there's no longer a state that value
 * would apply to). "Quiz passed" stays excluded (DESIGN_SPEC.md S8 - no
 * backend concept, do not build). Effective count: eight specified
 * (seven original plus CANCELLED, new), all built. */

/* Console tile row (component vocabulary proposal, approved after two
 * revision rounds - docs/briefs/COMPONENT_VOCABULARY_PROPOSAL.md).
 * Three tone-coded tiles above a console's detail cards, each a plain
 * in-page anchor to its own card - "the mobile app had 3 colored
 * boxes... in 1 second of viewing it, I knew exactly what was
 * happening" (Colin). Reuses the three tones the badge family above
 * already assigns to these exact questions, rather than inventing
 * tile-specific colours: attention/attention-tint (.badge--attention's
 * own pair), action/action-chip (.badge--action's own pair - "New
 * Updates" is new content, the same family as an in-progress chip), and
 * pending/pending-tint (.badge--pending's own pair). The two tokenised
 * backgrounds are not new - they are the exact values .badge--action/
 * .badge--pending already use above, now shared via design/tokens.json
 * rather than repeated as a second literal (Colin's own instruction:
 * each was already a two-consumer hex before this tile existed, and
 * mobile's own tokens.json import makes a third literal a drift risk).
 * Borders stay literal - single consumer each, mobile's own Card
 * supplies its own edge and won't consume them. First caller:
 * dashboard.html's player console. */
.tile-row {
  display: grid;
  grid-template-columns: repeat(3, minmax(0, 1fr));
  gap: var(--vault-space-12);
  margin-bottom: var(--vault-space-16);
}
/* 2.310 (14) (Colin, 2026-11-07): "four tiles on one row". The rule
 * above is `repeat(3, ...)` at every width, so the coach's Actions page -
 * Done, To verify, In progress, Not started - put its fourth tile on a
 * second row of its own at 1320.
 *
 * The same selector the phone rules at the foot of this file already use
 * to lift a four-tile row into 2x2, which is where the four-tile case was
 * anticipated and the desktop half was not. Added by API for its own page
 * and routed to Design, whose kit this is (2.283 (1)).
 *
 * **Kept as `:has()` rather than folded into a `--four` modifier**
 * (Design, 2026-11-07, the queue's own question). How many tiles a row
 * holds is already in the DOM, and the phone rules below read it from
 * there; a modifier would encode the same fact a second way, in a place
 * that can disagree with it - and the disagreement is silent in the one
 * direction that matters. A caller that forgets the class gets back
 * exactly the defect 2.310 (14) reports, with nothing to notice it,
 * because absence renders. `.tile-row:has(...)` cannot be forgotten.
 * A browser without `:has()` keeps the three-column grid, which is the
 * same safe failure the phone rule chose.
 *
 * **Scoped to exactly four.** `:nth-child(4)` alone means "four or
 * more", and `coach_evaluations_index.html` carries five (Published,
 * With your reviewer, Awaiting your review, Returned to you, and Not
 * started behind `show_not_started`) - measured, and recorded in
 * tests/test_tile_row_phone_layout.py::CONSUMERS since 2.253a. At five
 * `repeat(4, ...)` leaves a lone fifth tile in the first column of a
 * second row, where the 3-up grid that page was built with gives 3 + 2.
 * Colin's decision names four, so four is what fires: a fifth tile is a
 * row size nobody has drawn, and it keeps the shape it already had
 * rather than taking a new one from a rule written for another page.
 * The phone half needs no such scoping - 2x2 then one is what "a 4-tile
 * row goes 2x2" gives a fifth tile anyway. */
.tile-row:has(> .tile:nth-child(4)):not(:has(> .tile:nth-child(5))) {
  grid-template-columns: repeat(4, minmax(0, 1fr));
}
/* 2.283 (Colin, 2026-10-29): every value here is the board's, from
 * docs/boards/director/Main.dc.html - left-aligned, not centred, which is
 * what 2.243's pictures showed the built page getting wrong. */
.tile {
  display: block;
  text-decoration: none;
  /* Deliberately still --vault-radius-card, not --vault-radius-panel
     (walkthrough, 2026-10-21) - this token now means the tile/chip/
     button group specifically (20px); .card's own comment has the full
     reasoning, including why the value couldn't just change in place. */
  border-radius: var(--vault-radius-panel);
  padding: 14px 16px;
  /* The boards draw a tile's border in its own tone at full strength -
     #9B2C2C on the red tile, #0A6BE0 on the blue, #4B3BA8 on the purple -
     and each variant below already sets exactly that as its `color`. So
     the border is currentColor and a new tone needs no second line (the
     four pale border-colors this replaces were close enough to the tint
     to be invisible at 1px, which is how the console's tiles photographed
     as borderless beside a board whose tiles are outlined). */
  border: 1px solid currentColor;
}
.tile__count {
  display: block;
  font-size: 28px;
  font-weight: 700;
  line-height: 1.1;
}
.tile__label {
  display: block;
  font-size: 14px;
  font-weight: 600;
  line-height: 1.25;
  color: var(--vault-color-ink);
  margin-top: 2px;
}
/* Walkthrough, 2026-10-21 (Colin's cleared decision 4): "waiting on
   you"/"nothing to do"/"out with them" - a short sentence under the
   label and count, not a fourth number. **Muted ink, not the tile's tone**
   (Mobile, FINDINGS 2026-11-13: this comment said "same tone colour as the
   rest of the tile (inherited)" while the rule below set muted ink, and the
   rule is what ships) - lighter weight and a touch smaller too, so it reads as
   commentary on the count above it rather than a second label. One muted
   colour also means 2.317's neutral tile needs no rule of its own here. */
.tile__caption {
  display: block;
  font-size: 12.5px;
  font-weight: 400;
  line-height: 1.2;
  color: var(--vault-color-ink-muted);
}
.tile--attention {
  background: var(--vault-color-attention-tint);
  color: var(--vault-color-attention);
}
.tile--new {
  background: var(--vault-color-action-chip);
  color: var(--vault-color-action);
}
.tile--pending {
  background: var(--vault-color-pending-tint);
  color: var(--vault-color-pending);
}
/* Staff console tile (_console_buckets.html), second caller of the tile-
 * row pattern above. The staff Needs Attention tile uses error/error-tint
 * rather than attention/attention-tint - a muted brick red reads as more
 * urgent than the player console's amber, which is the point for a staff
 * audience triaging real work queues rather than seeing their own
 * activity. .tile--attention (amber) stays exactly as it is for the
 * player console; this is a sibling, not a recolour. */
.tile--attention-error {
  background: var(--vault-color-error-tint);
  color: var(--vault-color-error);
}
/* 2.318 (g) (Colin, 2026-11-13, after Jamie's review): the Actions index's
 * four state tiles are tinted by meaning - Done green, To verify amber, In
 * progress blue, Not started red. Three of the four had a tone already
 * (.tile--attention, .tile--new, .tile--attention-error); this is the fourth,
 * a sibling of them written the same way, on the same success token the
 * ledger bar's own Done segment uses.
 *
 * **Design's file, added by API** because (g) is the Actions index's item and
 * it is one rule; routed in docs/FINDINGS.md so the kit's owner can move or
 * rename it. A new class name, so nothing already in here changes. */
.tile--success {
  background: var(--vault-color-success-tint);
  color: var(--vault-color-success);
}
/* 2.317 (Colin, 2026-11-13: "Don't draw attention to a 0"). A tile is
 * coloured only when its count is above zero; at zero it is neutral - no
 * wash, the page's own line as its border, and a muted numeral. The label and
 * the caption are unchanged either way, because what the tile counts does not
 * depend on how many there are.
 *
 * `.tile.tile--zero`, two classes, so it wins over any tone rule whatever
 * order they end up in: a modifier that only works because it sits last in the
 * file is one stylesheet reshuffle from silently doing nothing. The same
 * decision runs on the phone (`CONF-11`), where Mobile builds it. */
.tile.tile--zero {
  background: transparent;
  color: var(--vault-color-line-strong);
}
.tile.tile--zero .tile__count {
  color: var(--vault-color-ink-muted);
}

/* The phone half of the tile row (2.242, Colin 2026-10-28). Without this
 * `.tile-row` kept `repeat(3, minmax(0, 1fr))` at every width, so at 390px
 * three tiles became three one-word-per-line columns - the defect Colin saw
 * in the comparison page's Phone view. §9's phone rule: there is no
 * horizontal waste to recover at that width, so the answer is a dense row
 * per item rather than a squeezed column per item.
 *
 * Colin's decision, 2026-10-28: a 3-tile row becomes one compact line each;
 * a 4-tile row goes 2x2. So one column is the default here and `:has()`
 * lifts a row of four into two. A browser without `:has()` support drops
 * only that one rule and gets the single column, which is the safe way to
 * fail rather than the squeezed grid.
 *
 * Written as a max-width addition rather than converting `.tile-row` and
 * `.tile` to the mobile-first shape `.stat__value` above uses: every
 * declaration here exists only inside this query, so desktop is untouched
 * by construction instead of by re-reading two heavily-commented base
 * rules.
 *
 * The "measured 2026-10-28: the only two consumers are
 * _console_buckets.html and dashboard.html, three tiles each, so the 2x2
 * half is drawn from the decision rather than from a row that exists yet"
 * that stood here is expired, and it was read as current when the desktop
 * four-up rule above was written. Four-tile and five-tile rows both exist
 * now; CONSUMERS in tests/test_tile_row_phone_layout.py is the live count
 * and fails on a new consumer, which is where to look instead of here. */
@media (max-width: 767px) {
  .tile-row {
    grid-template-columns: 1fr;
  }
  /* One compact line each: the count and its label on one line, left
     aligned, no centred card chrome. The caption ("waiting on you") takes
     the full width below, so a tile is at most two lines. */
  .tile-row > .tile {
    flex-direction: row;
    align-items: baseline;
    justify-content: flex-start;
    flex-wrap: wrap;
    text-align: left;
    gap: 10px;
    padding: 11px 14px;
  }
  .tile-row > .tile > .tile__label {
    flex: 1 1 auto;
  }
  .tile-row > .tile > .tile__caption {
    flex: 0 0 100%;
  }
  /* Four or more tiles: 2x2, and each tile keeps the centred stack it has
     on desktop - a compact row inside a half-width cell is cramped, which
     is the reason the two cases differ at all. */
  .tile-row:has(> .tile:nth-child(4)) {
    grid-template-columns: repeat(2, minmax(0, 1fr));
  }
  /* **2.327's two phone rules** (API's finding, 2026-11-16, left to Design
   * because they live here and two sessions in one stylesheet is how a
   * merge drops a side).
   *
   * **Opt-in, not blanket.** "Two buttons in a row split it at 390" would
   * reach every page that puts two buttons together, and most of them are
   * a consequential action beside a way out, where a half-width Cancel
   * reads as the same weight as the thing it cancels. So a page asks for
   * it: `u-buttons-split` on the row, `u-button-wide-phone` on a button
   * that should fill the width. Both do nothing above 767px. */
  .u-buttons-split > button,
  .u-buttons-split > .button--primary,
  .u-buttons-split > .button--secondary {
    flex: 1 1 0;
  }
  .u-button-wide-phone {
    display: block;
    width: 100%;
    text-align: center;
  }
  /* **The same selector as the desktop four-up rule, repeated** (API's
   * finding, 2026-11-16). A media query does not add specificity: the
   * desktop rule above is
   * `.tile-row:has(> .tile:nth-child(4)):not(:has(> .tile:nth-child(5)))`,
   * and `:has()` and `:not()` each take their argument's specificity, so it
   * scores one class and two pseudo-classes more than the rule just above -
   * and won at 390 on every page with exactly four tiles. Every four-tile
   * row drew four across on a phone, and the test said the rule existed.
   *
   * Matching the selector exactly makes the two tie, and this one is later
   * in the file, which is what decides a tie. The rule above still carries
   * five or more, where the desktop rule deliberately does not apply. */
  .tile-row:has(> .tile:nth-child(4)):not(:has(> .tile:nth-child(5))) {
    grid-template-columns: repeat(2, minmax(0, 1fr));
  }
  .tile-row:has(> .tile:nth-child(4)) > .tile {
    flex-direction: column;
    align-items: center;
    justify-content: center;
    flex-wrap: nowrap;
    text-align: center;
    gap: 4px;
    padding: 14px 8px;
  }
}
/* Club facts row (2.251's Console A, Colin 2026-10-28): the director
 * console's "Where <club> stands" - four cards, each a label, one headline
 * fact, one detail line, and the page it opens.
 *
 * Neither of the two components it sits between would do. `.tile` is a
 * number over a label, centred; these are a sentence over a sentence, and
 * a 28px count is the wrong emphasis for "1 of 22 running". `.card` is the
 * three bucket cards below, and four of those at .card's own padding would
 * outweigh the sections they summarise.
 *
 * Phone (<=767px) goes 2x2, which is Colin's own tile-row decision of the
 * same day - "a 3-tile row becomes one compact line each, a 4-tile row goes
 * 2x2" - applied to the second 4-up row on the same page rather than
 * answered a second way. The detail line is kept to a few words for that
 * reason: at 390px each cell is about 170px wide. */
.fact-row {
  display: grid;
  grid-template-columns: repeat(4, minmax(0, 1fr));
  gap: var(--vault-space-8);
  margin-bottom: var(--vault-space-16);
}
/* The safeguarding block's two halves: the same component two across, so
   each half reads as one of a pair rather than as its own card. */
.fact-row--pair {
  grid-template-columns: repeat(2, minmax(0, 1fr));
}
/* The director console's five facts: four about the club and the support tile (D4, 2.283). */
.fact-row--five {
  grid-template-columns: repeat(5, minmax(0, 1fr));
}
.fact {
  display: block;
  text-decoration: none;
  color: inherit;
  background: var(--vault-color-surface);
  border: 1px solid var(--vault-color-line);
  border-radius: var(--vault-radius-card);
  padding: 14px 16px;
}
.fact__label {
  display: block;
  font-size: 11px;
  font-weight: 600;
  letter-spacing: 0.07em;
  text-transform: uppercase;
  color: var(--vault-color-ink-muted);
}
.fact__headline {
  display: block;
  font-size: 15px;
  font-weight: 700;
  margin-top: 4px;
}
.fact__detail {
  display: block;
  font-size: 13px;
  color: var(--vault-color-ink-muted);
  margin-top: 4px;
}
/* The fact's own way back in. A link inside a card that is itself a link
   would be two destinations; a `.fact` that is an <a> carries this as its
   last line instead, so the whole card is the one target. */
.fact__link {
  display: block;
  font-size: 13px;
  color: var(--vault-color-action);
  margin-top: 8px;
}
/* C36's exception state - a minor no adult can reach - is the one fact on
   this row that is not neutral. Same attention pair .tone-attention-panel
   uses, so the console has one attention treatment and not two. */
.fact--attention {
  border-color: var(--vault-color-attention);
  background: var(--vault-color-attention-tint);
}
.fact--attention .fact__headline {
  color: var(--vault-color-attention);
}
@media (max-width: 767px) {
  .fact-row {
    grid-template-columns: repeat(2, minmax(0, 1fr));
  }
}

/* Pairs each _console_buckets.html card with its own tile's border colour,
 * so the eye joins a tile's number to the card it counts - .card's own
 * default border (--vault-color-line) doesn't carry that link. Literal
 * borders, matching the tile rules above - same single-consumer reasoning. */
.card--tone-attention-error {
  /* The tone itself, as .card--tone-new and .card--tone-pending below
     already do, and as the board draws it (#9B2C2C). It was the one pale
     value of the three, which photographed as a card with no border. */
  border-color: var(--vault-color-error);
}
.card--tone-new {
  border-color: var(--vault-color-action);
}
.card--tone-pending {
  border-color: var(--vault-color-pending);
}

/* A card filled with the action tint rather than bordered in a tone
 * (2.305 (3), docs/boards/c98/Main.dc.html's "Your club does not see
 * this"): the board draws its one aside note as a tinted panel with the
 * ordinary line border, which is .card plus a fill and nothing else.
 * The same fill .banner--notice and .empty-state--locked-panel already
 * use for "this is the system telling you something", not a new token. */
.card--tint-action {
  background: var(--vault-color-action-tint);
}
/* Player console redesign (2026-10-19, docs/briefs/PLAYER_CONSOLE_
 * REDESIGN_PROPOSAL.md's tile row, now in the admin console's own
 * tinted-card shape rather than the proposal's original white-card
 * version): second real consumer of .tile--attention (amber, not
 * .tile--attention-error's staff red - the comment above already
 * anticipated this). Matches .tile--attention's own border exactly,
 * same single-consumer-literal reasoning as its three siblings above. */
.card--tone-attention {
  border-color: var(--vault-color-attention);
}

/* 2.276 (Colin, 2026-10-29), docs/boards/child/: the three tinted
   sections carry the full-strength colour on their border, as the
   board draws them - orange, blue, purple - not the pale raw hex
   they had. Same change as .tile--* above, and through the tokens. */

/* Success banner (S5's Banner component - "success-tint for success").
 * organization_survey_detail.html, commit 12: three identical inline
 * `border-left:4px solid #2f855a` confirmations (reminder sent, reminders
 * sent, survey sent) in that one file - a real, repeated caller, not a
 * speculative build. Stacks on top of `.card` rather than replacing it,
 * so callers keep .card's background/border/radius/shadow and only add
 * the accent. Uses the real `success` token, not #2f855a literally -
 * close enough to be the same colour intent, per S1's "replace rather
 * than alias". nomination_review.html has the identical inline pattern
 * (plus a second, different green use for a projected-winner highlight)
 * but is not part of this pass; it becomes this class's second caller
 * when its own turn comes, not now. */
.banner--success {
  border-left: 4px solid var(--vault-color-success);
}

/* Error/attention banner - the variant DESIGN_SPEC.md S5 recorded as a
 * known gap when .banner--success was built ("three different inline
 * error treatments still exist... unifying them is a real decision, not
 * done in this pass"). First real caller: _membership_content.html's
 * failed/cancelled-checkout notice (docs/B10_WALKTHROUGH_FINDINGS.md
 * item 4) - "the most important sentence on the page when it applies"
 * was rendering as muted body text in the middle of a long page. Uses
 * the real `attention` token ("another attempt needed" is exactly
 * S1's own definition for it), not a new red - same "replace rather
 * than alias" rule .banner--success already follows. Does not retire
 * the plain .error box (a different, older treatment, form-validation
 * shaped) or my_survey_detail.html/player_evaluations.html's own inline
 * red boxes - unifying all three is the still-open decision this leaves
 * open, one caller closer. */
/* Design coherence audit, part 2 step 2: repointed from `attention` to
 * the real `error` token now that one exists. `attention` was the
 * nearest available colour when this class was first built
 * (_membership_content.html's failed/cancelled-checkout notice) - a
 * documented stand-in, not a considered choice, per this class's own
 * original comment ("DESIGN_SPEC.md S5 recorded as a known gap...
 * unifying them is a real decision, not done in this pass"). A failed or
 * cancelled payment is a genuine failure with nothing left to do but
 * retry, which is `error`'s own meaning; `attention` is reserved for
 * incomplete-but-actionable work (overdue, another attempt needed) -
 * keeping the two apart is the actual point of adding a fourth semantic
 * colour instead of leaving three things sharing `attention`. */
.banner--error {
  border-left: 4px solid var(--vault-color-error);
}

/* Colin's own instruction, 2026-09-29: the timezone-confirm prompt's
 * contrast bug is fixed (button colours, grid placement) but the
 * component itself was never actually designed - a bare .card with no
 * visual distinction from ordinary page content, sitting flush against
 * the header with a native unstyled feel, for a prompt that is many
 * people's first authenticated impression of the product. Uses the same
 * border-left-accent idiom as .banner--success/.banner--error rather
 * than inventing a new one, on the `action` token (S1: "Primary
 * buttons, links, selected state, focus") - this banner is a question
 * asking for input, not a success or a failure, and `action` is already
 * the app's "this wants you to do something" colour. `pending` was
 * considered and rejected: S1 already assigns it a specific, different
 * meaning ("Submitted, pending verification") and reusing it here would
 * blur that, not extend it. action-tint fills the card so it reads as a
 * system prompt at a glance, distinct from the plain white cards below
 * it on the same page. */
.banner--notice {
  background: var(--vault-color-action-tint);
  border-left: 4px solid var(--vault-color-action);
}
.banner--notice .eyebrow {
  display: flex;
  align-items: center;
  gap: 6px;
  color: var(--vault-color-action);
}

/* The attention tone of the same banner (2.305 (3), docs/boards/c98/
 * Refused.dc.html): "You have not accepted the new terms" is drawn in
 * the attention tint with the attention rule and its heading in the same
 * colour. The fourth kind rather than a one-off panel, because §9 sends
 * every banner through the macro and the macro selects on kind.
 *
 * .banner--error is the closest existing tone and is the wrong one: a
 * refusal is a state the reader chose and can undo, not a failure, and
 * error carries role="alert" by default (Colin, punchlist 2.135). */
.banner--attention {
  background: var(--vault-color-attention-tint);
  border-left: 4px solid var(--vault-color-attention);
}
.banner--attention > strong {
  color: var(--vault-color-attention);
}

/* Stat tile (S5). organization_survey_detail.html's Assigned/Completed/
 * Outstanding/Response-rate row is four identical number-over-label
 * pairs - the first real caller. */
.stat {
  display: flex;
  flex-direction: column;
  gap: var(--vault-space-4);
}
/* Value type treatment (post-review): the shell shipped with no size at
 * all ("a separate decision if one is wanted later" - this is that
 * decision). Reuses S2's section-heading role rather than inventing a
 * number - that role has been specified in tokens.css since the type
 * scale landed and never actually used anywhere. Deliberately not
 * coach_actions_index.html's 30px/own-card treatment: this page is
 * already the densest in the app, and four separate cards at that size
 * would add real vertical weight it doesn't have room for. Only this
 * page's .stat is updated - coach_actions_index.html's own pattern is
 * flagged in DESIGN_SPEC.md S5 as the follow-up, not changed here. */
.stat__value {
  display: block;
  font-family: var(--vault-type-section-heading-family);
  font-weight: var(--vault-type-section-heading-weight);
  font-size: var(--vault-type-section-heading-size-mobile);
  line-height: var(--vault-type-section-heading-line-height);
}
@media (min-width: 768px) {
  .stat__value {
    font-size: var(--vault-type-section-heading-size-desktop);
  }
}

/* Results distribution (S5, new - organization_survey_detail.html, the
 * "known gap" DESIGN_SPEC.md S5 flagged when .stat was first built).
 * Two shapes, both pure CSS + server-computed percentages (no client JS):
 * .histogram for SCORE (bar height = count / the question's own max
 * bucket count - works identically whether _score_distribution returned
 * 5 value-labelled buckets or 10 range-labelled ones, since the
 * component only ever sees "a list of {count, label}"), .option-meter
 * for SINGLE_SELECT/MULTI_SELECT/BOOLEAN (bar width = pct, with the
 * label always paired with "n of m" text - never a bare percentage,
 * S4's rule). BOOLEAN is folded into the same meter as a two-option
 * list (Yes/No) rather than kept as a third, separately-rendered shape. */
.histogram {
  display: flex;
  align-items: flex-end;
  gap: 8px;
  height: 90px;
  margin-top: 10px;
}
.histogram__bar-wrap {
  display: flex;
  flex: 1;
  min-width: 20px;
  height: 100%;
  flex-direction: column;
  align-items: center;
  justify-content: flex-end;
}
.histogram__count {
  font-size: 11px;
  color: var(--vault-color-ink-soft);
  margin-bottom: 2px;
}
.histogram__bar {
  width: 100%;
  min-height: 2px;
  background: var(--vault-color-action-tint);
  border: 1px solid var(--vault-color-action);
  border-radius: 3px 3px 0 0;
}
.histogram__label {
  margin-top: 4px;
  font-size: 11px;
  color: var(--vault-color-ink-faint);
  white-space: nowrap;
}

.option-meter {
  display: grid;
  gap: 8px;
  margin-top: 10px;
}
.option-meter__label {
  display: flex;
  justify-content: space-between;
  gap: 12px;
  font-size: 13px;
}
.option-meter__track {
  height: 8px;
  margin-top: 3px;
  background: var(--vault-color-line-soft);
  border-radius: var(--vault-radius-pill);
  overflow: hidden;
}
.option-meter__fill {
  height: 100%;
  background: var(--vault-color-action);
  border-radius: var(--vault-radius-pill);
}

/* Determinate progress bar (docs/FINDINGS.md, 2026-10-14 - promoted from
 * .legacy-progress-fill/-track, which had no modern equivalent despite
 * being load-bearing in five live templates). A real upload-progress bar a
 * family sees while a video uploads, not cosmetic debt - hidden by default,
 * shown and animated by JS as upload.onprogress reports real percentages
 * (static/video-upload.js for the one template already using it as a
 * shared driver; each other template's own small inline driver sets the
 * same two properties - .progress-bar's display, .progress-bar__fill's
 * width). Same visual recipe .option-meter__track/__fill above already
 * established (8px height, pill radius, action-blue fill) - a live upload
 * indicator and a static results meter are different jobs sharing the
 * identical look, not two components that happen to coincide. */
.progress-bar {
  display: none;
  height: var(--vault-space-8);
  margin-top: var(--vault-space-12);
  background: var(--vault-color-line);
  border-radius: var(--vault-radius-pill);
  overflow: hidden;
}
.progress-bar__fill {
  height: 100%;
  width: 0%;
  background: var(--vault-color-action);
}

/* Criterion trend (TREND_VIEW_BRIEF.md, DESIGN_SPEC.md S7; extended for
 * WALKTHROUGH_FINDINGS.md items 7/8/11, and round 3's own items 2/3) -
 * pure CSS + server-computed percentages, same discipline as .histogram
 * above, plus one inline SVG holding one <polyline> per line segment
 * (services/evaluation_results.py::CriterionTrend.line_segments - a
 * maximal run of adjacent points sharing team and scale, three points
 * minimum per run; see TrendLineSegment's own docstring for why two
 * never draws one, and why a run never crosses a team or scale
 * boundary). The SVG is stretched with preserveAspectRatio="none" over
 * the exact same box .trend-chart__point is positioned in, so its
 * 0-100 coordinate space lines up with the dots' left/bottom
 * percentages with no JS measuring anything. x_percent is
 * time-proportional (item 7), not ordinal - .trend-chart__x-label is
 * positioned to match, not spread evenly. .trend-chart__point-value is
 * deliberately dominant (bold, closest to the dot) and the line (drawn
 * in .trend-chart__line via inline style, not a class - the colour
 * needs a raw value regardless of whether it's the default token or a
 * per-team override) deliberately muted and thin - TREND_VIEW_BRIEF.md
 * Q3's refinement, applied uniformly rather than only once there are
 * enough points for a line to exist at all: these are discrete
 * judgements, not a continuous measurement, and the visual weight
 * should say so regardless of N. Team distinction is colour, not shape
 * - see .trend-chart__point's own comment below and DESIGN_SPEC.md's
 * Criterion trend chart entry for why that overrules the system's
 * usual "colour means status" convention. */
.trend-chart {
  display: flex;
  gap: 12px;
  margin-top: 14px;
}
.trend-chart__axis-labels {
  display: flex;
  flex-direction: column;
  justify-content: space-between;
  flex-shrink: 0;
  max-width: 110px;
  text-align: right;
  font-size: 11px;
  color: var(--vault-color-ink-faint);
  padding: 4px 0;
  overflow-wrap: break-word;
}
.trend-chart__plot-column {
  flex: 1;
  min-width: 0;
}
.trend-chart__plot {
  position: relative;
  height: 180px;
  border-left: 1px solid var(--vault-color-line);
  border-bottom: 1px solid var(--vault-color-line);
}
.trend-chart__line {
  position: absolute;
  inset: 0;
  width: 100%;
  height: 100%;
}
/* Team distinction (WALKTHROUGH_FINDINGS.md item 8, round 3's own item
 * 3): originally shape modifiers here, overruled deliberately - see
 * evaluation_results.py's _TEAM_LINE_COLORS comment and DESIGN_SPEC.md's
 * Criterion trend chart entry for the full record. .trend-chart__point
 * is always a plain circle now; colour (background, set inline per
 * point/legend entry from TrendPoint.line_color /
 * TrendTeamLegendEntry.line_color - None for the common single-team
 * case, which keeps exactly this default action-blue) is the only thing
 * that distinguishes teams. */
.trend-chart__point {
  position: absolute;
  width: 12px;
  height: 12px;
  margin-left: -6px;
  margin-bottom: -6px;
  background: var(--vault-color-action);
  border: 2px solid var(--vault-color-surface);
  border-radius: 50%;
}
.trend-chart__point-value {
  position: absolute;
  transform: translateX(-50%);
  font-size: 12px;
  font-weight: 600;
  color: var(--vault-color-ink);
  white-space: nowrap;
}
/* Partner-demo polish pass (2026-09-22): two same-valued points close in
 * time (services/evaluation_results.py::_assign_value_label_rows decides
 * which points these are) render this label 16px further from its own
 * dot, clearing row 0's text instead of overlapping it - "3"/"3" read as
 * "33" was the finding. translateY, not a second bottom: calc() - the
 * existing bottom already encodes this point's own real score position,
 * and a transform shifts the rendered label without touching that. */
.trend-chart__point-value--row-1 {
  transform: translateX(-50%) translateY(-16px);
}
/* Position: relative/absolute, not flex-equal-width (WALKTHROUGH_
 * FINDINGS.md item 7) - x_percent is time-proportional now, so a label
 * must sit at its own point's own percentage, the same technique
 * .trend-chart__point itself already uses, not spread evenly across
 * equal-width flex items the way ordinal spacing could get away with. */
.trend-chart__x-labels {
  position: relative;
  min-height: 46px;
  margin-top: 6px;
}
/* Colin's own instruction, 2026-09-29 (family/athlete visual pass): the
 * label nearest left:100% (the most recent date, so the single most-
 * looked-at one) was wrapping to three lines while every other label on
 * the identical axis stayed on one - confirmed live, not from the
 * screenshot alone (Playwright bounding boxes: 28x39px there against
 * ~57x13px elsewhere). Root cause: an absolutely-positioned box with
 * only `left` set (no `right`) gets shrink-to-fit sized from
 * `container width - left` (CSS2.1 10.3.7) - correct for a box that
 * grows rightward from `left`, but this one also carries
 * transform:translateX(-50%), which recenters it AFTER that width
 * calculation runs. The transform is invisible to layout, so a label
 * anchored at left:100% gets ~0px of computed available width no
 * matter how much visual room the transform then gives it back. A
 * fixed `width` (not `max-width`) sidesteps the shrink-to-fit path
 * entirely - every label now reserves the same 90px box regardless of
 * position, with centered text making the extra room on a short label
 * invisible. Safe to extend past the card's own edge: neither .card
 * nor .trend-chart sets overflow:hidden, confirmed by reading both. */
.trend-chart__x-label {
  position: absolute;
  top: 0;
  transform: translateX(-50%);
  width: 90px;
  text-align: center;
  font-size: 11px;
  color: var(--vault-color-ink-faint);
  overflow-wrap: break-word;
  padding: 0 2px;
}
/* Demo walkthrough (2026-09-18) Finding 4: two labels close enough in
 * time to overlap at row 0's single height (services/evaluation_
 * results.py::_assign_x_label_rows decides which points these are, from
 * x_percent alone - never from a fixed index/count, so a chart with
 * nothing crowded renders every label at row 0 exactly as before). The
 * 18px offset is calibrated for this label's own 11px font/single line -
 * enough to clear row 0's text without a gap that reads as unrelated. */
.trend-chart__x-label--row-1 {
  top: 18px;
}

/* Empty / locked state (S5/S6 - see _empty_state.html for the macro that
 * renders into this shell). Deliberately plain: DESIGN_SPEC.md S6 calls
 * locked-state copy "informational", not an alert - no tint background,
 * no border, just spacing so the paragraph/list/footnote don't run
 * together. */
.empty-state p {
  margin: 0 0 8px;
}
.empty-state ul {
  margin: 0 0 8px;
  padding-left: 18px;
}
.empty-state p:last-child {
  margin-bottom: 0;
}
/* Lock icon (DESIGN_BRIEF_2 item 1) sits inline in the intro sentence,
 * not a separate line - it marks the state the sentence already
 * describes rather than repeating it. Still true for Empty/Failed and
 * for Locked outside the panel variant below; the panel variant
 * promotes it to its own line instead, the next escalation step past
 * "text-only -> inline icon" this same comment originally described. */
.empty-state__lock-icon {
  vertical-align: -2px;
  margin-right: 6px;
}
/* Locked-state visual pass (component vocabulary proposal, approved
 * after two revision rounds - docs/briefs/COMPONENT_VOCABULARY_
 * PROPOSAL.md). Supersedes this file's own ".empty-state p" comment
 * above ("Deliberately plain... no tint background, no border") for
 * this one branch - see _empty_state.html's own comment at its real
 * caller for why. action-tint, not a new token - the same fill
 * .banner--notice already uses for "this wants something from you",
 * which is exactly what a live Subscribe link means here now that one
 * exists (already shipped, unrelated to this round). */
.empty-state--locked-panel {
  background: var(--vault-color-action-tint);
  border: 1px solid var(--vault-color-line);
  border-radius: var(--vault-radius-panel);
  padding: var(--vault-space-24);
}
.empty-state--locked-panel .empty-state__lock-icon {
  display: block;
  width: 28px;
  height: 28px;
  margin: 0 0 var(--vault-space-12);
  color: var(--vault-color-action);
  vertical-align: baseline;
}
/* Disabled Subscribe button + "Coming soon" caption (DESIGN_BRIEF_2 item
 * 1): no :disabled rule existed anywhere in the app before this, because
 * no button had ever been shipped disabled on purpose. Opacity + a
 * not-allowed cursor is the one visual difference from a live
 * `.button--primary` - the shape stays identical, since this is the same
 * button once the subscribe surface (OPEN-A24) ships, not a different
 * component.
 *
 * **No longer scoped to `.empty-state`** (2026-10-30). The paginator ships
 * Previous and Next disabled at the ends of a list, which is §9's rule
 * ("disabled is for a destination that does not exist") - and with the
 * rule scoped elsewhere they rendered identical to the live control. A
 * button that looks live and does nothing is worse than one that is
 * absent, which is the whole reason §9 distinguishes the two. The
 * recipient picker's own send button, which JavaScript disables when
 * nobody can be sent to, was invisible in the same way and is not now. */
button:disabled {
  opacity: 0.5;
  cursor: not-allowed;
}
.empty-state__coming-soon {
  margin-left: 10px;
  font-size: 13px;
}
/* Failed-to-load (PARTNER_BRIEF 2026-09-23, item 1's fourth state): the
 * icon alone carries the distinction from Locked's own lock icon -
 * `--vault-color-attention` (the same amber this system already uses for
 * "needs attention", not a new red introduced just for this) rather than
 * the older `.error` box's red/bordered treatment, since this state is
 * informational-with-urgency, not a form-validation failure. */
.empty-state--failed .empty-state__lock-icon {
  color: var(--vault-color-attention);
}

/* Projected-winner row highlight (S5, new). nomination_review.html,
 * commit 13: deliberately its own modifier, not `.banner--success` -
 * this colours a single row inline in a list to mark "this one wins
 * right now", a different job from a page-level confirmation banner,
 * even though both borrow the same green. `.row.row--winner` (compound
 * selector) so it reliably overrides `.row`'s border regardless of
 * where in the file either rule is defined. Background is #f0fbf4 in
 * the source template, close enough to `success-tint` to alias rather
 * than carry a second near-identical literal (S1's "replace rather than
 * alias", same reasoning as the near-white greys in earlier commits). */
.row.row--winner {
  border-color: var(--vault-color-success);
  background: var(--vault-color-success-tint);
}

/* Table (S5, first real caller: players_index.html, 5.12 post-review).
 * S5's own definition is "desktop only; becomes cards at 768px" - only
 * the desktop half is built here. `.table-wrap` gives it a horizontally-
 * scrolling container instead, so a narrow viewport doesn't clip or
 * force page-level horizontal scroll (S7's actual mobile bar) without
 * yet building the real card-collapse treatment. */
/* Edge-scroll affordance (WEB_UI_GAP_LIST.md "Built and rough" #3,
 * 2026-10-06): .table-wrap already scrolled horizontally - nothing was
 * ever lost - but at 1440px, DESIGN_SPEC.md §7's own "coach and
 * director reference" width, the team dashboard's nine-column table
 * clipped two columns with no visible hint two more existed past the
 * edge. Fixed once here, not on the one page where it was noticed, so
 * every .table-wrap gets it. Pure-CSS scroll shadows (no JS): two
 * gradients scroll with the content and mask the fixed shadow layers
 * beneath them, so a shadow only shows on the side there's still
 * something to scroll to - at the left edge there's nothing left to
 * reveal leftward, so no left shadow renders, and the same at the
 * right edge once fully scrolled. background-color is repeated as its
 * own layer (not left to the parent .card) because a transparent gap
 * between the two content-scrolling gradients would otherwise show
 * through to whatever sits behind the table. */
.table-wrap {
  overflow-x: auto;
  background-image:
    linear-gradient(to right, var(--vault-color-surface) 30%, rgba(255, 255, 255, 0)),
    linear-gradient(to left, var(--vault-color-surface) 30%, rgba(255, 255, 255, 0)),
    linear-gradient(to right, rgba(0, 0, 0, 0.12), rgba(0, 0, 0, 0)),
    linear-gradient(to left, rgba(0, 0, 0, 0.12), rgba(0, 0, 0, 0));
  background-position: left, right, left, right;
  background-repeat: no-repeat;
  background-color: var(--vault-color-surface);
  background-size: 40px 100%, 40px 100%, 14px 100%, 14px 100%;
  background-attachment: local, local, scroll, scroll;
}
table.table {
  width: 100%;
  border-collapse: collapse;
}
table.table th,
table.table td {
  text-align: left;
  padding: 10px 12px;
  border-bottom: 1px solid var(--vault-color-line);
}
table.table th {
  font-size: 12px;
  font-weight: 600;
  text-transform: uppercase;
  letter-spacing: .04em;
  color: var(--vault-color-ink-faint);
}

/* PARTNER_BRIEF (2026-09-23): a video sizes itself to the source's
 * native resolution once metadata loads - _vault_item_detail.html's own
 * two <video> elements (the JS-hydrated one and the <noscript> fallback,
 * one shared partial for both the athlete's and the guardian's pages -
 * no CONF-39 duplication here) only ever had `width:100%` inline, which
 * bounds width but nothing bounds height. A portrait phone clip (shot
 * vertically from the sideline - the common case, not the edge case)
 * renders around 1920px tall at typical container widths and takes over
 * the page, confirmed live: correct before play, then jumps the moment
 * metadata loads and the browser applies the source's real dimensions.
 * max-height keeps it within the viewport regardless of source
 * orientation; object-fit: contain letterboxes a portrait clip inside
 * that box instead of cropping it - a width-only cap can't do either,
 * since height is what was blowing out. Native fullscreen (the
 * `controls` attribute's own browser-supplied control) is unaffected -
 * inline and contained is the right default, fullscreen stays on
 * request, nothing here disables it. */
.video-player {
  width: 100%;
  max-width: 100%;
  max-height: 70vh;
  object-fit: contain;
  border-radius: 10px;
  background: #000;
}

/* Groundwork, 2026-10-04 (Colin's own instruction): the shared layout
 * vocabulary's own page-header shape, backing _page_header.html. The
 * exact flex/gap/margin values this round's own pages
 * (teams_dashboard.html, seasons.html, organization_surveys.html,
 * team_detail.html, season_detail.html) each wrote inline, independently,
 * for the same "eyebrow/title/subtitle on the left, actions on the
 * right" shape - named once, as new classes, rather than converting any
 * of those pages to use them in this pass (DESIGN_SPEC.md §9's own "do
 * not restyle and restructure in the same pass" rule). */
.page-header {
  display: flex;
  justify-content: space-between;
  align-items: end;
  gap: var(--vault-space-24);
  margin-bottom: var(--vault-space-24);
  flex-wrap: wrap;
}
.page-header__actions {
  display: flex;
  gap: var(--vault-space-12);
  align-items: center;
  flex-wrap: wrap;
}
/* v2 (2026-10-09, docs/briefs/PAGE_HEADER_REVISION_PROPOSAL.md,
 * approved): the back link's own tap target. Deliberately not a
 * button - Colin's own decision, back_href/back_label's own doc
 * comment names why - so this adds padding only, no background or
 * border, inheriting the plain <a>'s default colour/underline every
 * other back link in the product already has. inline-block + negative
 * margin keeps the padding from shifting the text's own position
 * (nothing else sits above it to misalign against, but the convention
 * costs nothing and holds if that ever changes). ~40px tall including
 * line-height - short of the usual 44px guidance, a deliberate
 * trade-off for a secondary affordance that must still read as a link,
 * not a button, at a glance. */
.page-header__back {
  display: inline-block;
  padding: 10px 8px 10px 0;
  margin: -10px 0 0 -8px;
}

/* Groundwork, 2026-10-04: the shared form-field shape backing
 * _form_field.html - a label, a control (input/select/textarea), and
 * optional help/error text, one block instead of the hand-typed
 * `style="padding:9px;border:1px solid #cfd6e4;border-radius:8px;
 * font:inherit"` repeated inline on nearly every <input> in the product
 * (round one item 2's "need to rethink the base design for all of these
 * layouts" - a form is exactly one of those layouts). Deliberately its
 * own class, not an edit to the existing bare `select, textarea { ... }`
 * rule above (which also does not cover `input` at all, part of the
 * same gap) - widening that rule is a behaviour change on every
 * existing page the instant it ships; this is new, additive, and opt-in
 * until the overhaul that follows V1 converts a page onto it on
 * purpose. Also fixes a real, if minor, finding made while building
 * this: the inline style string being replaced used `#cfd6e4` as a
 * literal hex, a colour with no token - `--vault-color-line` is the
 * real token it was always meant to be. */
/* BEGIN GENERATED SHARED FIELD RULE - see website/generate_shared_css.py, do not edit by hand */
.field {
  display: block;
  margin-bottom: var(--vault-space-16);
}
/* END GENERATED SHARED FIELD RULE */
.field__label {
  display: block;
  margin-bottom: var(--vault-space-4);
}
.field__hint {
  color: var(--vault-color-ink-soft);
}
/* Component vocabulary proposal, Stage 3: the visual marker
 * _form_field.html's own required parameter never had - same error
 * token every other "something needs your attention" mark in this file
 * already uses, not a new colour invented for an asterisk. */
.field__required {
  color: var(--vault-color-error);
}
/* Border colour, 2026-10-17-18 (Colin), corrected one layer down from
 * where it was first found. 10-17 doubled this selector to
 * .field__control.field__control and set border here to beat the base
 * input[type="text"], ..., select, textarea rule's (0,1,1) specificity
 * for <input> - which proved the component had been fully inert for
 * every <input> element all along; the colour measured on inputs
 * before that fix was never coming from this component. That rule is
 * the one actually governing colour, padding, border-radius and font
 * for 203 of 266 covered-type elements app-wide, this component's own
 * dozen included, so 10-18 fixed border-color on the base rule itself
 * (also --vault-color-line, the same greenfield-pick mistake this
 * component made independently) and removed it from here along with
 * the three other properties - padding: 9px, border-radius: 8px,
 * font: inherit - already byte-identical to what the base rule gives
 * every covered element for free. Checked before stripping: every
 * current .field__control usage (the macro's own text/date/select/
 * textarea branches, season_rollover_draft.html's two selects,
 * vault_admin_organizations.html's three inputs) is on a type the base
 * rule covers, so none loses anything. _form_field.html documents an
 * uncovered pass-through kind (e.g. "url") with no current caller;
 * if one is added, it needs its own border/padding/radius/font here,
 * not a free ride on the base rule.
 *
 * The doubled selector existed only to beat the base rule on
 * border-color. With that property now agreeing with the base rule by
 * construction - and gone from this rule entirely - nothing remaining
 * (display, width, box-sizing, margin-top) is a property the base
 * rule sets, so there is no longer a specificity fight to win. Undone
 * back to a single class. */
.field__control {
  display: block;
  width: 100%;
  box-sizing: border-box;
  margin-top: var(--vault-space-4);
}
.field__help {
  display: block;
  margin-top: var(--vault-space-4);
}
.field__error {
  display: block;
  margin-top: var(--vault-space-4);
  color: var(--vault-color-error);
}

/* _survey_question.html's own .group-error - the error-message slot for
 * a checkbox group's client-side min/max validation (my_survey_detail
 * .html's own inline <script>). A component, not a capture (C76's own
 * test): one shape, one role, both occurrences - the JS only ever
 * writes el.style.display ('block'/'none'), never reads the attribute,
 * so starting hidden via this class rather than an inline style changes
 * nothing about how the script shows or hides it - confirmed by reading
 * both call sites before naming this. */
.field-error-hidden {
  color: var(--vault-color-error);
  display: none;
}

/* Groundwork, 2026-10-04: the generic "nothing here yet" shape backing
 * _simple_empty_state.html - deliberately separate from
 * _empty_state.html (the Vault manifest's own Empty/Locked/Suspended/
 * Failed state machine, CONF-33-shaped and not reusable outside it).
 * Most "no rows" states in the product (teams_dashboard.html's "No
 * active teams", seasons.html's "No seasons yet", roster_
 * administration.html's "Nobody is rostered yet") are a single muted
 * sentence, optionally with one link - this is that shape, named once,
 * on `.card` the same way every other populated state already renders. */
.simple-empty-state {
  text-align: left;
}
.simple-empty-state__action {
  display: inline-block;
  margin-top: var(--vault-space-12);
}

/* Groundwork, 2026-10-04: pairs the "N of M shown" count with the
 * row-limit control, backing _data_table.html's own toolbar() macro -
 * the one flex row seasons.html/access_feed.html/roster_
 * administration.html each wrote inline, independently, for the
 * identical pairing. */
.data-table-toolbar {
  display: flex;
  justify-content: space-between;
  align-items: baseline;
  gap: var(--vault-space-12);
  flex-wrap: wrap;
  margin-bottom: var(--vault-space-8);
}

/* ---- Inline-style centralization, Phase A (2026-10-08) -----------------
 *
 * CSS only - no template is edited in this pass. vaultapp/templates/
 * carries 2,343 inline style="..." attributes (154 files, including
 * instruction/, measured 2026-10-08); 16 contain a template expression
 * and are out of scope by construction (a literal utility class can't
 * stand in for a computed value). The ten classes below are the
 * highest-frequency *static* patterns - between them, roughly 930 of
 * the remaining ~2,327 occurrences, by exact-string match. Phase B (a
 * separate commit) applies these to the player console templates only,
 * as a bounded pilot; nothing else is touched yet.
 *
 * Captured exactly as found, not rounded to design/tokens.json's own
 * 4/8/12/16/24/32 spacing scale. An off-scale value quietly becoming
 * its nearest token is a visual change wearing a refactor's clothes,
 * and it would land everywhere this class is ever used at once, with
 * nobody having agreed to it - reconciling the scale is a separate
 * decision for Colin, not this pass, and it needs the full off-scale
 * census (this session's own report) to be made from. Two of these ten
 * (8px, 4px) already equal a real token exactly and reference it
 * directly below - that's recording an equivalence that already
 * holds, not rounding to one. Every other pixel value here is a
 * literal, each marked off-scale in place so the decision stays
 * visible at the call site rather than disappearing into a
 * plausible-looking class name. */
.u-mt-0 { margin-top: 0; }
.u-m-0 { margin: 0; }
/* C73 (docs/decisions/C73_SPACING_IS_A_SCALE_AND_TEMPLATES_HOLD_NO_VALUES.md,
 * 2026-10-16): the single source for page rhythm is design/tokens.json's
 * six-step spacing scale (4/8/12/16/24/32), and every utility class name
 * carries a step, never a value - "u-mb-18" cannot become 20px without
 * lying; "u-mb-lg" can (item 4). This block replaces every value-named
 * margin/padding/gap utility this file ever accumulated across five
 * separate "inline-style centralization" phases (2026-10-08 through
 * 2026-10-12) - each of those phases is now gone from this file, not
 * merely superseded in place, because a value-named class left standing
 * beside its step-named replacement is exactly the kind of live
 * contradiction a future reader cannot resolve by reading the file.
 *
 * u-mt-0 / u-m-0 / u-mb-0 above and below are the one exception: zero is
 * not a step (item 4's own "zero stays zero"), so they are unchanged,
 * not renamed.
 *
 * Three shapes item 5's own table does not name, resolved here (reported
 * to Colin, approved 2026-10-16):
 *   - u-m-6-0 was a 2-value shorthand (margin: 6px 0, both top AND
 *     bottom) wrongly grouped with the 3-value single-direction
 *     shorthands C73 lists as collapsing outright - it needs both
 *     u-mt-sm and u-mt-sm at its three call sites, not one class.
 *   - Negative margins (u-mt-neg6, u-m-neg4-0-0) are not rhythm at all -
 *     "a negative margin is a correction to a layout... rhythm is
 *     positive space between blocks" (Colin). Both are real, scale-
 *     exempt component corrections now named and defined where they
 *     belong: .section-subtitle (a muted one-line explainer directly
 *     under its own <h2>, confirmed by rendering to still pull the text
 *     up by exactly 6px, matching every one of its 15 call sites
 *     byte-for-byte) and .input-hint (a muted hint directly under a raw,
 *     unwrapped <input>, 3 call sites, -4px).
 *   - Two-axis padding (u-padding-10-4, u-padding-6-10, u-padding-7) is
 *     padding, not margin - "padding within a thing belongs to the
 *     thing" (Colin), so none of the three survive as utilities at all;
 *     each moved to the one component it was actually padding
 *     (.nav-item--compact, .table--admin-usage's own cell rule, and
 *     .field__control respectively - see each class's own site below).
 *     u-padding-8-4 had zero call sites and is deleted outright, not
 *     migrated. */
.u-mt-xs { margin-top: var(--vault-space-4); }
.u-mt-sm { margin-top: var(--vault-space-8); }
.u-mt-md { margin-top: var(--vault-space-12); }
.u-mt-lg { margin-top: var(--vault-space-16); }
.u-mt-xl { margin-top: var(--vault-space-24); }
.u-mt-2xl { margin-top: var(--vault-space-32); }
.u-mb-xs { margin-bottom: var(--vault-space-4); }
.u-mb-sm { margin-bottom: var(--vault-space-8); }
.u-mb-md { margin-bottom: var(--vault-space-12); }
.u-mb-lg { margin-bottom: var(--vault-space-16); }
.u-mb-xl { margin-bottom: var(--vault-space-24); }
.u-mb-2xl { margin-bottom: var(--vault-space-32); }
.u-mr-md { margin-right: var(--vault-space-12); }
.u-grid-gap-sm { display: grid; gap: var(--vault-space-8); }
.u-grid-gap-md { display: grid; gap: var(--vault-space-12); }
.u-grid-gap-lg { display: grid; gap: var(--vault-space-16); }
.u-grid-gap-xl { display: grid; gap: var(--vault-space-24); }
/* Each flex-gap name keeps its own alignment/wrap suffix (center / wrap /
 * start / end-wrap) - that part is layout, not a spacing value, and C73
 * item 4's "direction is kept because it is semantic" applies the same
 * way to these as to margin direction. */
.u-flex-gap-sm-center { display: flex; align-items: center; gap: var(--vault-space-8); }
.u-flex-gap-sm-wrap { display: flex; gap: var(--vault-space-8); flex-wrap: wrap; }
.u-flex-gap-md-wrap { display: flex; gap: var(--vault-space-12); flex-wrap: wrap; }
.u-flex-gap-md-center { display: flex; gap: var(--vault-space-12); align-items: center; }
.u-flex-gap-md-start { display: flex; gap: var(--vault-space-12); align-items: start; }
.u-flex-gap-md-end-wrap { display: flex; gap: var(--vault-space-12); align-items: end; flex-wrap: wrap; }
.u-flex-gap-2xl-wrap { display: flex; gap: var(--vault-space-32); flex-wrap: wrap; }
/* .section-subtitle / .input-hint - see the block comment above for why
 * these are component classes, not utilities. Values match every
 * pre-sweep call site exactly (verified rendered, CONF-43) - this is a
 * mechanical relocation of an existing value, not a redesign. */
.section-subtitle { margin-top: -6px; }
.input-hint { margin: -4px 0 0; }
/* parent_player_detail.html's own compact reuse of .nav-item (see that
 * class's own comment) - tighter padding than the sidebar's 0 12px, no
 * other difference. Replaces u-padding-10-4, which named a value, not
 * this component. */
.nav-item--compact { padding: 10px 4px; }
/* vault_admin_usage.html's own two identical tables (both tabs of /vault-
 * admin/usage) - replaces that page's own legacy-table-collapse, its
 * inline-styled second <table>, u-padding-6-10, and legacy-cell-left-6-10
 * / legacy-cell-right-6-10 all at once, since all five existed only to
 * describe these same two tables and splitting the fix across phase 1
 * and phase 3 would leave the page inconsistent in between. Text
 * alignment on the header cells uses u-text-left (already existed) and
 * u-text-right (new here, paired for the same reason u-text-left and
 * u-text-center already are). */
.table--admin-usage { width: 100%; border-collapse: collapse; }
.table--admin-usage th, .table--admin-usage td { padding: 6px 10px; }
.u-text-right { text-align: right; }
.u-text-center { text-align: center; }

/* C73 phase 2 (2026-10-16): the pure-spacing inline styles, same method
 * as phase 1's utility rename. Margin-left utilities, parallel to the
 * mt/mb set above - nothing used margin-left as a scale value before
 * this pass. */
.u-ml-xs { margin-left: var(--vault-space-4); }
.u-ml-sm { margin-left: var(--vault-space-8); }
.u-ml-md { margin-left: var(--vault-space-12); }
.u-ml-lg { margin-left: var(--vault-space-16); }
.u-ml-xl { margin-left: var(--vault-space-24); }
.u-ml-2xl { margin-left: var(--vault-space-32); }
.u-pt-0 { padding-top: 0; }

/* Negative margins, investigated individually per phase 1's corrected
 * hypothesis - each is a correction to a layout, not rhythm, so each
 * gets a named component class carrying the exact (or, where two call
 * sites measured 1-2px apart, snapped-to-one, same precedent as
 * .section-subtitle's own -6px) value rather than a utility step. */
.form-note {
  /* action_library.html / action_library_edit.html: a muted note
   * directly under a checkbox-group's closing </div>, pulled up to
   * sit with the group rather than at full paragraph rhythm. */
  margin-top: -8px;
}
.page-header-note {
  /* A note directly under _page_header.html's own render() output -
   * coaching_library.html and director_library.html measured -16px,
   * player_detail.html measured -18px for the same relationship;
   * snapped to one value rather than forking the class. */
  margin-top: -16px;
}
.card-lead-note {
  /* my_survey_detail.html: the first visible line inside a .card,
   * pulled up against the card's own top padding. Not merged with
   * page-header-note despite the similar magnitude - it compensates
   * for a different element (.card's padding, not _page_header.html's
   * trailing space), and C73's own negative-margin finding was that
   * these need individual investigation, not a shared assumption. */
  margin-top: -14px;
}
.list--indented {
  /* A plain <ul>'s own hanging indent - padding within the list, not
   * rhythm between it and its neighbours. evaluation_criterion_detail
   * .html already separated its own margin-top from this value before
   * this pass; account.html and vault_admin_organization_activity.html
   * (18px, snapped to 20px for the same reason as the negatives above)
   * had not. */
  padding-left: 20px;
}
/* Routed back by Instruction from the phantom-class guard's own two
 * remaining findings (ALLOWLIST entries removed there once this lands) -
 * both real, named components, not legacy debt, so they sit here rather
 * than in the "Captured components" block below. */
/* instruction/_queue.html's plan-entry row - three variants, currently
 * one inline style= with three independent conditions. Not one of
 * phase 2's pure-spacing cases (opacity and a border are not spacing)
 * and would otherwise wait for the 280-compound-styles phase; written
 * now rather than treated as an interrupt, per Colin (2026-10-16).
 * data-entry-id / data-state stay on the element - three test files
 * locate rows by them. */
.plan-entry {
  display: block;
}
/* .card .row { border: none } (above, Colin's own 2026-10-15 radius fix)
 * is a compound selector, specificity (0,2,0) - a single class loses to
 * it regardless of source order. Instruction found this by measuring
 * (computed border: 0px none, not the 3px solid this rule names) after
 * the DOM-diff check passed clean; C73 Decision 11 is amended for it
 * (2026-10-16). Fixed by matching the exact losing context rather than
 * narrowing the reset: .card .row.plan-entry--this-week is (0,3,0),
 * beats it, and says precisely which row the override is for. The
 * pattern for the next component that needs a border inside a card:
 * compound onto .card .row, don't touch the reset itself - a reset with
 * a growing :not() list is harder to read than each override owning its
 * own specificity. */
.card .row.plan-entry--this-week {
  border-left: 3px solid var(--vault-color-attention);
}
.plan-entry--settled {
  opacity: .7;
}
.plan-entry--collapsed {
  padding: 10px 16px;
}
/* instruction/dispatch.html's per-watch-login note. Deliberately not a
 * general small-note utility - every existing 12px rule in this file is
 * component-specific, and a general one is the type scale C73 Decision
 * 6 deferred (same family as u-fs-13/u-fs-14 above). Specific now; folds
 * into the general one when the type scale gets its own pass. */
.watch-login {
  font-size: 12px;
}
/* vault_admin_accounts.html's memberships table: four cells' padding,
 * each individually inline before this pass. A new, narrowly-scoped
 * class rather than touching that table's own legacy-table-collapse,
 * which is phase 3's - unlike vault_admin_usage.html in phase 1, this
 * table's legacy class is not itself part of what changed here, so it
 * is left alone. Values are an unaltered port of what was there: the
 * last column's own 4px 0 (not a mirror of the first column's 4px 8px
 * 4px 0) is preserved as found, not corrected. */
.table--accounts-memberships td:first-child { padding: 4px 8px 4px 0; }
.table--accounts-memberships td:not(:first-child):not(:last-child) { padding: 4px 8px; }
.table--accounts-memberships td:last-child { padding: 4px 0; }

/* C73 phase 3 (2026-10-17): the legacy-* classes. Two named components -
 * not drift, the opposite of it (C76): identical value, identical role,
 * every occurrence, confirmed by rendering both/all pages. */
.form-section-label {
  /* vault_admin_organizations.html / vault_admin_sports.html: a muted
   * mini-label directly before a <label> field group ("First admin",
   * "Engagement", "Scored question", "Open-notes question"). The -2px
   * compensates for <p>'s own default bottom margin in this one
   * relationship - a real mechanism, not a capture, so the value is
   * kept exactly as measured, not rounded. */
  margin: 6px 0 -2px;
}
.card-subheading {
  /* vault_admin_platform_content.html: an <h3> used as an intra-card
   * subsection break ("Promote an organization's template/criterion/
   * item"), after that card's own empty-state line and before its own
   * description text - same role at all three occurrences, rendered
   * and confirmed. h3 carries margin:0 in this file's own base rule
   * (above), so the top value compensates for nothing; 22 is a chosen,
   * unscaled value with the 6-step scale (4/8/12/16/24/32) sitting
   * right next to it - snapped to 24 (C76: spacing is a scale, and a
   * scale that exists is the one case a legacy- capture does get
   * snapped rather than kept literal). Rendered on all three occurrences
   * before and after; stated here per C76's one condition on this kind
   * of change. */
  margin: 24px 0 4px;
}

/* Ten new utilities, closed list (2026-10-17, approved) - missing
 * suffix/step variants of the u-flex-gap-*/u-grid-gap-* families above,
 * and four off-scale max-widths. A new composite name from here needs a
 * decision, not a judgement call at a call site - the cap is what stops
 * -baseline-wrap-between from quietly growing a -reverse sibling. */
.u-grid-gap-xs { display: grid; gap: var(--vault-space-4); }
.u-flex-gap-xl-between-start { display: flex; justify-content: space-between; align-items: flex-start; gap: var(--vault-space-24); }
.u-flex-gap-lg-end-wrap { display: flex; align-items: end; gap: var(--vault-space-16); flex-wrap: wrap; }
.u-flex-gap-md-baseline-wrap-between { display: flex; align-items: baseline; flex-wrap: wrap; gap: var(--vault-space-12); justify-content: space-between; }
.u-flex-gap-md-center-wrap { display: flex; align-items: center; flex-wrap: wrap; gap: var(--vault-space-12); }
.u-flex-gap-sm-end { display: flex; align-items: end; gap: var(--vault-space-8); }
/* Max-width is not a scale (C76) - no vocabulary of steps to snap any of
 * these four to, so each keeps its own literal value in its own name,
 * the same property .u-min-w-260 (above) already has. */
.u-max-w-420 { max-width: 420px; }
.u-max-w-480 { max-width: 480px; }
.u-max-w-520 { max-width: 520px; }
.u-max-w-560 { max-width: 560px; }
.u-max-w-760 { max-width: 760px; }

/* Four structural renames, no declaration changes - each already a real,
 * reusable shape (not a value-drift capture), just carrying a legacy-
 * name. Confirmed none land on an <input>/<select>/<textarea> (the
 * padding-vs-specificity risk documented below, in the Phase A2 block,
 * stays confined to that block). */
.table--collapse { width: 100%; border-collapse: collapse; }
.tab-strip { display: flex; gap: 4px; border-bottom: 2px solid var(--vault-color-line); margin-bottom: 20px; flex-wrap: wrap; }

/* C73 final phase (2026-10-17): .tab-strip's own tab items - confirmed
 * a real sibling pair (active/inactive), not a drifted capture: two
 * occurrences of each exact string, the same border-bottom-plus-
 * negative-margin-bottom mechanism (the 3px border and the -2px pull
 * it under the strip's own 2px border-bottom, closing the gap between
 * them) at every site, not a value nobody chose. Base carries the
 * inactive look, --active overrides it - same convention as
 * .nav-item/.nav-item--active above. */
.tab-link {
  padding: 10px 16px;
  text-decoration: none;
  font-weight: 400;
  color: var(--vault-color-ink-soft);
  border-bottom: 3px solid transparent;
  margin-bottom: -2px;
}
.tab-link--active {
  font-weight: 600;
  color: var(--vault-color-ink);
  border-bottom: 3px solid var(--vault-color-action);
}
.card-row {
  /* legacy-card-row-16-10-12: a card-row shell - border, radius,
   * padding and flex all travel together at every call site. Exact
   * port of the existing declaration (below, in the Captured
   * components block) - not reconstructed from the class name, which
   * names neither the 16px gap nor the 8px radius correctly on its own. */
  display: flex;
  align-items: start;
  justify-content: space-between;
  gap: 16px;
  border: 1px solid var(--vault-color-line);
  border-radius: 8px;
  padding: 10px 12px;
}
.bordered-box {
  padding: 14px 16px;
  border: 1px solid var(--vault-color-line);
  border-radius: 10px;
}
.auth-panel {
  /* legacy-centered-panel-56-440: the account/auth split-panel's own
   * centered box. Exact port - see that class's own comment below. */
  max-width: 440px;
  margin: 56px auto 0;
}
.flex1-minw220 {
  /* legacy-flex1-minw220: no padding, so none of the field-sizing
   * specificity risk the Phase A2 block below documents - a plain
   * layout shape, not a captured accident. */
  flex: 1;
  min-width: 220px;
}

/* The one form-input shape behind two inline spellings (105 combined
 * occurrences: 55 with box-sizing:border-box, 50 without). Verified
 * before unifying rather than assumed: `* { box-sizing: border-box }`
 * (top of this file) already applies globally, so the 50 instances
 * missing the explicit declaration were always rendering border-box
 * anyway - one real rendered box, two spellings, never a width
 * difference waiting to surface. Named as a component being captured,
 * not designed - deliberately not .field / .field__control
 * (_form_field.html's own reserved names, DESIGN_SPEC.md §11, for the
 * redesign this is explicitly not that) and deliberately called out as
 * legacy in its own name as well as this comment, so the eventual
 * restyle doesn't find a plausible-looking .field-input already
 * sitting in components.css and mistake it for a decision rather than
 * a capture. 9px padding, 6px top margin: both off-scale, both simply
 * what this input happens to do today. */
.legacy-field-input {
  margin-top: 6px; /* off-scale - captured, not endorsed */
  padding: 9px; /* off-scale - captured, not endorsed */
  width: 100%;
  box-sizing: border-box;
}

/* ---- Inline-style centralization, Phase A2 (2026-10-08) ----------------
 *
 * Phase A's twelve classes were derived from a count across the whole
 * app, before any template was edited. Applying them to five groups (55
 * template renders, 54 distinct files) showed the limit of that census:
 * coverage ran near-total on the lightest templates, then 52%, then
 * 39.6% on the densest group - twelve classes captured the app's most
 * common patterns, not the patterns these particular templates actually
 * carry. This set is derived the same way Phase A was, but from a fresh
 * census taken *after* Phase A, across the same 54 files, of every
 * style="..." string Phase A's classes don't already match (2,343 - 442
 * removed by groups 1-5 - the 1,452 Phase A already matched across the
 * rest of the app = the universe this count draws from is these 54
 * files' own remainder, 449 occurrences, 208 distinct strings).
 *
 * Cutoff is count >= 3, not a round number chosen by eye: the frequency
 * distribution itself breaks there - 11 patterns at count 3, a jump to
 * 46 at count 2, 122 at count 1. Below 3, a repeat is as likely to be
 * two templates coincidentally matching as it is a real shared pattern.
 * These 40 classes cover 235 of the 449 remaining occurrences (52.3%).
 * The other 214 (168 distinct strings, almost all count 1-2) are the
 * long tail this phase deliberately leaves alone - see the second pass's
 * own commit messages for the measured residue after applying this set.
 *
 * Same two rules as Phase A, unchanged: values are captured exactly as
 * found, never rounded to the nearest token: 12/8/16 below already equal
 * a real token and reference it directly; every other pixel value is a
 * literal, marked off-scale in place. And a captured component is named
 * and commented as captured, not designed - "legacy-" prefixed, so nothing
 * here reads as a decision to the redesign that hasn't been made. */

/* Phase A2's own margin-top and 3-value-shorthand utilities (the two
   sub-sections that used to sit here) are gone with the rest of the
   value-named spacing utilities this file carried - see C73's own block
   comment near the top of this file, where every one of them now
   lives under its step name instead. */

/* -- Other single/simple-property utilities --------------------------- */
.u-justify-self-start { justify-self: start; }
.u-w-max-content { width: max-content; }
.u-w-100 { width: 100%; }
.u-nowrap { white-space: nowrap; }
/* The link-styled-as-plain-text idiom - an <a> meant to inherit its
   surrounding text color and carry no underline. Mechanical, same as
   Phase A's own u-text-center; not a design decision about links. */
.u-link-plain { text-decoration: none; color: inherit; }
.u-text-left { text-align: left; }
.u-fs-13 { font-size: 13px; } /* off-scale - captured, not endorsed */
.u-fs-14 { font-size: 14px; } /* off-scale - captured, not endorsed */
.u-min-w-260 { min-width: 260px; } /* off-scale - captured, not endorsed */
.u-d-inline { display: inline; }

/* -- Captured components: compound, multi-declaration shapes that recur
   as one unit. Each is a real, verified-identical string match (the
   census that produced this file counts whole style="..." strings, so
   "6 occurrences" means 6 call sites with this exact combination, not 6
   similar-looking ones) - named for what it looks like, never for what
   it's for, so the redesign finds a description instead of a decision.

   C73 phase 3 (2026-10-17) renamed and relocated several of this
   block's own rules above (.table--collapse, .tab-strip, .card-row,
   .bordered-box, .auth-panel) once C76 established the test for doing
   that: one shape, one role, every occurrence, confirmed by rendering -
   not a captured accident. .legacy-flex-between-start-20 and
   .legacy-field-group-420 converted onto the step scale directly (both
   were grid/flex-gap-plus-margin shapes already covered by existing or
   newly-added step utilities), not kept as named components. */

/* A second, visually distinct text-input shape from Phase A's own
   .legacy-field-input - this one draws a visible border, radius and
   inherited font. Its border color was a literal hex (#cfd6e4) when this
   shape was first captured; tokenized 2026-10-08 (--vault-color-line-
   strong, design/tokens.json) rather than left as a literal, since a
   literal in shared, named vocabulary propagates - the app-wide spacing
   scale stays deliberately unreconciled for the same reason that's a
   bigger, still-open decision, but a single-value color duplicated
   across two classes was not. Two occurrences of this shape exist,
   differing only in margin-top (7px vs 10px, the latter also stating
   box-sizing:border-box explicitly, redundant under this file's own
   global reset) - kept as two classes rather than guessing they're
   interchangeable.

   Doubled-class selector, here and on every other Phase A2 class below
   that sets padding, border or border-radius on a field: this file's own
   base input/select/textarea rule (input[type="email"], ...) is an
   attribute selector, specificity (0,1,1) - higher than a single class
   (0,1,0). An inline style="..." always wins regardless (max specificity),
   which is exactly what hid this: converting one of these from an inline
   style to a plain class silently let the base rule's padding:9px/
   border-radius:8px win instead, on every field using this class -
   found only because the second-pass pixel diff caught forgot_password.
   html, reset_password_landing.html and guardian_reclaim_landing.html
   all rendering 11px padding as 9px. .x.x bumps specificity to (0,2,0),
   which beats (0,1,1) regardless of which of input/select/textarea the
   class lands on, without needing to know or duplicate the element type. */
.legacy-field-input-bordered-mt7.legacy-field-input-bordered-mt7 {
  display: block;
  width: 100%;
  margin-top: 7px; /* off-scale - captured, not endorsed */
  padding: 10px; /* off-scale - captured, not endorsed */
  border: 1px solid var(--vault-color-line-strong);
  border-radius: 8px; /* off-scale - captured, not endorsed */
  font: inherit;
}
.legacy-field-input-bordered-mt10.legacy-field-input-bordered-mt10 {
  display: block;
  width: 100%;
  box-sizing: border-box;
  margin-top: 10px; /* off-scale - captured, not endorsed */
  padding: 10px; /* off-scale - captured, not endorsed */
  border: 1px solid var(--vault-color-line-strong);
  border-radius: 8px; /* off-scale - captured, not endorsed */
  font: inherit;
}

/* A grid with a larger gap and top margin - 3 occurrences, both off-scale. */

/* A grid, wider gap, larger top margin - 3 occurrences, both off-scale. */

/* 2026-10-09 census, same second pass as the u-grid-gap-10/u-flex-gap-*
   additions above - the remainder of the layout+spacing bucket's own
   top repeats (docs/FINDINGS.md), each 4+ call sites, named like their
   .legacy-grid-N-mtM siblings above. */
.legacy-flex-gap-8-end-mt12 { display: flex; gap: var(--vault-space-8); align-items: end; margin-top: var(--vault-space-12); }

/* 2026-10-10 census, the third pass (docs/FINDINGS.md): the 3+ tranche
   measured against what remained after the first two. Two of these
   (legacy-field-input-block, legacy-field-input-w160) are the same
   .legacy-field-input shape this pass found missing a declaration
   (box-sizing, redundant under this file's own `* { box-sizing:
   border-box }`) or carrying one extra (display:block) or a fixed
   width instead of 100% - one original pattern that drifted three ways
   in the templates, not three different needs, named as siblings of it
   rather than as unrelated shapes. legacy-field-input-bordered-mt6
   joins -mt7/-mt10 the same way, for the identical reason. */
.legacy-mt4-fs13 { margin-top: 4px; font-size: 13px; } /* off-scale - captured, not endorsed */
.legacy-field-input-bordered-mt6.legacy-field-input-bordered-mt6 {
  display: block;
  width: 100%;
  box-sizing: border-box;
  margin-top: 6px; /* off-scale - captured, not endorsed */
  padding: 9px;
  border: 1px solid var(--vault-color-line-strong);
  border-radius: 8px; /* off-scale - captured, not endorsed */
  font: inherit;
}
.legacy-mt6-block { margin-top: 6px; display: block; } /* off-scale - captured, not endorsed */
.legacy-field-input-block { display: block; margin-top: 6px; padding: 9px; width: 100%; } /* off-scale - captured, not endorsed */
.legacy-field-input-w160 { margin-top: 6px; padding: 9px; width: 160px; } /* off-scale (160px) - captured, not endorsed */
.legacy-flex1-minw220 { flex: 1; min-width: 220px; } /* off-scale - captured, not endorsed */

/* A flexed field with a fixed minimum width - 3 occurrences, padding
   off-scale, min-width off-scale. Padding (9px) happens to equal the
   base input rule's own 9px, so the specificity gap above this block's
   own comment describes was invisible here by coincidence - doubled
   anyway, so it stays correct if either value ever changes. */
.legacy-flex-field-220.legacy-flex-field-220 { flex: 1; padding: 9px; min-width: 220px; } /* off-scale (9px, 220px) - captured, not endorsed */

/* A small metadata row - label-sized text, centered items, a tight gap -
   3 occurrences, both non-text values off-scale. */

/* A small chip/pill's own padding and text size - 4 occurrences, both
   off-scale. Doubled defensively, same reasoning as legacy-flex-field-220
   above - and unlike when that defensive doubling was first written,
   this one now does land on a form element: all four occurrences are
   <select> (access_feed.html x3, _row_limit.html), confirmed 2026-10-17.
   C76 amendment (decision 2 names the property, not an enumerated list
   of families): stays legacy-, same as legacy-field-input. */
.legacy-chip-4-6-13.legacy-chip-4-6-13 { padding: 4px 6px; font-size: 13px; } /* off-scale - captured, not endorsed */

/* A third bordered-input shape, wider padding, an asymmetric bottom
   margin - 3 occurrences, all three spacing values off-scale, border
   drawn from the real token this time (not the hex literal above). */
.legacy-bordered-input-11-6-20.legacy-bordered-input-11-6-20 {
  width: 100%;
  padding: 11px; /* off-scale - captured, not endorsed */
  margin: 6px 0 20px; /* off-scale - captured, not endorsed */
  border: 1px solid var(--vault-color-line);
  border-radius: 8px; /* off-scale - captured, not endorsed */
}

/* Phase 5 (2026-10-12): the 3+-occurrence patterns from a fresh scan of the
   552 remaining instances after phase 4 (docs/DESIGN_SPEC.md S11) - the
   same threshold phase 3 used. Every name here follows the existing
   value-based convention. legacy-progress-fill/-track, the one pair
   originally named for what it was rather than its values, was promoted
   to a real component (.progress-bar/.progress-bar__fill, 2026-10-15,
   docs/FINDINGS.md 2026-10-14) and all five of its callers converted -
   removed from here rather than left as a now-unused pair.
   legacy-mt6-pad9-w100 is doubled defensively, same reasoning as
   legacy-flex-field-220 above - it lands on <select> elements today
   (single-class specificity already beats a bare `select` tag selector,
   but not `input[type=...]`'s own attribute selector, so doubling costs
   nothing and removes the risk if a future caller moves it onto an
   input). */
.u-d-block { display: block; }
.legacy-box-border-mt6-w100 { box-sizing: border-box; margin-top: 6px; width: 100%; } /* off-scale - captured, not endorsed */
.u-whitespace-pre-line { white-space: pre-line; }
.legacy-grid-gap-8-mt10 { display: grid; gap: 8px; margin-top: 10px; } /* off-scale - captured, not endorsed */
.legacy-fw600-mt5 { font-weight: 600; margin-top: 5px; } /* off-scale (5px) - captured, not endorsed */
.u-cursor-pointer { cursor: pointer; }
.u-flex-1 { flex: 1; }
.legacy-fs13-mt3 { font-size: 13px; margin-top: 3px; } /* off-scale (3px) - captured, not endorsed */
.legacy-fs28-fw600 { font-size: 28px; font-weight: 600; } /* off-scale (28px) - captured, not endorsed */
.u-d-inline-block { display: inline-block; }
.legacy-grid-gap-14 { display: grid; gap: 14px; } /* off-scale (14px) - captured, not endorsed */
.legacy-grid-gap-2 { display: grid; gap: 2px; } /* off-scale (2px) - captured, not endorsed */
.legacy-flex-gap-10-center-wrap { align-items: center; display: flex; flex-wrap: wrap; gap: 10px; } /* off-scale (10px) - captured, not endorsed */
.legacy-flex-gap-12-baseline-wrap-between { align-items: baseline; display: flex; flex-wrap: wrap; gap: var(--vault-space-12); justify-content: space-between; }
.u-mb-0 { margin-bottom: 0; }
.legacy-block-mt12 { display: block; margin-top: var(--vault-space-12); }
.legacy-mt6-pad9-w100.legacy-mt6-pad9-w100 { margin-top: 6px; padding: 9px; width: 100%; } /* off-scale (6px) - captured, not endorsed; doubled, same defensive reasoning as legacy-flex-field-220 - padding could compete with the base select/input rule */
.legacy-flex-gap-6-wrap-mt8 { display: flex; flex-wrap: wrap; gap: 6px; margin-top: 8px; } /* off-scale (6px, 8px) - captured, not endorsed */
.legacy-m-6-0-0-pre-line { margin: 6px 0 0; white-space: pre-line; } /* off-scale - captured, not endorsed */
.legacy-m-4-0-0-24 { margin: 4px 0 0 24px; } /* off-scale - captured, not endorsed */

/* ---- Inline-style consolidation, Phase 7 (2026-10-22) -----------------
 * Punchlist 2.68. Phase 6 (2026-10-12) stopped with zero patterns at the
 * 3+ threshold and said to re-open only on a measurement finding the pool
 * had regrown from concurrent work. Measured 2026-10-22: 285 instances
 * across 101 files, 201 distinct patterns, 17 of them at 3+ covering 67
 * instances. The raw count FELL over those ten days (406 -> 285) while
 * the 3+ bucket went 0 -> 17, which is why DESIGN_SPEC says the count is
 * the wrong instrument.
 *
 * Same normalization phases 4-6 used, including phase 5's hex-to-token
 * widening, and the same value-based naming - two exceptions named below
 * where the shape is real and a value name would describe nothing.
 *
 * Four instances of `display:none` were NOT captured, deliberately. All
 * four (guardian_claims.html:175, _checkout_poll.html:31,
 * _vault_item_detail.html:169, staff_lookup.html:75) are toggled by JS
 * setting `style.display` directly, so a class would have to be removed
 * by that JS as well - a behaviour change, not the mechanical swap these
 * phases are. Left inline and recorded here so the next measurement does
 * not re-propose them. */
.u-fs-085em { font-size: 0.85em; } /* off-scale and relative - captured, not endorsed */
.u-fw-600 { font-weight: 600; }
.u-fw-400 { font-weight: 400; } /* normal weight inside something that sets 600 - a disclosure summary's own meta (2.305 (3)) */
.u-items-start { align-items: flex-start; }
.u-p-9 { padding: 9px; } /* off-scale - captured, not endorsed */
.u-min-w-280 { min-width: 280px; }
/* Three more of the same off-scale family, for roster.html's filter row
 * (punchlist 3.5). A filter control narrower than its own label wraps
 * the label, which is what these stop. */
.u-min-w-160 { min-width: 160px; } /* off-scale - captured, not endorsed */
.u-min-w-180 { min-width: 180px; } /* off-scale - captured, not endorsed */
.u-min-w-240 { min-width: 240px; } /* off-scale - captured, not endorsed */ /* off-scale - captured, not endorsed, same as .u-min-w-260 above */
.u-grid-2 { grid-template-columns: 1fr 1fr; }
.u-flex-gap-10 { display: flex; gap: 10px; } /* off-scale gap - captured, not endorsed */
.u-color-attention { color: var(--vault-color-attention); }
.u-color-action { color: var(--vault-color-action); }
/* 2.318 (g): the Team Actions table's numbers read in their column's colour,
 * "as text, never pills". The two that had no utility yet, beside the two
 * that did - same shape, same tokens the tiles and the ledger bar use, so a
 * column, its tile and its bar segment cannot be three different greens.
 * Design's file, added by API with the item that needs them (FINDINGS). */
.u-color-success { color: var(--vault-color-success); }
.u-color-error { color: var(--vault-color-error); }
/* A control stripped of its own box so a parent supplies one - the three
 * publish-preview pages put a <select> inside a bordered row. */
.u-reset-box { border: 0; padding: 0; min-width: 0; }
/* A list with no marker and no indent, which every board draws and no
 * class carried: .u-reset-box is a control stripped of its box, and
 * .team-grid is a grid that happens to include it. Three callers on
 * 2.305 (3)'s page alone. */
.list-plain { list-style: none; margin: 0; padding: 0; }
/* A section divided from the one above it by a rule rather than a card
 * edge. Two sizes because the product has two, 16 and 18, and snapping
 * them to one is a visible change this consolidation is not licensed to
 * make (DESIGN_SPEC: do not restyle and restructure in the same pass). */
.u-divide-top-16 { padding-top: var(--vault-space-16); border-top: 1px solid var(--vault-color-line); }
.u-divide-top-18 { padding-top: 18px; border-top: 1px solid var(--vault-color-line); } /* off-scale - captured, not endorsed */
/* Shape-named, not value-named: a dashed box is "nothing here yet, and
 * something could be", which is what all three callers use it for. */
.box-dashed { padding: 14px; border: 1px dashed var(--vault-color-line); border-radius: 8px; } /* off-scale radius - captured, not endorsed */
/* Shape-named for the same reason: the read-only box a recorded answer
 * sits in, on the three pages that show a submitted response. */
.value-box { padding: 10px 12px; background: var(--vault-color-page); border-radius: 7px; } /* off-scale - captured, not endorsed */
/* Joins the .legacy-field-input-bordered-mt{6,7,10} family above - the
 * same bordered control, with a minimum width instead of a top margin. */
.legacy-field-input-bordered-w220 {
  padding: 9px; /* off-scale - captured, not endorsed */
  border: 1px solid var(--vault-color-line-strong);
  border-radius: 8px; /* off-scale - captured, not endorsed */
  font: inherit;
  min-width: 220px; /* off-scale - captured, not endorsed */
}
/* The attention treatment on a panel, which .card--tone-attention does
 * not cover: that class is border-color alone, a different colour, and no
 * background. Not a .card modifier, because one of the three callers puts
 * it on a .row. */
.tone-attention-panel {
  border-color: var(--vault-color-attention);
  background: var(--vault-color-attention-tint);
}

/* A form's optional half, collapsed (C88 question 5, 2026-10-22).
 * "Required stays name and category... so a simple Action is two fields
 * and a save."
 *
 * A <details>, so it works with JS off and keyboard behaviour is the
 * browser's - the same choice .nav-group made for the My Children
 * collapse, and for the same reason: nothing here is worth a script.
 *
 * The marker is removed in both spellings browsers use and replaced by
 * a rotating caret on the summary itself, because the native triangle
 * sits outside the text box and makes the summary's own hit area
 * ambiguous next to a form label. */
.form-more {
  border-top: 1px solid var(--vault-color-line);
  padding-top: var(--vault-space-16);
}
.form-more__summary {
  cursor: pointer;
  list-style: none;
  font-weight: 600;
  display: flex;
  align-items: center;
  gap: var(--vault-space-8);
  min-height: 32px;
}
.form-more__summary::-webkit-details-marker {
  display: none;
}
.form-more__summary::before {
  content: "";
  width: 0;
  height: 0;
  border-left: 5px solid currentColor;
  border-top: 4px solid transparent;
  border-bottom: 4px solid transparent;
  transition: transform .12s ease;
}
/* Any <details> carrying the shared summary, not only a .form-more:
 * 2.305 (3)'s agreement documents are disclosures inside a .card rather
 * than a form's collapsed half, so they take the summary without the
 * top border and padding .form-more adds. Every .form-more call site is
 * itself a <details>, so this is strictly broader and changes none of
 * them. */
details[open] > .form-more__summary::before {
  transform: rotate(90deg);
}
@media (prefers-reduced-motion: reduce) {
  .form-more__summary::before { transition: none; }
}

/* The shared recipient picker (C91, 2026-10-22). One row per person,
 * checkbox and name, the checkbox aligned to the first line so a row
 * whose annotation wraps to two lines does not float its box to the
 * middle. Captured from organization_survey_detail.html's own inline
 * style="display:flex;align-items:flex-start;gap:7px", which is the
 * only place this shape existed before it was shared. */
.recipient-row {
  display: flex;
  align-items: flex-start;
  gap: 7px; /* off-scale - captured, not endorsed */
}
/* The list scrolls rather than growing without bound: a roster of
 * eighty-five is a page of checkboxes otherwise, and the send button
 * below it would be off-screen on every visit. */
.recipient-list {
  max-height: 280px; /* off-scale - captured, not endorsed */
  overflow: auto;
}
/* 2.246 (Colin, 2026-10-28): the folded variant. The scroll box above
 * exists because an unbounded roster puts the send button off-screen;
 * where the caller folds instead, there is nothing to scroll, so the cap
 * and the scrollbar both go.
 *
 * auto-fill with a min track, never `repeat(3, …)`: at 390px a fixed
 * three-column grid is exactly the one-word-per-line squeeze 2.242 fixed
 * on the console tiles, and auto-fill drops to one column on its own
 * without a media query to keep in step. */
.recipient-list--grid {
  display: grid;
  grid-template-columns: repeat(auto-fill, minmax(220px, 1fr));
  gap: var(--vault-space-8);
  max-height: none;
  overflow: visible;
}


/* 2.276 (Colin, 2026-10-29), docs/boards/child/Main.dc.html: the child's
   page below Needs Attention is two columns - the record and what is
   still out with the coach on the left, Recent Activity and About this
   record on the right.

   Flex rather than grid, with the basis doing the phone layout: at 390px
   neither column can reach its basis, so both take a full row and the
   order is the DOM's - the record first, which is what a reader opened
   the page for. No media query, so there is no second layout to keep in
   step with this one (S9's phone rule: the same purpose, a different
   form, not a different page). */
.child-columns {
  display: flex;
  flex-wrap: wrap;
  gap: var(--vault-space-24);
  align-items: flex-start;
}
.child-columns__main {
  /* 420, not the board's 620: the board is a 1320px artboard with no
     sidebar, and the app's content column at that width is about 890px
     once the nav and the page gutters are taken out. At 620 the two
     bases plus the gap came to 964 and the aside wrapped underneath -
     which rendered as one column and read as the rule not working.
     Measured from the rendered page, not from the board's own numbers. */
  flex: 999 1 420px;
  min-width: 0;
}
.child-columns__aside {
  flex: 1 1 300px;
  min-width: 0;
}

/* 2.276 (Colin, 2026-10-29), docs/boards/child/Main.dc.html: the child's
   page draws its tiles as a big left-aligned number over a label over a
   sub-label, in a tint with a coloured border - not the small centred
   ones the staff consoles use. A modifier on the row rather than a
   second tile component, so the three tints stay one definition. */
/* 2.283: all but one of this modifier's declarations became the default
   when the kit took the boards' own values, so they are gone from here -
   a value defined twice is a value that drifts. What is left is the one
   real difference between two boards: the child's page draws a 30px
   count (docs/boards/child/Main.dc.html) where the director's draws 28
   (docs/boards/director/Main.dc.html). Unifying them is Colin's call,
   not a stylesheet's. */
.tile-row--board .tile__count {
  font-size: 30px;
}

/* ------------------------------------------------------------------ *
 * 2.283's kit (Colin, 2026-10-29). Every value is taken from a board
 * in docs/boards/, so five builders copying a board land on the same
 * page. The pieces above - .tile, .card, .badge, the buttons, .table,
 * .sidebar - are the same kit; only what the boards drew differently
 * changed. These three are what the boards have and the stylesheet
 * did not.
 * ------------------------------------------------------------------ */

/* The lede under a page's h1. 70ch because a board's own measure is
 * 70ch (docs/boards/director/A-Teams.dc.html and three others). */
.page-lede {
  margin: 6px 0 0;
  font-size: 15px;
  color: var(--vault-color-ink-muted);
  max-width: 70ch;
}

/* The phone form of a table (2.280 (3), from §9's rule of 2026-10-21):
 * "a dense row per item - one line of identity, one of state, no card
 * chrome". The default .table turns each cell into its own labelled
 * block, which is the stack of cards that rule rejects; .table--dense
 * keeps a real table, which the same sentence rejects too. This is the
 * third form, opted into per table.
 *
 * The first cell is the identity line. Every other cell joins one state
 * line, separated by a middot, with no column name printed - the whole
 * point is that a reader does not need the header row to read a name
 * and a state. */
@media (max-width: 767px) {
  .table:not(.table--dense) thead , .table--rows thead {
    display: none;
  }
  .table:not(.table--dense), .table--rows,
  .table:not(.table--dense) tbody, .table--rows tbody,
  .table:not(.table--dense) tr, .table--rows tr,
  .table:not(.table--dense) td , .table--rows td {
    display: block;
    width: 100%;
  }
  .table:not(.table--dense) tr , .table--rows tr {
    padding: 10px 0;
    border-bottom: 1px solid var(--vault-color-line-soft);
  }
  .table:not(.table--dense) td , .table--rows td {
    padding: 0;
    border: 0;
  }
  .table:not(.table--dense) td::before , .table--rows td::before {
    content: none;
  }
  .table:not(.table--dense) td:first-child , .table--rows td:first-child {
    font-weight: 600;
  }
  .table:not(.table--dense) td:not(:first-child) , .table--rows td:not(:first-child) {
    display: inline;
    font-size: 13px;
    color: var(--vault-color-ink-muted);
  }
  .table:not(.table--dense) td:not(:first-child):not(:nth-child(2))::before , .table--rows td:not(:first-child):not(:nth-child(2))::before {
    content: " · ";
  }
  .table:not(.table--dense) td:empty , .table--rows td:empty {
    display: none;
  }
}

/* One form on a phone, the other on a wide screen (2.324 (l), Colin
 * 2026-11-15; `docs/boards/actions-2324l/AllTeams-A-Phone.dc.html`).
 *
 * The kit's dense row prints no column name, on the premise that a reader
 * does not need one to read a name and a state - true of "no coach", and
 * false of five bare numbers: "22 · 5 · 8 · 3 · 2" says nothing without the
 * header row a phone has no room for. So the board gives the phone its own
 * line, where each number carries its word, the ones that need the reader
 * come first, and a zero is left out rather than printed grey.
 *
 * Ordering and omission are decisions about meaning, so they are made in
 * the template where they can be tested, and these two utilities only
 * choose which form is shown. Design's file, added by API with the item
 * that needs them, as the colour utilities above were (FINDINGS). */
.u-phone-only { display: none; }
@media (max-width: 767px) {
  .u-phone-only { display: block; }
  .u-wide-only { display: none; }
  /* **A table cell needs the longer selector.** The dense row above sets
   * `display: block` on every `td` of the tables it governs, at a higher
   * specificity than a bare utility class, so `.u-wide-only` alone lost to
   * it and the page drew both forms at once - measured, not reasoned, on
   * `docs/screens/club-actions-all-teams-phone/compare`. Scoped rather
   * than given `!important`, which would win everywhere and be the next
   * person's puzzle. */
  .table:not(.table--dense) td.u-wide-only, .table--rows td.u-wide-only { display: none; }
}

/* A closed pillar's lock in the sidebar (CONF-46): the row stays, says
 * why it is closed to a screen reader, and carries the board's own
 * 12px lock at the right. */
.nav-item__lock {
  margin-left: auto;
  flex: 0 0 auto;
}
.nav-item--locked {
  color: var(--vault-color-ink-soft);
}

/* The club switcher above the nav (C99 A, Colin 2026-10-29 -
 * docs/boards/c99/Main.dc.html). Every value is the board's. Rendered only
 * where the viewer has staff standing in more than one club, so for almost
 * everybody this rule set never matches anything.
 *
 * The summary is 44px minimum: it is a control somebody taps on a phone,
 * and the board sets that height explicitly rather than leaving it to the
 * text. `list-style: none` on both the element and ::-webkit-details-marker
 * removes the disclosure triangle, which the board does not draw - "Change"
 * is the affordance. */
.club-switcher {
  margin: 10px 12px 0;
}
.club-switcher > summary {
  display: flex;
  gap: var(--vault-space-8);
  align-items: center;
  padding: 9px 12px;
  border: 1px solid #CFD6E4;
  border-radius: 10px;
  cursor: pointer;
  min-height: 44px;
  box-sizing: border-box;
  list-style: none;
}
.club-switcher > summary::-webkit-details-marker {
  display: none;
}
/* The board wraps the label and the name in one span, so the summary lays
 * out two children - that pair, and Change - rather than three in a row.
 * `min-width: 0` lets a long club name wrap inside it instead of pushing
 * Change off the edge. */
.club-switcher__who {
  min-width: 0;
}
.club-switcher__label {
  display: block;
  font-size: 11px;
  letter-spacing: .07em;
  text-transform: uppercase;
  color: var(--vault-color-ink-muted);
  font-weight: 600;
}
.club-switcher__name {
  display: block;
  font-size: 13.5px;
  font-weight: 600;
}
.club-switcher__change {
  margin-left: auto;
  color: var(--vault-color-ink-muted);
  font-size: 12px;
}
.club-switcher__options {
  margin-top: 6px;
  border: 1px solid var(--vault-color-line);
  border-radius: 10px;
  overflow: hidden;
}
/* Each option is a submit button rather than a link, because choosing is a
 * write - so these undo the button chrome the boards draw as a plain row. */
.club-switcher__option {
  display: block;
  width: 100%;
  text-align: left;
  padding: 10px 12px;
  font-size: 13.5px;
  color: var(--vault-color-ink-muted);
  border: 0;
  background: none;
  cursor: pointer;
}
.club-switcher__option + .club-switcher__option {
  border-top: 1px solid var(--vault-color-line-soft);
}
.club-switcher__option[aria-current="true"] {
  background: var(--vault-color-action-chip);
  color: var(--vault-color-action);
  font-weight: 600;
}
.club-switcher__note {
  margin: 8px 2px 0;
  font-size: 12.5px;
  color: var(--vault-color-ink-muted);
}

/* Choosing a club (C99 A - docs/boards/c99/Choose.dc.html). The board makes
 * the whole card the control, so this is a submit button wearing a card: a
 * POST, because the choice is audited either way, with none of the button
 * chrome. "Work in this club →" is a blue label and not a filled button,
 * which is what keeps one page from carrying one blue button per club. */
.club-choice {
  display: block;
  width: 100%;
  text-align: left;
  background: var(--vault-color-surface);
  border: 1px solid var(--vault-color-line);
  border-radius: var(--vault-radius-card);
  padding: 20px 22px;
  cursor: pointer;
  font: inherit;
  color: inherit;
}
.club-choice__head {
  display: flex;
  flex-wrap: wrap;
  gap: var(--vault-space-12);
  align-items: baseline;
  justify-content: space-between;
}
.club-choice__name {
  font-size: 17px;
  font-weight: 700;
}
.club-choice__go {
  font-size: 13.5px;
  font-weight: 600;
  color: var(--vault-color-action);
}
.club-choice__meta {
  display: block;
  font-size: 13.5px;
  color: var(--vault-color-ink-muted);
  margin-top: 4px;
}

/* The destination at the right of a console row (director/Main.dc.html).
 * `.row--link` already puts it there with space-between; this is its blue,
 * and `flex: 0 0 auto` so a long sentence on the left wraps rather than
 * squeezing the label. */
.row__go {
  flex: 0 0 auto;
  font-size: 13px;
  color: var(--vault-color-action);
}

/* The empty-teams grid on Rosters (director/A-Rosters.dc.html). A short
 * line per team rather than a table: every one of them is in the same
 * state, so a row per item with two columns repeating "No season set" and
 * "No coach" spends several hundred pixels saying one thing. `auto-fill`
 * makes it three columns on a desktop and one at 390px with no second
 * rule. */
.team-grid {
  list-style: none;
  margin: 0;
  padding: 0;
  display: grid;
  grid-template-columns: repeat(auto-fill, minmax(260px, 1fr));
  gap: 2px 24px;
}
.team-grid__item {
  display: flex;
  justify-content: space-between;
  align-items: baseline;
  gap: var(--vault-space-12);
  padding: 7px 0;
  border-top: 1px solid var(--vault-color-line-soft);
}

/* 2.306 (coordinator, under §9: "a link to a page highlights the item it was
 * about"). The fragment does all of it: `#claim-<id>` scrolls the row into
 * view and `:target` tints it, so a page whose rows carry ids gets the
 * behaviour by being linked to one - no query parameter, no script, and
 * nothing to keep in step on the sending side but the id.
 *
 * The action tint, which is what every other "the system is pointing at this"
 * surface uses; it is a cue about how the reader arrived and never a state of
 * the row, so it carries no border and no weight. The cells rather than the
 * row, because a <tr> takes no background of its own once its cells have one. */
[id]:target > td,
[id]:target > th { background: var(--vault-color-action-tint); }

/* 2.307 (1) (Colin, 2026-11-05): "the Needs Attention tile keeps its jump to
 * the section on the same page, and the section it lands on is highlighted".
 * What he hit was the jump moving the page barely a screen, so nothing
 * appeared to happen - the anchor worked and the arrival was invisible.
 *
 * **A ring, not the tint the rows above use.** The tint is #F3F7FF and the
 * page behind a card is #F5F8FC, so a tinted card is *less* distinguishable
 * from the page than the white one it replaces - the row rule works because
 * cells sit on a card's own white. The chip-strong blue is the one token in
 * this family that reads against both surfaces.
 *
 * **Not the focus ring's outline**, which is the same action blue at the same
 * offset and means something else entirely (keyboard focus); two cues that
 * look alike for two different facts is worse than one that is only similar.
 *
 * `scroll-margin-top` is half of it: a card jumped to flush against the top
 * of the viewport reads as a page that reloaded rather than scrolled. Any
 * card with an id gets this by being linked to - the console's three sections
 * are the callers today (`_console_buckets.html`), and nothing is per-page. */
.card[id]:target {
  box-shadow: 0 0 0 3px var(--vault-color-action-chip-strong);
  scroll-margin-top: var(--vault-space-24);
}

/* 2.306 (Colin, 2026-11-04): "a blue band for a running team, orange for one
 * waiting on a coach". Teams groups its teams by the state each one is in
 * (web.py's `team_groups`) and the four groups read identically - the state is
 * in the heading's words and nowhere else. The band carries the group's colour
 * down the left edge of its heading, its note and its rows together, so which
 * state a block of teams is in is visible before a word of it is read.
 *
 * Two colours, because the decision names two: the club's work is running, or
 * it is waiting on somebody. The other two groups keep the neutral line - a
 * band that says nothing is owed, rather than no band and a block that looks
 * like it belongs to the one above it. */
/* 2.315 (g) (Colin, 2026-11-10): "the light border every other page uses, not
 * the left-hand vertical rule". It was a 3px coloured rule down the left of
 * each group (2.306), blue for running and orange for waiting on a coach -
 * a second border vocabulary on a page already inside a card, and a colour
 * carrying a meaning the heading above it already says in words. Now it is
 * the card's own border, so a group on this page looks like a block on every
 * other one. */
.group-band {
  background: var(--vault-color-surface);
  border: 1px solid var(--vault-color-line);
  border-radius: var(--vault-radius-panel);
  padding: 16px 18px;
  margin-bottom: var(--vault-space-16);
}

/* 2.306 (Colin, 2026-11-04): "Needs a coach" on one line however the name
 * wraps. The state is the shorter half and the name is the half that should
 * take the extra line, so the state neither shrinks nor wraps - .row__go's own
 * reasoning above, for the same shape at the other end of a row. */
.team-grid__state {
  flex: 0 0 auto;
  white-space: nowrap;
}

/* 2.315 (h) (Colin, 2026-11-10): one team per row rather than the auto-fill
 * columns above. A team waiting on a coach carries its own Add a coach, and a
 * 260px column is not wide enough to hold a name and a control without the two
 * colliding on a long name. Same row, same rule, one column. */
.team-grid--rows {
  grid-template-columns: 1fr;
}

/* Design 11/12 (FINDINGS 2026-10-30), docs/boards/actions/Family-A.dc.html:
   a family Action is a row, not a table row - a title with its pills
   inline, the dates at the right, and one button whose label is the next
   step. The board draws no column headers at all.

   At phone width it stacks, which is the same purpose in a different
   form (S9): there is no horizontal room to recover, so the dates and
   the button fall under the title rather than squeezing beside it. */
.family-action-row {
  display: flex;
  flex-wrap: wrap;
  gap: var(--vault-space-12);
  align-items: baseline;
  padding: var(--vault-space-12) 0;
  border-top: 1px solid var(--vault-color-line-soft);
}
.family-action-row__what {
  /* 340, measured rather than chosen. The board has 1320px and no
     sidebar; this page has about 850 inside the card, and 420 + the
     dates + the button is 854 - so widening this to keep a title's two
     chips on one line pushed the *button* onto a line of its own, on
     every row. The button at the right is the board's structure and the
     chip wrap is one row's cosmetics, so 340 stays and the wrap is a
     consequence of the sidebar, like the guardian page's height. */
  flex: 1 1 340px;
  min-width: 0;
}
.family-action-row__dates {
  flex: 0 0 auto;
  font-size: 13px;
}
/* The disclosure reads as a link, which is what the board draws - a
   blue "What to do ▾" under the title, not a grey row of its own. */
/* The coach's redo note, as the board draws it (actions/Family-A.dc.html):
   a tinted block under the title, not a line of body text. It is *why*
   there is work to do, so it reads as an instruction rather than as more
   of the row's own description. The attention tint, because the row it
   sits in is one the family is behind on. */
.family-action-row__returned {
  margin-top: var(--vault-space-8);
  padding: var(--vault-space-8) var(--vault-space-12);
  background: var(--vault-color-attention-tint);
  border-radius: var(--vault-radius-panel);
  font-size: 13px;
}
.family-action-row__more > summary {
  margin-top: 5px;
  color: var(--vault-color-action);
  font-size: 13px;
  cursor: pointer;
  /* The board draws the caret *after* the label - "What to do \25BE" -
     where the browser's own marker sits before it and points right. The
     default marker is removed and the caret written on, which also lets
     it flip when the disclosure opens. Both the modern property and the
     WebKit pseudo-element, because Safari still reads only the latter. */
  list-style: none;
}
.family-action-row__more > summary::-webkit-details-marker { display: none; }
.family-action-row__more > summary::after { content: " \25BE"; }
.family-action-row__more[open] > summary::after { content: " \25B4"; }

/* The paginator (2.285, Colin 2026-10-30). One page of rows, the count,
 * and the two ways through - under the table it belongs to. The count sits
 * at one end and the controls at the other on a desktop, and they stack at
 * phone width rather than squeezing, which is the same answer every other
 * toolbar here gives. */
.paginator {
  display: flex;
  flex-wrap: wrap;
  align-items: center;
  justify-content: space-between;
  gap: var(--vault-space-12);
  margin-top: var(--vault-space-12);
}
.paginator__count {
  margin: 0;
  font-size: 13px;
  color: var(--vault-color-ink-muted);
}
.paginator__controls {
  display: flex;
  align-items: center;
  gap: var(--vault-space-12);
}
.paginator__where {
  font-size: 13px;
  color: var(--vault-color-ink-muted);
}

/* The help ? (2.310 and §9, Colin 2026-11-07: "It should be obvious
 * what's going on from the functionality without someone needing to read
 * an in-line user manual"). A section's explanation is no longer prose
 * above its numbers; it sits behind a small ? beside the heading, closed
 * until the reader asks for it.
 *
 * A <details>, so it works with JS off and the keyboard behaviour is the
 * browser's - the same choice .nav-group and .form-more already make,
 * and for the same reason.
 *
 * **The heading is the summary's own content.** <summary> takes phrasing
 * content *or* exactly one h1-h6, so the ? cannot be a sibling of the
 * heading inside it: it is a <span> inside the heading, which is valid
 * and makes the whole heading row the tap target - 44px at phone width
 * without a second rule.
 *
 * **The body opens below the heading, at the container's full width, and
 * pushes what follows down.** The alternative - a panel floating over
 * the page from the ? - covers the numbers the reader came for, which at
 * 390px is most of the screen. That is also why the body is not laid out
 * beside the mark: a 20px-wide column is not a paragraph.
 *
 * The summary's accessible name is the heading's own text, so the
 * disclosure announces the section it explains ("Rounds in flight,
 * collapsed"); .help__mark is aria-hidden because a question mark read
 * aloud says nothing. */
.help__summary {
  cursor: pointer;
  /* Both spellings, because Safari still reads only the WebKit one. */
  list-style: none;
  /* Full width and 32px tall, so the tap target is the whole heading
   * row rather than the 20px mark - the same shape, and the same
   * number, .form-more__summary already uses for a disclosure in a
   * card. Padding with a cancelling negative margin was the other
   * candidate and it moves the ink: the heading would sit 10px nearer
   * whatever precedes it than the utility the caller passed says. */
  display: flex;
  align-items: center;
  min-height: 32px;
}
.help__summary::-webkit-details-marker {
  display: none;
}
/* The mark, inside the heading. An outlined circle rather than a filled
 * one: it is a way to ask, not an action, and §9's one blue button on a
 * page is not this. em units so it tracks the heading it sits beside
 * rather than needing a size per level. */
.help__mark {
  display: inline-flex;
  align-items: center;
  justify-content: center;
  width: 1.3em;
  height: 1.3em;
  margin-left: var(--vault-space-8);
  border: 1px solid var(--vault-color-line-strong);
  border-radius: var(--vault-radius-pill);
  background: var(--vault-color-surface);
  color: var(--vault-color-ink-soft);
  font-family: inherit;
  font-size: 0.8em;
  font-weight: 700;
  line-height: 1;
  vertical-align: 0.12em;
}
.help__summary:hover .help__mark,
.help__summary:focus-visible .help__mark {
  border-color: var(--vault-color-action);
  color: var(--vault-color-action);
}
/* Open: the mark reads as the thing that is currently showing something,
 * so it carries the action tint rather than changing its glyph - a ? that
 * becomes an x asks the reader to work out which one closes it. */
.help[open] > .help__summary .help__mark {
  border-color: var(--vault-color-action-chip-strong);
  background: var(--vault-color-action-chip);
  color: var(--vault-color-action);
}
/* 2.319 (i): a page heading needs the space under it that `.page-header` has,
 * and takes the same token rather than a second value that can drift from it.
 * Only the h1 form - an h2 or h3 help heading is a section inside a card. */
.help--page {
  margin-bottom: var(--vault-space-24);
}
.help__body {
  margin-top: var(--vault-space-8);
  padding: var(--vault-space-12);
  border: 1px solid var(--vault-color-line);
  border-radius: var(--vault-radius-panel);
  background: var(--vault-color-page);
  /* A measure, not the container's width: an explanation is prose, and
   * prose at 1320px wide is unreadable. */
  max-width: 62ch;
  font-size: 13.5px;
  line-height: 1.5;
  color: var(--vault-color-ink-muted);
}
.help__body > :first-child { margin-top: 0; }
.help__body > :last-child { margin-bottom: 0; }

/* The team drop's preview strip (`C110` D16, Colin 2026-11-09: the club
 * staff's page "shows thumbnails, so a club's media staff can pick from what
 * they see"), from the board's own markup
 * (docs/boards/c110/staff-drop/Main.dc.html).
 *
 * A component rather than three utilities: the tile has a fixed size the
 * board draws, and a 4px gap no utility carries. The size is the board's
 * 52x40 - small enough that a game day stays one table row, which is the
 * whole reason the page can show a club's worth of media at all. */
.media-strip {
  display: flex;
  align-items: center;
  gap: var(--vault-space-4);
}
.media-tile {
  width: 52px;
  height: 40px;
  flex: 0 0 auto;
  border: 1px solid var(--vault-color-line-strong);
  border-radius: 6px;
  background: var(--vault-color-page);
  object-fit: cover;
}
/* A video has no still: no VideoProvider thumbnail call exists, so the tile
 * says what it is rather than drawing an empty frame that reads as a broken
 * image (the same choice the family's own drop made). */
.media-tile--video {
  display: flex;
  align-items: center;
  justify-content: center;
  font-size: 11px;
  color: var(--vault-color-ink-soft);
}
