/* ══════════════════════════════════════════════════════════════════════════════════════════════
   402 SHARED·PROVENANCE — the reviewer + source trust footrow. (D-439 · D-465 · D-585)
   ──────────────────────────────────────────────────────────────────────────────────────────────
   EXTRACTED 2026-08-16 (Andrew: "we need the source and reviewed by"). Design-locked 2026-07-15;
   the styles lived provisionally on workbench/shared/provenance.html and every surface that wanted
   a footrow rolled its own — the collection reader had `.crd-prov`, the technique player had
   `.tqp-prov` with DIFFERENT copy ("Source: X" with a prefix the design does not have, and a
   reviewer with no shield and no "Reviewed by"). Two lookalikes of one component, drifting
   already. This file is the one definition; surfaces emit `.prov-*` and stop deciding.

   THE RULES IT CARRIES, none of them cosmetic:

   · SOURCE bottom-LEFT (theirs) · REVIEWER bottom-RIGHT (ours), and NEVER fused (D-439/D-465).
     Fusing them reads as a source ENDORSEMENT — "the NHS says our paediatrician approved this" —
     which is a fabrication, and it is the shape a well-meaning one-liner naturally takes.
   · NO DATE on the card (D-451). Methodology lives once, in About (405).
   · 🔴 RENDERS FROM DATA OR NOT AT ALL (D-560). No fields ⇒ no row. Never a stamped badge.

   ⚠️ THE HISTORY THIS EXISTS TO PREVENT IS SPECIFIC AND RECENT. D-541: the technique player
   rendered ONE hardcoded page for EVERY technique and stamped it "Source: NHS" twice and
   "Reviewed by a paediatrician" — on content nobody had read. Tap "How to swaddle safely", get the
   reflux video, credited to a clinician who had never seen it. That is why the absent state is the
   default and why nothing here has a fallback: in a health app a FALSE TRUST CLAIM IS WORSE THAN
   NONE, and it is worse in a way that cannot be walked back after it ships.

   ⚠️ NOT IN THIS FILE: `.prov-none`, the dashed gold "missing provenance" flag. That is a
   WORKBENCH-ONLY defect marker and it stays on the WB page. The app renders nothing here, on
   purpose — a published unit missing provenance is caught and fixed BEFORE live, not annotated
   in front of a parent.

   ⚠️ AND NO ITEM IN THE CORPUS CARRIES `source` OR `reviewer` TODAY, so on real data this file
   renders NOTHING on every surface. That silence is the honest state, not a gap in the CSS. The
   fields are D-703's, they are Andrew-authored, and writing one without a real review is the
   fabrication D-703 exists to stop.
   ══════════════════════════════════════════════════════════════════════════════════════════════ */

/* flex-start + a wide gap so a long source wraps to two lines and never crowds the reviewer */
.prov-row {
  margin-top: var(--space-16);
  padding-top: var(--space-12);
  border-top: 1px solid var(--border-hairline); /* token-exception: a hairline */
  display: flex;
  align-items: flex-start;
  justify-content: space-between;
  gap: var(--space-24);
}

/* capped so a spelt-out body ("National Institute for Health and Care Excellence") wraps rather
   than pushing the reviewer off the right edge */
.prov-src {
  font-family: var(--font-label);
  font-size: var(--size-11);
  font-weight: var(--weight-600);
  color: var(--text-tertiary);
  line-height: var(--leading-140);
  max-width: 50%;
}

.prov-rev {
  display: inline-flex;
  align-items: flex-start;
  gap: var(--space-4);
  font-family: var(--font-label);
  font-size: var(--size-11);
  font-weight: var(--weight-700);
  color: var(--accent);
  line-height: var(--leading-140);
  text-align: right;
}

.prov-rev svg { width: var(--size-14); height: var(--size-14); flex: none; }