/* ═════════════════════════════════════════════════════════════════════════════ VENDORED VERBATIM — DO NOT EDIT THIS FILE HERE. Source : Alterspective-Design-System-Showcase — assets/adl.css Commit : 11db15509c54dec7dfcb9532da8fa5f7f67b4958 2026-08-11 — feat(ui): glass context rails — hover-reveal, click-to-anchor (⚓), AI Insights card (ADL-DEC-059) (#36) Digest : sha256 df84086383df02e8b2b9587c30da2c9ae61876274b147dbb9d176bcf4c1ea0b4 (459 lines, LF-normalised, the upstream file as committed) Everything after this banner is byte-for-byte the upstream file. The banner is the only local addition, and it exists because the last copy carried no provenance at all: this file had drifted 4 changed lines and 306 MISSING lines from its source — 153 lines vendored against 459 upstream, everything from §4 onwards absent — and nothing in the repository could say so, because nothing recorded what "its source" meant. A vendored file with no recorded revision cannot drift visibly; it can only drift. (An earlier draft of this banner said 287 lines and "§4-§12". Neither reproduced: the old copy stopped at §3 and upstream continues past §13, so the gap is §4-end.) TO RE-SYNC when upstream moves: replace everything after this banner with the upstream file, then update Commit, Digest and the line count above. TO CHECK for drift — BANNER_LINES is the placeholder replaced below: tail -n +39 public/adl.css | tr -d '\r' | sha256sum The digest is over LF-NORMALISED bytes deliberately. core.autocrlf is true on Windows here, so a digest taken over working-tree bytes changes with the checkout, and the check would fail on a clean clone while nothing was wrong. Normalised, it equals `git show :assets/adl.css | sha256sum` in the Showcase repository exactly — the comparison that actually matters. FIXING SOMETHING IN HERE MEANS FIXING IT UPSTREAM. A local edit is invisible to every other consumer of the design system and is silently destroyed by the next re-sync. Page-local corrections belong in app.css, which is linked AFTER this file for exactly that purpose (see the pill notes there). ═════════════════════════════════════════════════════════════════════════ */ /* ═════════════════════════════════════════════════════════════════════════════ assets/adl.css — the SINGLE SOURCE for the ADL mock-card anatomy and the tenant-skin (brand) token scopes. Supersedes the copied-block mechanism (DEC-024 → DEC-026): pages this file instead of carrying a verbatim copy. Repo-relative, no CDN, no build — pages still open from disk. HOW THIS FILE IS ORGANISED (for the next editor, human or AI): §1 ANATOMY — the mock-card component classes. NEVER themeable: no brand may change geometry here (radius, spacing, label placement, rail width). Change = design change = review round. §2 BRANDS — tenant-skin token scopes. A brand ONLY re-declares tokens (colour, type, button radius). To add a brand: copy one scope pair (light + dark), rename the [data-brand="…"] key, and register it in assets/adl-brands.js. Nothing else. §3 SWITCHER — chrome for the brand bar that adl-brands.js renders. CONTRACTS THE PAGES RELY ON: - Every class that carries layout declares `display` explicitly (ballot stimuli must stay -based inside