/* * Copyright 2026 Ali Sasanian * SPDX-License-Identifier: LicenseRef-PolyForm-Strict-1.0.0 * * egzos design tokens · v0.13 · 2026-11-08 (v0.12 · v0.11 · v0.10 · v0.9: 2026-09-23 · v0.8 · v0.7 · v0.6 · v0.5 · v0.4: 2026-09-22 · v0.3: 2026-09-11) * * v0.13 re-values the file to direction A′ (Blueprint, quieter lines), decided by the Chief on * 2026-11-08. No token is renamed, so components re-skin by value. The eight colours are the kit's * (Egzos/egzos-platform, design/concepts/lib/blueprint.mjs — a proprietary sibling repository): canvas * and paper are its ground, ink its text, ink-2 and ink-3 its muted text, rule its edge, rule-soft its * lo (a plate's one inner crease), act its lit thing. Strokes are hairlines (--egz-bw 1px), corners are * round (--egz-radius-plate 14px, -control 8px, -pill 20px) and nothing is offset: --egz-off and * --egz-off-lg are 0px and not 0, because a calc() adds them to a length and calc(16px + 0) is invalid, * which drops the whole declaration. Motion is two clocks: a discrete change is 700ms on * cubic-bezier(.32, .72, 0, 1), with --egz-motion-discrete as the one shared value for a CSS * transition, and a continuous one is a spring, k 100, c 18, m 1 (UX-METHOD §3, R8). Added: * --egz-mid, --egz-grid, the three radii, --egz-motion-discrete, --egz-spring-k / -c / -m, and the six * --hairline-* properties, declared below the shadcn bridge as aliases so no scheme block copies them. * Both schemes carry the A′ values for now: the Chief has not picked a light palette (direction B * Vellum is the candidate), the violet canvas repeats them, and color-scheme is dark in both because * the ground is. OPEN, for the Chief: A′ names no failure colour, so --egz-alarm keeps its dark-scheme * red, which is 4.0:1 on the ground, under the 4.5:1 the tap spec asks of red text (hsl(0 100% 66%) * is the first lightness that reaches it). * * Re-checked against v0.13, as README.md's Version rule asks. Type on a solid ink fill is * --egz-canvas, the claim four specs cite at v0.12, holds: ink is the text colour, canvas the ground * (11.9:1), and --egz-act-on still equals --egz-canvas. So does the §5 claim that this file carries no * size scale: v0.13 adds corners and clocks, no dimension. * * v0.12 is the FOURTH failure of the line above, and the worst-behaved: the seven versions * v0.10 … v0.4 were re-dated to 2026-09-23 as a side effect of v0.11's bump. Nothing edited * them. `same day` is a POINTER AT THE HEADER DATE, so when the header moved forward a day the * whole run moved with it, silently, and seven revisions written on the 22nd (and one on * 09-11, saved only by carrying an explicit date) claimed a day they did not happen on. The * three earlier failures were erasures a reader could notice; this one leaves the line looking * exactly as well-formed as before. * * So `same day` is no longer allowed to span a date boundary in this file: each run now states * its own date, on the evidence of this file's pushed blob at `6fa85751` — the branch's last * commit dated on or before 2026-09-22 — which reads `v0.8 · 2026-09-22 (v0.7 · v0.6 · v0.5 · * v0.4: same day · v0.3: 2026-09-11)`. A2's audit now reads DATES and not only version numbers: * a header against its own newest entry, a history that must run forward, a version the chain * names but nothing can date, and both index tables' date columns. It had thirty-four wrong * dates to find across thirteen files and found none, because every number agreed. * * v0.10 restores v0.8 to the line above. v0.9 replaced the leading version in place instead of * prefixing it, erasing the one revision whose whole content WAS this history line — "v0.8 fixes * that history line rather than extending it" is still in this file, describing a revision the line * no longer admitted. A history entry is a RECORD, and an erasure is a move: a sweep that moves one * is the sweep being wrong. Third instance of this class in this file, which is why A2's pre-push * audit now derives this chain too (it had only ever checked files with **Changelog** entries, and * this file has none — a CSS comment was the blind spot). * * v0.8 fixes that history line rather than extending it. Under this set's convention a list * terminated by one date assigns it to every version in it, so v0.7 ... v0.4 all read 2026-09-11 -- * false for four of the five, and contradicted by this file's own header at 2b51955 (v0.4 · * 2026-09-22), by DESIGN-SOURCES' 2026-09-22 row for the v0.6 bridge, and by the tap spec's record * that v1.6 shipped with tokens v0.5. Each bump prefixed the previous version without carrying its * date, so v0.4's date was absorbed into v0.3's group at 8299031 and stayed absorbed. These are * RECORDS (README.md, the Version rule): a sweep that moves one is the sweep being wrong. * The ONE code artifact A2 ships (decisions §P). Consumed as CSS variables by the lifeboat * (src/egzos/web, src/egzos/authz templates) and mapped into the Tailwind theme by the * flagship (Egzos/egzos-platform). The platform consumes this file; it never forks it. * * Direction: A′, Blueprint with quieter lines (decided by the Chief 2026-11-08; it replaced Docket v2, * bound 2026-09-11). Colour law: one lit thing — --egz-act, the single bright mark, which asks a * human — over one ground (--egz-canvas), three steps of line (--egz-rule, --egz-mid, --egz-rule-soft) * and two of type (--egz-ink, then --egz-ink-2 and --egz-ink-3, which coincide); --egz-alarm reports * a failure, and A′ names no colour for it yet. No yellow, lime or acid anywhere. Structure: 1px * hairline strokes, round corners (plate 14px, control 8px, pill 20px), no shadow, no offset, no glow, * no gradient. Schemes: both carry the A′ values for now (see v0.13). * * Scheme selection (both UIs): * the light scheme (default; A′ values for now) * the dark scheme * the dark scheme; the canvas setting stays, no cast under A′ * data-canvas is a USER SETTING (decided 2026-09-11), persisted per client; it has no effect * in the light scheme. prefers-color-scheme picks the initial scheme when none is stored. */ :root { /* ── canvas & ink ─────────────────────────────────────────────── */ /* LIGHT SCHEME: the Chief has not picked a light palette; it carries the A′ values for now (direction B Vellum is the candidate). */ --egz-canvas: #0E2A47; /* page: the ground */ --egz-paper: #0E2A47; /* queue rail, inset fields, secondary surfaces: the ground again; a plate is told from the page by its edge, not its fill */ --egz-ink: #DCEAF7; /* text */ --egz-ink-2: #9CBBD9; /* secondary text: muted */ --egz-ink-3: #9CBBD9; /* labels, metadata: muted (≥ 4.5:1 on canvas; 7.3:1) */ --egz-rule: #7FA3C8; /* the edge: structural borders, a plate's outline */ --egz-rule-soft: #2F5478; /* the lo: a plate's one inner crease, hairlines inside tables and lists, a disabled edge */ --egz-mid: #557A9F; /* the mid: a crease inside a solid, a secondary stroke (v0.13) */ --egz-grid: #1C3C5F; /* the grid: the ground's faint lattice (v0.13) */ --egz-shadow-ink: #0E2A47; /* the offset's colour: the ground, so a stray offset is invisible on it (v0.13) */ /* ── the lit thing and the failure colour ─────────────────────── */ --egz-act: #FFFFFF; /* the lit thing, the one bright mark: a human is asked: sign, approve, open a window, focus, the outward arrow */ --egz-act-on: #0E2A47; /* --egz-act-on is the type colour ON --egz-act, and on nothing else. * * TYPE ON AN INK FILL IS --egz-canvas (v0.9). Five specs render a word inside a solid ink * fill — the `verified` stamp, `active`, a held capability stamp, the matrix's explicit-coverage * cell, the gate's count badge — and until v0.9 none of them named a colour for that word, so * every one left it to a builder. The one spec that did answer reached for --egz-act-on, which * is legible today only because --egz-act-on and --egz-canvas happen to carry the same value in * all three schemes. They have unrelated jobs: --egz-act-on tracks whatever --egz-act becomes, * and --egz-canvas is the page. Re-value --egz-act to a lighter blue needing dark type by day * and --egz-act-on flips to near-black — at which point every ink-filled stamp renders black on * black, and nothing in the file would have predicted it. * * --egz-canvas is correct by construction: ink is the opposite of the canvas in every scheme, * so type in the canvas colour is always legible on an ink fill and always will be. Same for * --egz-alarm-on: it is the type colour on --egz-alarm, never on ink. */ --egz-alarm: hsl(0 100% 60%); /* Swiss Red: quarantine, presence lapsed, quota, unreachable, anomaly. A′ names no failure colour: the dark scheme's red in both schemes until the Chief picks (4.0:1 on canvas) */ --egz-alarm-on: #0E2A47; /* ── structure (A′: hairline, round, flat) ────────────────────── */ --egz-radius: var(--egz-radius-control); /* the default corner. Was 0 */ --egz-radius-plate: 14px; /* a plate: panel, card, sheet (v0.13) */ --egz-radius-control: 8px; /* a control: button, field (v0.13) */ --egz-radius-pill: 20px; /* a pill: the floating control, a chip (v0.13) */ --egz-bw: 1px; /* structural border weight: a hairline, as --egz-hair. Was 2px */ --egz-hair: 1px; /* hairlines inside tables and lists; skeleton borders (v0.4) */ --egz-off: 0px; /* no offset: box-shadow var(--egz-off) var(--egz-off) 0 var(--egz-shadow-ink) paints nothing. 0px, not 0: a calc() adds it to a length. Was 4px */ --egz-off-lg: 0px; /* the page frame's offset, the same. Was 8px */ --egz-focus: var(--egz-bw) solid var(--egz-act); /* outline; outline-offset 3px; 1px under A′ */ /* ── type ─────────────────────────────────────────────────────── */ --egz-font-ui: "IBM Plex Sans", system-ui, -apple-system, "Segoe UI", sans-serif; --egz-font-mono: "IBM Plex Mono", ui-monospace, "SFMono-Regular", Menlo, monospace; --egz-fs-1: 12px; --egz-fs-2: 13px; --egz-fs-3: 14px; --egz-fs-4: 16px; --egz-fs-5: 20px; --egz-fs-6: 28px; --egz-fs-7: 44px; /* Weight is a CLOSED set of two, not a range (D-T4). Until v0.11 there was no weight token at * all, while every spec's §16 forbade a "literal weight" — a rule with nothing to obey. The * renders did what a builder would: four different weights across seven files (400, 500, 600, * 700), chosen ad hoc, and two of them SYNTHESISED because the family was never loaded at that * weight. A browser-faked bold is a rendering artefact wearing a design decision's clothes, and * this is a record set in IBM Plex, not an approximation of one. * Two, because two is what the distinction needs (principle 3: weight is STRUCTURE, not * emphasis) and because these are the two the family is loaded at. 500 is neither regular nor * semibold; it is the weight you reach for when no rule stops you. * Load exactly these: IBM Plex Sans and IBM Plex Mono at 400 and 600. A value outside this set * is a design-gap against THIS file, never a literal in a component. */ --egz-w-regular: 400; --egz-w-semibold: 600; --egz-lh: 1.5; --egz-tracking-caps: .12em; /* mono labels, uppercase */ --egz-tabular: "tnum" 1; /* font-feature-settings for ids, hashes, timestamps, countdowns */ /* ── space ────────────────────────────────────────────────────── */ --egz-sp-1: 4px; --egz-sp-2: 8px; --egz-sp-3: 12px; --egz-sp-4: 16px; --egz-sp-5: 24px; --egz-sp-6: 40px; --egz-sp-7: 64px; /* ── motion (A′: two clocks, R8) ──────────────────────────────── */ --egz-motion-state: 700ms; /* a discrete change: a state change. v0.13: was 0ms, a stamp lands */ --egz-motion-reveal: 700ms; /* a discrete change: expand, fan, slide, a panel opening. Was 120ms */ --egz-motion-arrival: 700ms; /* a discrete change: an item arrives. Was 200ms */ --egz-motion-presence: 4000ms; /* presence: one full circuit of the window beam along the frame. Unchanged: the beam is a Docket v2 artefact and a CSS animation, which UX-METHOD §7.1 will reject in time; the Chief decides */ --egz-ease: cubic-bezier(.32, .72, 0, 1); /* v0.13: was linear, which R8 rejects */ --egz-motion-discrete: var(--egz-motion-state) var(--egz-ease); /* the ONE shared value for a CSS transition (UX-METHOD §4): 700ms cubic-bezier(.32, .72, 0, 1), and 0ms under reduced motion because it is built from --egz-motion-state */ --egz-spring-k: 100; /* a continuous input (the pointer, the camera, a drag) is a spring: k 100, c 18, m 1 (R8). */ --egz-spring-c: 18; /* Unitless, read by the motion kernel with getComputedStyle; CSS never animates with them */ --egz-spring-m: 1; --egz-t-hold: 700ms; /* v0.5: press-and-hold fill (flagship enhancement only). The other interaction timeouts — the two-step arm, the lifeboat's fragment refresh, the new-items debounce — are deliberately NOT tokens: the arm is validated server-side and the refresh is an htmx trigger, so a CSS variable there would imply a deployment could restyle a security timeout. See the tap spec §5 (D-T2). */ /* ── media (moodboard imagery, marketing only) ────────────────── */ --egz-img-filter: grayscale(.15) contrast(1.02); color-scheme: dark; /* the ground is dark in every scheme under A′, so native controls and scrollbars are too (was light) */ } /* NOTE — this comment annotates the block BELOW it, which is the rule a CSS comment binds to. The @media (prefers-color-scheme: dark) fallback further down duplicates every value in THIS block, value for value. Any change here must be made there too; a2-conformance checks it. It must NOT be read as applying to the violet block: violet is a user setting with no effect without data-canvas, so mirroring a violet value into the fallback would give an OS-dark viewer with no stored scheme a cast they never chose — the shape of the v0.4 defect where that block was missing --egz-img-filter. (Placement corrected in v0.8; it had sat above the violet block, where the default reading made 'this block' the wrong one.) */ html[data-scheme="dark"] { --egz-canvas: #0E2A47; --egz-paper: #0E2A47; --egz-ink: #DCEAF7; --egz-ink-2: #9CBBD9; --egz-ink-3: #9CBBD9; --egz-rule: #7FA3C8; --egz-rule-soft: #2F5478; --egz-mid: #557A9F; --egz-grid: #1C3C5F; --egz-shadow-ink: #0E2A47; /* the ground: a stray offset is invisible on it (was the ink: the record as a paper cutout) */ --egz-act: #FFFFFF; --egz-act-on: #0E2A47; --egz-alarm: hsl(0 100% 60%); --egz-alarm-on: #0E2A47; --egz-img-filter: grayscale(.2) brightness(.82) contrast(1.06); color-scheme: dark; } html[data-scheme="dark"][data-canvas="violet"] { /* v0.13: A′ has no violet cast. The canvas setting stays (scheme.ts still writes it); this block repeats the A′ values until the Chief decides whether the setting survives. */ --egz-canvas: #0E2A47; --egz-paper: #0E2A47; --egz-ink-3: #9CBBD9; --egz-rule-soft: #2F5478; --egz-act-on: #0E2A47; --egz-alarm-on: #0E2A47; } @media (prefers-color-scheme: dark) { html:not([data-scheme]) { --egz-canvas: #0E2A47; --egz-paper: #0E2A47; --egz-ink: #DCEAF7; --egz-ink-2: #9CBBD9; --egz-ink-3: #9CBBD9; --egz-rule: #7FA3C8; --egz-rule-soft: #2F5478; --egz-mid: #557A9F; --egz-grid: #1C3C5F; --egz-shadow-ink: #0E2A47; --egz-act: #FFFFFF; --egz-act-on: #0E2A47; --egz-alarm: hsl(0 100% 60%); --egz-alarm-on: #0E2A47; --egz-img-filter: grayscale(.2) brightness(.82) contrast(1.06); /* v0.4: was missing, so an OS-dark viewer with no stored scheme got the light filter on imagery */ color-scheme: dark; } } /* --------------------------------------------------------------------------- * shadcn/ui bridge — DECLARED, not documented (D-T3, tokens v0.6) * * Until v0.6 these keys existed only as a comment at the foot of this file. A comment * declares nothing: a catalogue component resolved `--background` against the empty string, * and a4s's only way to make the pick render was to retype all twenty-five lines into the * flagship theme — a fork by transcription, which this file's header forbids in the same * breath as a fork by copy. They are declarations now. The flagship maps the `--egz-*` layer * into Tailwind and gets this layer for free; nothing is retyped anywhere. * * Every value is an ALIAS, so the block is scheme-independent by construction: `var()` * resolves at use time against whichever scheme block is winning on , which is why * there is no dark copy of this list and must never be one. A dark copy would be a second * place to forget. * * The lifeboat pays nothing: an unreferenced custom property is inert — no paint, no layout, * no cascade cost. It is catalogue-exempt and simply never reads these names. * --------------------------------------------------------------------------- */ :root { /* surfaces */ --background: var(--egz-canvas); --foreground: var(--egz-ink); --card: var(--egz-canvas); --card-foreground: var(--egz-ink); --popover: var(--egz-canvas); /* Command #382, Popover, Context Menu, Tooltip */ --popover-foreground: var(--egz-ink); --muted: var(--egz-paper); --muted-foreground: var(--egz-ink-3); /* the one act colour, and the one failure colour — there is no third */ --primary: var(--egz-act); --primary-foreground: var(--egz-act-on); --destructive: var(--egz-alarm); --destructive-foreground: var(--egz-alarm-on); --secondary: var(--egz-paper); --secondary-foreground: var(--egz-ink); --accent: var(--egz-paper); --accent-foreground: var(--egz-ink); /* edges */ --border: var(--egz-rule); /* v0.13: the plate's edge; rule-soft is now the inner crease */ --input: var(--egz-rule); /* the field's edge: 1px under A′ (v0.13); was the 2px ink edge, not a hairline */ --ring: var(--egz-act); /* radius. The A′ set is three values (control 8, plate 14, pill 20), and shadcn derives * --radius-sm/md/lg/xl with calc(var(--radius) ∓ 2px|4px), which at 8px would invent 4, 6, 10 and 12: * four more. Each is pinned to one of the three (v0.13). Until then they were pinned to 0, because * the same calc went NEGATIVE at radius 0 and the component fell back to the browser default. */ --radius: var(--egz-radius); /* 8px */ --radius-sm: var(--egz-radius-control); --radius-md: var(--egz-radius-control); --radius-lg: var(--egz-radius-plate); --radius-xl: var(--egz-radius-pill); /* sidebar family — the flagship app shell's primary pick * (Egzos/egzos-platform · spec/design/search-list.md §17 — a proprietary sibling repository, * qualified because this file ships to every consumer of this public repository, a 0.1.0 fork * included, where an unqualified pointer resolves to nothing with no indication why). */ --sidebar: var(--egz-paper); --sidebar-foreground: var(--egz-ink); --sidebar-primary: var(--egz-act); --sidebar-primary-foreground: var(--egz-act-on); --sidebar-accent: var(--egz-canvas); --sidebar-accent-foreground: var(--egz-ink); --sidebar-border: var(--egz-rule); --sidebar-ring: var(--egz-act); /* DELIBERATELY ABSENT: --chart-1 … --chart-5. shadcn ships five categorical hues; egzos has * two colours and both are reserved (--act asks a human, --alarm reports a failure), so a * categorical ramp cannot be built from this palette without inventing a third. A pick that * reads a chart key is a design-gap against THIS file — A2 answers it with a non-chromatic * encoding (shape, hatch, position) or the Chief takes a colour-law decision. Do not alias * them to ink tints to make a component render: five greys that mean five categories is a * colour law nobody agreed, arriving through a component's back door. * * A pick needing any other unlisted key is likewise a design-gap against this file — never * a literal in the component, and never a retyped block in the flagship theme. * * SOURCE ORDER (v0.7, and binding on the flagship). This block only takes effect if it * out-cascades shadcn's own definitions, which a current shadcn / Tailwind v4 setup emits * from `@theme inline` into `:root` — the SAME selector at the SAME specificity, where * source order alone decides. So: **this file loads AFTER the shadcn theme layer**, or that * layer's colour and radius derivations are dropped. Either satisfies the requirement; * neither is optional. Get it wrong and nothing errors — `--radius-sm` quietly resolves to * `calc(0.625rem - 4px)` and a rounded corner appears in a square system, which is the * failure this block exists to prevent, arriving by a different door. A2 states the * requirement; where in the platform's CSS entry it is satisfied is a4s's to implement and * a2-conformance's to check. Since v0.13 the symptom of getting it wrong is a radius outside the A′ * set (14, 8, 20), not a rounded corner in a square system; the requirement is unchanged. */ } /* --------------------------------------------------------------------------- * Hairline figures — six custom properties, DECLARED (UX-METHOD §3 and §8, tokens v0.13) * * A Hairline figure themes through exactly these six, "set once from the app's tokens". They are * aliases, so — like the shadcn bridge above, and for the same reason — the block is * scheme-independent by construction: var() resolves against whichever scheme block is winning on * , and there is no scheme copy of this list and must never be one. The kit sets the same six on * each figure (blueprint.mjs, figure()); a figure inside the page inherits them from here instead. * --------------------------------------------------------------------------- */ :root { --hairline-plate: var(--egz-canvas); /* the plate's fill: it hides what is behind it (R6) */ --hairline-hi: var(--egz-act); /* lit: the one bright thing (R4) */ --hairline-edge: var(--egz-rule); /* a silhouette's outline */ --hairline-mid: var(--egz-mid); /* a crease inside it */ --hairline-lo: var(--egz-rule-soft); /* the dimmest line: an inner outline, a guide, an off state */ --hairline-stroke: 1; /* stroke width, unitless: SVG user units, non-scaling */ } @media (prefers-reduced-motion: reduce) { :root { --egz-motion-state: 0ms; /* v0.13: the discrete clock is no longer instantaneous, so it is zeroed here, and --egz-motion-discrete with it */ --egz-motion-reveal: 0ms; --egz-motion-arrival: 0ms; --egz-motion-presence: 0ms; /* the beam becomes a static marked segment; the countdown stays numeric */ } } /* * Tailwind mapping (flagship; A4s wires this in tailwind config / @theme): * colors.canvas → var(--egz-canvas) colors.paper → var(--egz-paper) colors.ink → var(--egz-ink) * colors.ink-2 / ink-3 / rule / rule-soft / mid / grid / act / act-on / alarm / alarm-on likewise * borderRadius.DEFAULT → var(--egz-radius) (8px) borderRadius.plate / control / pill → --egz-radius-plate / -control / -pill * borderWidth.structural → var(--egz-bw) borderWidth.hair → var(--egz-hair) * boxShadow.off → var(--egz-off) var(--egz-off) 0 var(--egz-shadow-ink) (paints nothing from v0.13: --egz-off is 0px) * fontFamily.ui / mono → the two stacks above fontSize.1…7 → --egz-fs-1…7 * transitionDuration.state / reveal / arrival / presence → the motion tokens transition (shorthand) → var(--egz-motion-discrete) * shadcn/ui bridge: DECLARED ABOVE (D-T3, v0.6), not listed here. The block above this comment * is the bridge; it was a comment in v0.4–v0.5 and nothing declared the keys. A pick needing a * key the block does not declare is a design-gap against THIS file, never a literal in the * component and never a retyped block in the flagship theme. */