stack/agents/dewey/work/wui/VERIFY.md

Brand-board verification

Author: Dewey. Date: 2026-09-08. Scope: local interactive brand prototype, ready for review, not an integrated product or approved identity.

Reproduce

From /mnt/storage/src/mosaic-stack:

node --check agents/dewey/work/wui/brand.js
node --check agents/dewey/work/wui/board.js
node agents/dewey/work/wui/checks/verify.mjs
node agents/dewey/work/wui/checks/detail-checks.mjs
node agents/dewey/work/wui/checks/final-checks.mjs

Environment used: Node 26.8.1, Chromium 152.0.7977.75 on Arch Linux. The CDP helper uses /usr/bin/chromium, or the CHROMIUM environment variable. Tests create a fresh temporary browser profile, open the entry page by file URL, save evidence here, and remove the profile at completion. No authenticated browser profile or Mosaic backend is used. There are no npm dependencies for this board or these tests.

checks/assets.mjs regenerates standalone SVGs and palette JSON from brand.js. checks/fetch-fonts.py refreshes the public Google Fonts WOFF2/OFL assets and needs network access. Neither is needed just to view or verify the current board. checks/references.mjs repeats the public reference-site inspection and also needs network access.

Results

Requirement	Result	Evidence
Three distinct logo concepts	pass for review presentation	Assembly, Relay, Aperture rendered in identity cards, standalone SVGs, small monochrome samples, and selectable header/wordmark specimen. Author inspected screenshots. Recognition and final brand suitability are still Jason's decision.
Sans-serif alternatives and coherent icons	pass	Three locally loaded families with four weights each and license files. Eight labeled navigation icons use one original 24-unit/1.75-stroke family, shared with feedback symbols.
Ten theory-backed palettes, all three appearances	pass	Ten palettes carry named harmony relationships and explicit hue anchors in brand.js and the board. Thirty token sets generated. Palette/mode controls tested through their real change/click handlers.
Token contrast, A11Y-01/02	pass for declared pairs	840 unrounded ratio checks in verification.json. Normal text/action-label minimum observed 4.503925718129755:1, boundary/focus minimum 3.704773189325781:1. Checks cover normal/muted/action/accent/status text across canvas, surface, raised surfaces; borders/focus across those surfaces; on-action text on action. Decorative dividers and disabled controls use criterion exceptions.
Rendered text contrast	pass for sampled DOM text	274 text/control samples per combination, 8,220 total. Computed foreground and solid ancestor backgrounds, qualifying large-text thresholds, minimum observed 4.504658476260286:1. This is not an image/gradient algorithm or a full accessibility audit.
Color-independent meaning, A11Y-03	pass	Selection uses Selected labels and pressed/checked state. Status uses text plus icons. Author inspected the rendered component section.
Keyboard operation and focus, A11Y-05/06	pass for tested flows	Native select and radio-arrow changes; Enter/Space selection buttons; Tab/Shift+Tab between sample fields; error recovery; skip link moves focus to main. Focus ring measured at 3 px solid. Short-screen sidebar focus stayed inside the 768 px viewport after adding sidebar scrolling. Chromium accessibility tree reports accessible names for all 60 interactive elements. This is not a screen-reader test.
Target size, A11Y-04	pass for measured controls	Enabled buttons, fields, radio inputs, sidebar links, brand link and review link meet 24 px dimensions at 320 and 1440 widths. Inline footer/prose links use the inline-text exception. Most controls are at least 44 px tall.
Responsive/full-width layout, A11Y-07	pass for measured layouts	All 30 combinations at 320, 390, 768, 960, 1440, 1920, 2560, 3440 CSS px. 240 layout checks, no page overflow. Ten additional checks cover both sides of the 700, 960, 1200, 1700, and 2400 breakpoints. Author inspected mobile, desktop, and 3440-wide screenshots; five palette columns use ultrawide space without unbounded prose.
Text enlargement/spacing, A11Y-07/08	pass for tested overrides	320 px reflow with WCAG text-spacing overrides, 200% root size, and separately doubling every element's computed font size including authored px values. Initial full-text expansion overflowed the review grid and small-size labels; minimum-width/wrapping fixes passed retest and screenshot inspection. Browser-native zoom was not exercised.
Reduced motion, A11Y-10	pass	Emulated reduced-motion media preference; theme and keyboard controls remain usable. Board state changes do not rely on animation.
Loading/empty/error/success/disabled states	pass as labeled specimens	Tested all five selectable states, loading's disabled action, success button, preserved input after error. No actual request or project creation occurs.
Long content and failure fallback	pass for tested cases	Wrapped headings/labels and doubled text; font loads blocked and browser storage denied in isolated Chromium, with working controls and no page overflow at 320. Malformed saved preference JSON recovers to default. Input has a 160-character limit and native horizontal text-field scrolling.
Review notes and token downloads	pass	Browser downloads captured in evidence/downloads and contents checked. Note download contains test review text; palette JSON includes Light/Dim/Dark tokens. These files are author test fixtures, not Jason's decisions.
Offline resource/link integrity	pass	CDP resource tree showed 11 loaded resources, all local file URLs. Source asset links and internal anchors resolve. Three real font families load. No remote font/CDN/backend request belongs to the board.
Public site completeness	not applicable to this local review artifact	Single file with section anchors, no public deployment or app router. About/Contact/legal pages, HTTP 404 behavior, consent management, billing and account workflows are outside this prototype. Storage use is explained beside controls and notes. No legal compliance claim.
Reference sites	pass for public-page inspection, app behavior not verified	RESEARCH.md and reference screenshots. No v1 inspection. T3 application theme switching and Buzz's installed application wizard were not exercised.

Visual fixes and test corrections

• Early reference screenshots captured entrance animations, not completed pages. Re-captured after a bounded animation allowance.
• CDP Enter needed its character value, and Shift uses modifier bit 8. Fixed the test driver rather than changing working controls to accommodate a faulty test.
• Added scrollable desktop preview controls so keyboard focus remains available on short screens.
• Doubled text exposed a review-column minimum-width problem and logo-size row wrapping. Fixed and rechecked.
• A 320 px screenshot exposed a clipped Conversations label. Changed narrow icon layout to two columns and allowed long labels to wrap. Rechecked the actual rendering.
• Corrected draft hue anchors so named complementary/triadic/split-complementary relationships match their stated angles. Regenerated assets and re-ran contrast/layout suites.
• File-URL resources did not populate the expected Performance entries. The final resource check uses Chromium's Page.getResourceTree instead; its resource URLs are all local.

Verification limits and remaining decisions

• Tested one installed desktop Chromium engine. Safari, Firefox, native mobile devices, touchscreen behavior, assistive technology, OS forced colors, and browser-native zoom are not verified.
• Browser viewport emulation and doubled text are useful evidence, not equivalent to testing every physical display and accessibility configuration.
• Contrast assertions and Chromium accessibility-tree names do not establish full WCAG conformance. They cover the listed cases.
• Only Latin font subsets are bundled. Multilingual typography and fallback coverage need a later product brief.
• Logos are original author studies, not independently reviewed, user-tested, trademark-cleared, or final production assets. Small-size optical refinements await selection.
• No independent technical review, issue lifecycle, repository delivery suites, integration, deployment, commits, or pushes were performed for this standalone design package. Those delivery gates remain separate if the prototype is promoted into product work.
• Jason has not approved branding. B09 and all dashboard tasks remain behind that human gate.