Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when the user wants to create or update a brand guideline: brand personality, color system, typography, logo usage, imagery and visual tone, component rules, accessibility contrast standards, standard asset sizes, naming and file organization. The output is the foundation document every design skill and the Brand Hub reads from. Trigger on 'brand guideline', 'brand book', 'brand identity', 'style guide', 'brand standards', 'brand palette', or 'update the brand guide'.
.claude/skills/minhnv0807-46-brand-guideline-global/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-10 | ✗→✓ | ▲ Improved | 60% | 0% |
| case-13 | ✗→✓ | ▲ Improved | 146% | 0% |
| case-17 | ✗→✓ | ▲ Improved | 166% | 0% |
| case-19 | ✗→✓ | ▲ Improved | 92% | 0% |
| case-05 | ✓→✓ | = Same ✓ | 8% | 0% |
A brand guideline is not a beautiful document to archive — it is a daily working tool. A good one lets everyone produce on-brand work without asking the lead. This is the foundation document: every design skill (42-50) and the Brand Hub read from it.
Read .agents/product-marketing-context-global.md, the logo files, and any existing brand material. If information is missing, ask up to 4 questions:
34-ai-marketing-os-global) so every role and agent reads one source.Three precise adjectives plus three adjectives to avoid, and the feeling a viewer should have when they see the work. If positioning is not settled, read 09-customer-insight-global and the existing positioning first.
From the logo and the personality, set the palette (primary, secondary, background, text) and the fonts (heading font], body font], plus a fallback). Every color carries usage rules and a contrast check.
Logo versions and don'ts; photography, icon, and background style; components (button, card, tag, divider) so every asset stays consistent.
Asset size table by channel, naming convention, folder tree. Then set up the brand kit in the team's design tool.
File name: brand-guideline-[brand-name]-[YYYYMMDD].md. Bracketed values are placeholders — fill them with the brand's real colors and fonts. The illustrative palette below is a neutral EXAMPLE only; replace it.
markdown# Brand Guideline — [Brand name] Last updated: [date] | Owner: [name] | Version: [x.y] ## Part 1 — Brand foundation Brand personality: [Brand] is [three precise adjectives] — not [three adjectives to avoid]. (Example only: "Sharp, Practical, Empowering — not Corporate, Fluffy, Generic") Brand voice in visual work: - Seeing our design, a viewer should feel: [...] - A viewer should never feel: [...] ## Part 2 — Color system Primary palette: | Name | Hex | RGB | Use when | Do not use when | |------|-----|-----|----------|-----------------| | [primary color] | [hex] | [rgb] | Headings, CTAs, dominant grounds | Ground for long-form body text | | [secondary color] | [hex] | [rgb] | Accents, highlights, figures, icons | Large-area backgrounds | | [background color] | [hex] | [rgb] | White ground, cards, content areas | — | Secondary palette: | Name | Hex | Use when | |------|-----|----------| | [strong text color] | [hex] | Body text and headings on light grounds | | [muted text color] | [hex] | Subtext, captions, metadata | | [light background color] | [hex] | Light section grounds, cards, tags | | [gradient] | [primary] to [secondary] | Hero, CTA section, covers | (Illustrative example only — replace with the real brand values: primary navy #0F4C81, secondary amber #F2A104.) Approved combinations: - [primary color] ground + light text - Light ground + [primary color] heading + [secondary color] accent - Gradient [primary to secondary] + light text - [light background color] + [primary color] text Forbidden combinations: - [secondary color] as a large-area background - Light text on [light background color] (insufficient contrast) - Three or more dominant colors in one asset - Any color outside the palette without approval - Neon or oversaturated colors that break the brand tone ## Part 3 — Typography Primary fonts: [heading font] (headings) / [body font] (body). Fallback: [fallback font], sans-serif. | Style | Weight | Size range | Used for | |-------|--------|-----------|----------| | Display | Black 900 | 48-72px | Hero headline, large figures | | H1 | Bold 700 | 36-48px | Page or post headline | | H2 | Bold 700 | 24-32px | Section heading | | H3 | SemiBold 600 | 18-22px | Card heading | | Body | Regular 400 | 14-16px | Paragraphs | | Caption | Regular 400 | 11-13px | Metadata, timestamps | | CTA | Bold 700 | 14-16px | Button text | Typography don'ts: - No fonts outside [heading font] / [body font] / fallback - No ALL CAPS for long paragraphs - No justified text - No body text below 13px (print) or 14px (mobile) - One typeface family, many weights — do not mix families ## Part 4 — Logo usage | Version | Use when | File | |---------|----------|------| | Full logo (color) | Header, cover, first slide, landing page | SVG + PNG @2x | | Full logo (light/reversed) | On dark grounds or [primary color] | SVG + PNG @2x | | Icon only (color) | Watermark, favicon, story, avatar | SVG + PNG @2x | | Icon only (light) | Dark grounds, small sizes | SVG + PNG @2x | Rules: - Minimum clear space 16px on every side — never crop tight - Minimum width 80px (below that it loses definition) - Place on a light ground or [light background color] — never on a busy photograph Logo don'ts: no rotation, no stretching or squashing, no added shadow/outline/glow, no colors outside the approved versions, no placement that muddies the mark. ## Part 5 — Imagery and visual style - Photography: [tone — for example warm, natural, candid but professional] — no filters that fight the brand tone - Prefer real people; avoid generic stock that dilutes the brand - Never use images carrying another source's watermark - Icons: [line / filled rounded] — single-color in [primary color] or [secondary color] - Backgrounds: [approved grounds] — no busy patterns; keep flat, maximum two blended colors - Watermark logo stays subtle in a corner ## Part 6 — Component rules Primary button: fill [primary color], light bold 15px text, radius [x]px, padding 12x24px, hover ~10% darker Secondary button: transparent, 1.5px border [primary color], text [primary color], hover fill [light background color] Card: light ground, 1px soft border, radius 12px, soft shadow, padding 24px Tag/badge: ground [light background color], text [primary color] 12px medium, pill radius, padding 4x12px Divider: 1px soft neutral (roughly 8% black) ## Part 7 — Standard asset sizes Social: square feed 1080x1080 (1:1), portrait feed 1080x1350 (4:5), story/reels/TikTok 1080x1920 (9:16), Facebook cover 820x312, LinkedIn post 1200x627, Pinterest pin 1000x1500, YouTube thumbnail 1280x720 Meta ads: feed 1080x1080, story 1080x1920 (keep content ~14% off the top and bottom edges), carousel 1080x1080 (2-10 cards), link banner 1200x628 Documents: slide 16:9 1920x1080, A4 portrait/landscape, email width 600px (Full table with safe zones and resize strategy: read `50-asset-resize-global`.) ## Part 8 — Accessibility, dos and don'ts, file organization Accessibility (WCAG 2.2 Level AA): - Body text contrast at least 4.5:1 against its background - Large text (18pt / 24px, or 14pt / 18.66px bold and above) at least 3:1 - UI components, icons, and graphical objects carrying meaning at least 3:1 - Never convey meaning by color alone — pair it with a label, icon, or pattern - Minimum 14px display type on mobile; captions on all video - Note: accessibility is legally enforced in many markets (ADA and Section 508 in the US, the European Accessibility Act in the EU) — treat AA as the floor, not the goal DO: - Use generous whitespace — do not fill every gap - Keep hierarchy obvious: H1 dominant, supporting text smaller and lighter - Keep text legible on a phone (>=14px) - Keep spacing consistent on an 8px grid - Pick one alignment and hold it - Preview on mobile before sending for approval - Put the logo on every asset DON'T: - More than three colors in one asset - Text over busy imagery with no overlay - Fonts or colors outside the guideline without approval - JPG exports below 85% quality Naming: [Brand]_[Campaign]_[AssetType]_[Size]_[Version]_[Status] Status tags: _DRAFT / _REVIEW / _APPROVED / _ARCHIVED Folders: /[Brand]/Brand Assets/ (logo, fonts, icons) /[Brand]/Campaigns/[CampaignName]/Assets/Working|Final|Archived/
Once the guideline is approved, load it into whichever tool the team uses:
Brief from the lead
-> Designer confirms objective, audience, platform, mood
-> Concept MVP (HTML or fast draft — 30 minutes max)
-> Lead approves the concept
-> Production in Canva / Figma / Adobe Express against this guideline
-> Export at the correct sizes, named to convention
-> Save the source link into the asset library, files into shared storageproduct-marketing-context-global and 34-ai-marketing-os-global: this guideline is a core Brand Hub file.42-image-brief-global, 43-carousel-brief-global, 44-video-editor-brief-global, 45-design-tool-brief-global: every design brief reads colors, fonts, and logo rules from here.47-design-review-global: brand consistency is scored against this document.49-html-email-template-global and 12-landing-page-brief-global: email and landing page design tokens come from here.50-asset-resize-global: full size matrix and safe zones.30-design-master-global: AI image generation reads brand identity from this guideline.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 42,885 | 40,128 | -6% | 1 | 1 | 0% | 8,305 | 9,911 | +19% | 0 | 0 | — |
case-02 | fail→fail | 48,403 | 32,084 | -34% | 1 | 1 | 0% | 8,299 | 8,374 | +1% | 0 | 0 | — |
case-03 | fail→fail | 48,072 | 52,978 | +10% | 1 | 1 | 0% | 8,275 | 10,615 | +28% | 0 | 0 | — |
case-04 | fail→fail | 13,593 | 13,944 | +3% | 1 | 1 | 0% | 2,425 | 4,499 | +86% | 0 | 0 | — |
case-05 | pass→pass | 31,293 | 18,614 | -41% | 1 | 1 | 0% | 6,092 | 6,569 | +8% | 0 | 0 | — |
case-06 | fail→fail | 19,405 | 15,682 | -19% | 1 | 1 | 0% | 3,398 | 6,097 | +79% | 0 | 0 | — |
case-07 | fail→fail | 21,978 | 18,055 | -18% | 1 | 1 | 0% | 3,970 | 6,394 | +61% | 0 | 0 | — |
case-08 | fail→fail | 38,218 | 21,404 | -44% | 1 | 1 | 0% | 8,237 | 7,733 | -6% | 0 | 0 | — |
case-09 | pass→pass | 29,279 | 15,379 | -47% | 1 | 1 | 0% | 5,116 | 5,726 | +12% | 0 | 0 | — |
case-10 | fail→pass | 21,360 | 17,415 | -18% | 1 | 1 | 0% | 3,846 | 6,164 | +60% | 0 | 0 | — |
case-11 | fail→fail | 16,196 | 13,521 | -17% | 1 | 1 | 0% | 2,768 | 5,587 | +102% | 0 | 0 | — |
case-12 | fail→fail | 16,910 | 12,932 | -24% | 1 | 1 | 0% | 3,094 | 5,718 | +85% | 0 | 0 | — |
case-13 | fail→pass | 15,744 | 18,498 | +17% | 1 | 1 | 0% | 2,822 | 6,955 | +146% | 0 | 0 | — |
case-14 | fail→fail | 21,707 | 9,442 | -57% | 1 | 1 | 0% | 1,144 | 4,997 | +337% | 0 | 0 | — |
case-15 | fail→fail | 16,328 | 14,078 | -14% | 1 | 1 | 0% | 2,682 | 4,986 | +86% | 0 | 0 | — |
case-16 | fail→fail | 11,741 | 24,345 | +107% | 1 | 1 | 0% | 2,062 | 7,827 | +280% | 0 | 0 | — |
case-17 | fail→pass | 15,609 | 21,905 | +40% | 1 | 1 | 0% | 2,622 | 6,983 | +166% | 0 | 0 | — |
case-18 | pass→pass | 15,921 | 12,647 | -21% | 1 | 1 | 0% | 2,714 | 5,153 | +90% | 0 | 0 | — |
case-19 | fail→pass | 11,670 | 4,333 | -63% | 1 | 1 | 0% | 1,996 | 3,835 | +92% | 0 | 0 | — |
case-20 | fail→fail | 2,664 | 35,765 | +1243% | 1 | 1 | 0% | 450 | 11,319 | +2415% | 0 | 0 | — |
case-21 | fail→fail | 38,313 | 39,431 | +3% | 1 | 1 | 0% | 8,212 | 11,335 | +38% | 0 | 0 | — |
case-22 | fail→fail | 29,595 | 25,190 | -15% | 1 | 1 | 0% | 6,745 | 8,576 | +27% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 21 counted toward the lift figure. The other 1 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +18 percentage points is the difference between those two pass rates over the 21 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.