Design System for GTM

Overview

When AI agents generate sales collateral, pitch decks, one-pagers, social graphics, and emails, they need a brand system to reference. Without it, every output has different colors, different fonts, different voice — and your brand erodes across every customer touchpoint. This skill defines a complete AI-consumable design system with visual identity, voice/tone, templates, and guardrails. Designed to be read by both humans and AI agents.

Authoritative Foundations

• Google DESIGN.md Specification — Shapes deliverables for this skill — When AI agents generate sales collateral, pitch decks, one-pagers, social graphics, and emails, they need a brand system.
• Stitch 9-Section Brand Format — Shapes deliverables for this skill — When AI agents generate sales collateral, pitch decks, one-pagers, social graphics, and emails, they need a brand system.
• Brad Frost — Atomic Design — Atomic Design
• Nathan Curtis — Design System Monorepo Architecture — Design System Monorepo Architecture
• Dan Mall — Design System Governance — Design System Governance

When to Use

Trigger phrases: "create a design system for our AI-generated content", "define our brand for agents", "build a brand guide for sales materials", "ensure consistent voice across AI outputs", "DESIGN.md specification", "design tokens for agents", "AI brand system", "brand context for Claude"

Step-by-Step Process

Phase 1: Visual Identity System

Color Palette — must include hex, RGB, and HSL:

PRIMARY PALETTE
  Primary:        #0066FF  rgb(0,102,255)   hsl(216,100%,50%)
  Primary Dark:   #0044CC  rgb(0,68,204)    hsl(220,100%,40%)
  Primary Light:  #3388FF  rgb(51,136,255)  hsl(216,100%,60%)

SECONDARY / ACCENT
  Accent:         #FF6600  rgb(255,102,0)   hsl(24,100%,50%)

NEUTRALS
  Ink (text):     #1A1A2E  rgb(26,26,46)    hsl(240,28%,14%)
  Muted Text:     #6B7280  rgb(107,114,128) hsl(220,9%,46%)
  Surface:        #FFFFFF
  Background:     #F9FAFB
  Border:         #E5E7EB

SEMANTIC
  Success:        #10B981  rgb(16,185,129)  hsl(160,84%,39%)
  Warning:        #F59E0B  rgb(245,158,11)  hsl(37,92%,50%)
  Danger:         #EF4444  rgb(239,68,68)   hsl(0,80%,60%)
  Info:           #3B82F6  rgb(59,130,246)  hsl(217,91%,60%)

CSS Variable Output (for agents generating HTML/CSS):

:root {
  --color-primary: #0066FF;
  --color-primary-dark: #0044CC;
  --color-primary-light: #3388FF;
  --color-accent: #FF6600;
  --color-ink: #1A1A2E;
  --color-muted: #6B7280;
  --color-surface: #FFFFFF;
  --color-bg: #F9FAFB;
  --color-border: #E5E7EB;
  --color-success: #10B981;
  --color-warning: #F59E0B;
  --color-danger: #EF4444;
}

Typography System:

HEADINGS: Inter (Google Fonts)
  H1: 48px / 1.1 line-height / 700 weight / letter-spacing: -0.02em
  H2: 36px / 1.2 / 600 / -0.01em
  H3: 24px / 1.3 / 600 / 0
  H4: 20px / 1.3 / 600 / 0

BODY: Inter
  Body L: 18px / 1.6 / 400
  Body: 16px / 1.6 / 400
  Body S: 14px / 1.5 / 400
  Caption: 12px / 1.5 / 400

MONOSPACE (code, data): JetBrains Mono
  Code: 14px / 1.5 / 400

CSS:
  @import url('https://fonts.googleapis.com/css2?family=Inter:wght@400;500;600;700&display=swap');
  font-family: 'Inter', -apple-system, BlinkMacSystemFont, 'Segoe UI', sans-serif;

Spacing Scale (8px base): 4, 8, 12, 16, 24, 32, 48, 64, 96, 128

Border Radius:

• Buttons, inputs: 8px
• Cards: 12px
• Modals: 16px

Shadows:

• Card: 0 1px 3px rgba(0,0,0,0.08)
• Elevated: 0 4px 12px rgba(0,0,0,0.1)
• Modal: 0 8px 32px rgba(0,0,0,0.12)

Phase 2: Voice and Tone Guide

Brand Voice Attributes (5 — be specific, not vague):

1. Direct — no hedging, no "we believe," no "it could be argued." Say the thing.
2. Data-backed — every claim has a number or source. "Most teams" → "73% of teams."
3. Opinionated — take positions. "We recommend X, not Y" not "There are several approaches."
4. Concrete — "47% reply rate" not "industry-leading engagement."
5. Confident but not arrogant — "Here's what works" not "We're revolutionizing."

Tone by Context:

Context	Tone	Example
Sales deck	Confident, data-heavy	"Teams using our platform see 47% reply rates within 30 days."
Case study	Narrative, specific	"Acme Corp was losing $2.3M/year to churn. Here's how they fixed it."
Email nurture	Helpful, low-pressure	"I noticed you downloaded our ROI calculator. Here are 3 ways customers use it."
Error messages	Clear, apologetic, helpful	"That didn't work. Here's why and how to fix it."
Social media	Punchy, shareable	"Stop guessing your ICP. Start scoring it."

Banned Words and Phrases:

• "Revolutionary," "high-impact," "disruptive" (vague, overused)
• "Best-in-class," "industry-leading" (meaningless without data)
• "Synergize," "leverage" (as a verb), "holistic" (jargon)
• "We're excited to announce" (nobody cares if you're excited)
• Passive voice: "Results were achieved" → "We achieved results"

Preferred Language Patterns:

• Active voice always. "We built" not "It was built."
• Present tense. "The platform processes" not "The platform will process."
• Short sentences. Under 25 words when possible.
• Bullets over paragraphs in decks and one-pagers.

Phase 3: Asset Templates

Slide Master Template:

• 16:9 aspect ratio
• Header: Inter 36px, primary color, left-aligned
• Body: Inter 18px, ink color
• Footer: company logo (bottom-right), slide number (bottom-left), 12px muted
• Data slides: charts in primary + accent palette, no decorative elements

One-Pager Template:

• A4/US Letter
• Header: logo + product name
• Hero: value prop (H2), one-line description, hero image
• Body: 3-column feature grid
• Social proof: 3 customer logos + 1 testimonial
• CTA: prominent button/contact info
• Footer: company info, legal

Email Signature:

[First Name] [Last Name]
[Title] | [Company Name]
[Phone] | [Website]
[Calendly link]
[Company logo — 80px wide]

Data Visualization Style:

• Primary + neutral palette (limit to 5 data colors)
• No 3D charts, no pie charts (use bar/line/scatter)
• Direct labels (no legends when possible)
• Source line: "Source: [name], [date]" in 10px muted

Phase 4: Brand Guardrails

Always Acceptable (AI can generate without review):

• Social graphics using brand colors, typography, and voice
• Email copy following tone guidelines
• Blog post drafts (human reviews before publishing)
• Data visualizations in brand style
• Landing page sections following template

Requires Human Review:

• Customer-facing decks (check data accuracy)
• Pricing pages (legal/commercial review)
• Press releases (legal review)
• Customer quotes/testimonials (customer approval)
• Any content referencing specific customers

Never Allowed:

• Competitive claims without cited sources
• Revenue/growth numbers not publicly disclosed
• Customer names without explicit permission
• Guarantees or promises ("100% deliverability")
• Any content that implies endorsement by a named authority

Integration with Claude Code Design

When using Claude Code to generate branded assets, load this skill alongside claude-design and popular-web-designs:

1. claude-design → drives the design process and artifact creation
2. design-system-gtm (this skill) → provides brand tokens and rules
3. popular-web-designs → provides visual vocabulary from known brands

The agent should output CSS variables matching this design system's tokens and follow the voice/tone guidelines for any copy it generates.

Output Format

DESIGN SYSTEM — [Company Name]

File: DESIGN.md (version controlled in repo)

Sections:
1. Brand Overview (2-3 sentence positioning)
2. Visual Identity (colors, typography, spacing, radii, shadows)
3. CSS Variable Output (copy-paste ready)
4. Voice and Tone (attributes, per-context, banned words)
5. Asset Templates (slide, one-pager, email, social, data viz)
6. Brand Guardrails (acceptable / review / never)
7. AI Agent Usage (how agents should reference this system)

Implementation Checklist

• All colors have hex, RGB, and HSL values
• CSS variable block is copy-pasteable into any project
• Typography includes Google Fonts import URL and font stack
• Voice attributes are specific and testable ("direct" > "professional")
• Banned words list exists (10+ entries)
• Tone guide covers at least 5 contexts
• Asset templates include specific dimensions and specs
• Guardrails have 3 clear tiers (always acceptable / review / never)
• AI agent usage instructions included (how to reference, what to load)
• File is version-controlled and linked from README

Quality Check

Before delivering, verify:

• Output matches the user's stated request
• Named frameworks or sources are reflected in the recommendation
• The deliverable is specific enough for an agent to execute
• Any assumptions, risks, or dependencies are explicit
• No unsupported claims, invented facts, or private/internal references are included

Common Pitfalls

1. Vague voice guidelines. "Professional and friendly" is useless. Agents need specific, testable attributes: "Short sentences. Active voice. No passive constructions. Data before adjectives." Fix: For each voice attribute, provide 3 examples of "this, not that."

2. RGB without hex. AI agents reference hex codes for HTML/CSS generation. Including RGB is good. Including only RGB is not. Fix: Hex is the primary color format for agents. RGB and HSL are supplementary.

3. No CSS variable output. Agents generating HTML need copy-pasteable CSS. Wading through a color palette table to extract hex codes produces errors. Fix: Include a :root {} CSS block with all design tokens.

4. Ignoring dark mode. Many agent-generated artifacts (dashboards, presentations) look different in dark mode. Fix: Include dark mode color variants or document that the brand is light-mode-only.

5. No guardrails. Without explicit rules, agents will improvise — and sometimes generate content that's off-brand, legally risky, or factually wrong. Fix: Clear tiered guardrails: acceptable, review, never.

6. Design system as PDF. If your design system is a Figma file or PDF, AI agents can't read it. Fix: DESIGN.md in markdown, version controlled, linked from all agent configuration files.

Execution Artifacts

• references/framework-notes.md — named frameworks, citation anchors, and operating assumptions
• templates/output-template.md — copy-paste deliverable structure for the user
• scripts/check-output.py — local checklist validator for required sections This skill includes lightweight artifacts the agent can load on demand: Use the artifacts when the user asks for an implementation-ready deliverable, a repeatable workflow, or a quality check rather than generic advice.

Related Skills

• claude-design — Design process for one-off HTML artifacts
• popular-web-designs — 54 known-brand design systems for visual vocabulary
• design-system-builder — Build complete design systems from references
• brand-kit — Logo design, color palette, typography, asset templates
• ui-ux-gtm — Landing pages, forms, signup flows, dashboards
• effective-ui-design — Professional UI design guidelines