Design System Playbook

Design Token Architecture For Multi-Brand Systems

This guide explains the architecture behind the DS Genie and why its output is stable at enterprise scale. You will learn how token layers reduce duplication, improve governance, and accelerate design-to-dev workflows.

Context: this is the logic used by the generator workflow on the main app. Read it before Step 2 if you need to align team conventions.

Architecture Model

A robust token system separates responsibilities into four layers:

1) Primitive

Raw values only: color ramps, spacing, radius, type sizes, motion duration. No semantic intent.

2) Theme

Light and dark behavior. Surface, text, and border mappings that preserve contrast rules per mode.

3) Semantics

Intent layer per brand mode: primary action, success, warning, error, and neutral states.

4) Components

Component aliases (button, input, list, tab) linked to semantics and theme, not directly to primitives.

Why Teams Break Token Systems

Anti-pattern: direct primitive links

Components mapped directly to raw values create fragile dependencies. One rebrand becomes a full UI rewrite.

Anti-pattern: semantic drift

Inconsistent naming and intent across squads makes governance impossible and slows release confidence.

How This Maps To The Generator

Generator Step What You Decide Architecture Impact
Step 1: Brand Count How many brand modes you need Defines semantic mode scalability
Step 2: Configure Color, type, radius, advanced options Builds primitive and semantic relationships
Step 3: Preview Component groups and token behavior Validates alias strategy before export
Step 4: Download Plugin ZIP export Creates deterministic Figma variables setup

Execution Checklist

Naming

Use stable token paths and avoid visual-only names. Favor intent over appearance.

Accessibility

Keep contrast targets in token logic, not in manual component overrides.

Governance

Version token collections and review semantic changes like API changes.

Adoption

Train teams on semantic usage patterns before scaling component libraries.

FAQ

Can I skip semantic tokens?

You can, but you will lose clean brand switching and increase maintenance cost over time.

Is this only for large companies?

No. Small teams benefit early because architecture debt grows fast without clear token layers.

Next action: generate your plugin.

Open DS Genie and apply this architecture in a real workflow. The output ZIP can be imported directly into Figma.

Open the plugin generator