Dictionary
Design vocabulary, key terms, techniques, and AI language for designers.
34 resources
Agentic Design System
A design system built for two audiences: humans and AI agents. Tokens an agent can parse, components with documented intent, and rules coding tools actually respect.
AI Agent
An AI that does not just answer, it acts: reads files, runs tools, makes changes, and works toward a goal across multiple steps with limited supervision.
Animation Vocabulary
The working vocabulary of UI motion. Around 80 terms for entrances, easing, springs, transitions, scroll, and performance, so you can name the exact motion you want instead of asking for 'smoother'.
Context Window
Everything the AI can 'see' at once: your prompt, the files, the images, and the conversation so far. When it fills up, older details fall out of view.
Design Tokens
The smallest units of a design system. Named values that store visual design decisions like colors, spacing, typography, and shadows.
LLM (Large Language Model)
The kind of AI behind Claude, ChatGPT, and Gemini. A model trained to predict text, which makes it good at writing, coding, and reasoning over language.
MCP (Model Context Protocol)
A standard that helps AI assistants connect to tools like Figma with rich context. The bridge between your design files and AI coding tools.
Chromatic
A visual regression testing service built for Storybook. Catches unintended UI changes before they ship.
CLAUDE.md
A markdown file that teaches Claude Code about a project. Turns generic AI output into project-aware AI output.
Context Engineering
Deciding what an AI agent knows when it works: which files, rules, and examples load into its context window, and when. The discipline behind CLAUDE.md, skills, and MCP.
Drift (Design System Drift)
The slow gap that opens between your design system as documented and your product as shipped: hardcoded values, rogue variants, one-off components nobody approved.
Design System Observatory
A monitoring dashboard that tracks the health of your design system. Measures token coverage, component adoption, accessibility, and design-code parity.
Environment Variables (.env files)
A .env file is a small text file that holds secrets like API keys outside your code. Your app reads from it at runtime. Never commit it to Git.
GitHub Actions
GitHub's built-in CI/CD platform. Runs workflows in response to events like a push, a pull request, or a schedule.
Guardrails
Hard limits on what an AI agent can do, enforced by the system rather than the prompt: scoped permissions, protected files, required reviews, and kill switches.
Hallucination
When an AI states something false with full confidence: a made-up API, a token that does not exist, a citation that was never written. Fluent, plausible, and wrong.
Human-in-the-Loop
A workflow where an AI agent does the work but a human approves it before it counts: the agent proposes, you decide. The standard pattern for mid-trust agents.
Observer Mode
Running an AI agent read-only against your real design system before giving it any authority. The agent audits and reports; it cannot change anything.
Personal Access Token (PAT)
A secret string that authenticates a user with a service like GitHub or Figma in place of a password.
Playwright
A browser automation tool for end-to-end tests, visual regression, and automatic component screenshots.
Primitive vs Semantic Tokens
The difference between raw value tokens and purpose-driven tokens. Understanding this distinction is essential for scalable design systems.
Prompt
The instruction you give an AI. The clearer and more specific it is, the better the result. A system prompt sets the standing rules for the whole session.
RAG (Retrieval-Augmented Generation)
A technique that lets an AI look things up before it answers. It retrieves the relevant documents first, then generates from them, instead of relying on memory.
Self-Healing Flywheel
A loop where agents detect design system drift, fix it at the source, and add a rule so the same drift cannot happen again. The system gets stronger with every cycle.
Skill (Claude Code)
A reusable instruction file that teaches Claude how to do one job your way, invoked with a slash command. Your taste, your review process, your rules, made portable.
Storybook
A tool for developing, documenting, and testing UI components in isolation. The default home for design system component docs.
Style Dictionary
An open-source build system that transforms design tokens from JSON into any platform format (CSS, iOS, Android, and more).
Subagent
A second AI agent spun up to handle one focused task in its own context, then report back. Used to search, review, or work in parallel without cluttering the main session.
Supabase
A backend platform with database, auth, and storage features. A pre-built infrastructure kit for web applications.
Tokens (the AI kind, not design tokens)
The small chunks of text an AI reads and writes, and what your usage is measured in. Not to be confused with design tokens, which are reusable style values.
Tokens Studio
A Figma plugin that manages design tokens inside Figma and syncs them to GitHub, GitLab, or other sources of truth.
Trust Levels
A five-step scale (Observer, Advisor, Junior, Senior, Autonomous) that defines what an AI agent is allowed to do, and what it has to prove before it gets more authority.
Vercel
A deployment and hosting platform for modern web applications. The 'publish to web' button with strong defaults.
Vibe Coding
Building software by describing what you want in plain language and letting the AI write the code, steering by feel and result rather than by reading every line.