Grounded Design System

Makes a team's design system AI-ready inside their own repository, so agents building UI read answers instead of guessing. Audits where the system is silent, sets up the context layer, catalog, enforcement and gap log one phase at a time, logs missing pieces instead of inventing around them, and runs readiness tests. Use when the user asks to make a design system AI-ready or agent-ready, audit a design system or component library for agents, write or fix AGENTS.md, CLAUDE.md or DESIGN.md for UI work, add component metadata or a registry, enforce tokens or lint rules for agents in CI, log a mis

9 starsFrontend & Design Plugin marketplace

Install

Claude Code plugin
/plugin marketplace add noemuch/skills

The repo is a Claude Code plugin marketplace. Run /plugin afterwards to pick the plugin that includes this skill.

skills CLI (Claude Code, Cursor, Codex and more)
npx skills add noemuch/skills --skill grounded-design-system
Manual install
git clone --depth 1 https://github.com/noemuch/skills.git /tmp/skills
mkdir -p ~/.claude/skills
cp -r /tmp/skills/skills/grounded-design-system ~/.claude/skills/grounded-design-system

Personal skills live in ~/.claude/skills. Use .claude/skills in a project folder to share a skill with your team.

From the SKILL.md

# Grounded design system

This skill makes a design system answer the three questions every agent building UI asks, in the team's own repository and stack. It audits, sets up, logs gaps and tests, one action per run.

## The doctrine in eight lines

1. An agent building UI asks: **What am I using here? What am I not allowed to do? What happens when the answer does not exist?** Every artifact this skill produces answers one of the three.
2. A gap is `missing` (the decision was never made), `disconnected` (it exists where an agent cannot reach it: a Figma comment, a Slack thread, someone's head), `stale` (superseded or generated, yet the most findable answer) or `violated` (in the repository, but bypassed or unenforced). An area scores `answered` only when the repository states, keeps current and follows the decision; a probe that cannot run read-only scores `not determined`.
3. Flag, don't invent. A blocker honestly reported is a good outcome. The patch is the failure.
4. Documentation for intent, enforcement for anything expensive to get wrong. If an agent breaking a rule would take a reviewer more than a minute to catch, the rule leaves prose and moves into types, lint, tests, hooks or CI.
5. A rule is mechanical: a binary condition, a grep-able pattern, a number or an explicit list. Write down only what an agent gets wrong by default.
6. Clarity before completeness: fix 3 to 5 gaps per pass, never all of them.
Read the full skill on GitHub

Details

Skill name
grounded-design-system
Repository
noemuch/skills
Path
skills/grounded-design-system/SKILL.md
License
MIT
Repo updated
4 October 2026

FAQ

What does the Grounded Design System skill do?

Makes a team's design system AI-ready inside their own repository, so agents building UI read answers instead of guessing. Audits where the system is silent, sets up the context layer, catalog, enforcement and gap log one phase at a time, logs missing pieces instead of inventing around them, and runs readiness tests. Use when the user asks to make a design system AI-ready or agent-ready, audit a design system or component library for agents, write or fix AGENTS.md, CLAUDE.md or DESIGN.md for UI work, add component metadata or a registry, enforce tokens or lint rules for agents in CI, log a mis

How do I install the Grounded Design System skill?

Claude Code plugin: /plugin marketplace add noemuch/skills. Skills go in ~/.claude/skills (personal) or .claude/skills (project), and Claude loads them automatically when a task matches the description.

Is Grounded Design System popular?

It lives in noemuch/skills, which has 9 GitHub stars and 0 forks.

Related skills

Gpui

9

AprilNEA/gpui-skills

GPUI UI framework best practices for building desktop applications. Use when writing GPUI code, creating UI components, handling state/events, async tasks, animations, lists, forms, testing, or working with Zed-style Rust GUI code.

Frontend & Design

Figjam Builder

9

prasantloki/figjam-builder

Build or modify FigJam boards via the Figma Plugin API. Use for creating, editing, extending, restructuring, or fixing the layout of a FigJam board. Triggers on "make a FigJam", "build a board", "update this FigJam", "fix the layout", "/figjam-builder", or a FigJam URL paired with a visual or content change.

Frontend & Design

Gpui Component

9

AprilNEA/gpui-skills

gpui-component library patterns for building reusable UI components. Use when creating buttons, inputs, dialogs, forms, or following Longbridge component library conventions in GPUI applications.

Frontend & Design

Elec Sheep Longform

9

bleakbelladonnals/elec-sheep-longform

Create, revise, and quality-check Elec Sheep branded Chinese educational carousels and vertical long-form infographics while preserving the exact robot-sheep IP, warm hand-drawn blue-orange visual system, editable SVG typography, and batch rendering workflow. Use when the user mentions 电子羊, Elec Sheep, the established mascot/IP, 小红书科普长图, an 8-page knowledge carousel, same-character/same-style continuation, or asks to revise existing Elec Sheep pages without identity or layout drift.

Frontend & Design

Rice Calculator Coach

9

Grafuja/Product-Manager-Skills

Interactive RICE (Reach, Impact, Confidence, Effort) calculator for product initiatives. Use this skill whenever the user wants to calculate RICE scores, prioritize features, rank initiatives, compare backlog items, or asks "how do I prioritize" or "what should I build first". Also trigger when user mentions RICE framework, scoring initiatives, or needs help with prioritization decisions. Guide them through each component step-by-step with examples and context.

Frontend & Design

Designgate

8

sukenik-collab/claude-designgate

Mandatory human-in-the-loop UI design checkpoint. Use automatically — before writing any UI code — whenever the task involves building, creating, adding, or redesigning a user-facing screen, page, flow, component, interface, or frontend, unless every interaction maps to an already-decided pattern. Verifies a brand brief exists, reads prior UX decisions to avoid re-asking resolved questions, surfaces only genuinely unresolved interaction decisions (batched, max 5), generates a complete screen-generation prompt for the configured design MCP, and enforces a hard human approval gate before impleme

Frontend & Design

Browse the most popular Claude skills, all frontend & design skills, or write your own with the skills generator.