Multi-runtime design operating system and CLI for Claude Code, OpenAI Codex, Gemini CLI, and Google Antigravity.
Install once to equip any AI coding agent with 12+ domain design skills, deterministic verification gates, 2:1 axonometric art direction, and live Figma / SVG / 3D / SwiftUI / packaging workflows.
DESIGN:OS is free to use and modify. Source-available under PolyForm Shield 1.0.0. Install with npx @easestart/design-os init.
Quickstart · What it makes · Install · How it works · Commands · Limits
AI coding agents write functional code quickly, but their default visual output suffers from inconsistent spacing, hallucinated tokens, rainbow badges, and zero verification gates. DESIGN:OS turns aesthetic taste and design systems engineering into deterministic, runtime-agnostic contracts.
| Dimension | Generic AI / Manual Workflow | DESIGN:OS CLI (DESIGN:OS) |
|---|---|---|
| Visual Quality | Generic purple-gradient AI slop and ad-hoc CSS | Token-locked typography, 8pt modular grid, and 2:1 axonometric art direction |
| Context Budget | Dumping massive style guides into every prompt | Progressive skill disclosure (SKILL.md router + on-demand domain packs) |
| Verification | Visual eyeballing with zero automated checks | Deterministic CLI gates (self_check.py, contrast, and anti-flop audits) |
| Multi-Agent Parity | Locked to a single IDE or proprietary prompt format | One canonical skill tree across Claude Code, Codex, Gemini CLI & Antigravity |
| Privacy & Telemetry | Opaque cloud telemetry or unverified third-party APIs | Zero telemetry — 100% local execution (PRIVACY.md audited) |
In Claude Code:
/plugin marketplace add jangtrinh/design-os
/plugin install design-os@design-os
Or from your terminal (Codex, Gemini CLI, Antigravity, OpenCode, or standalone CLI):
git clone https://github.com/jangtrinh/design-os.git
cd design-os
npx @easestart/design-os initOpen your project in your AI coding agent and describe what you need:
"Initialize DESIGN:OS in this repo and wire our design tokens and verification gates for Claude Code and Codex."
"Audit our current frontend components against DESIGN:OS typography, spacing, and WCAG 2.2 AA contrast gates."
"Scaffold a new product landing page with 2:1 axonometric B&W technical illustrations and SEO/AEO/GEO metadata."
python3 scripts/self_check.pyEvery deliverable produced by design-os is backed by a deterministic verification gate rather than freehand AI guesswork.
Prompt: "Initialize DESIGN:OS in this repo and wire our design tokens and verification gates for Claude Code and Codex."
| Deliverable / Mode | Best for | Output Format | Verification Gate |
|---|---|---|---|
| Multi-Runtime Skill Pack | Wiring Claude Code, Codex, Gemini CLI, and Antigravity to one design standard | .claude-plugin + skills/ |
python3 scripts/self_check.py |
| Design System Tokens & Gates | Enforcing spacing, typography, and WCAG 2.2 AA contrast in CI | JSON / CSS / CLI exit code |
Zero contrast or grid violations |
| Cross-Engine Orchestration | Routing tasks to Figma, SVG Animation, 3D Blender, Apple SwiftUI, or Origami CAD | Verified domain artifacts |
13-engine ecosystem parity |
| Design-System Guidelines Pack | Teaching any design agent the rules a library cannot show — owner rulings, look-alike components, measured anti-patterns — and shipping them as a Figma library guidelines bundle | design/guidelines/*.md + figma-guidelines/ |
ui ds guidelines check — no invented names, no tampered facts, Figma's 8-file / 200 KB limits |
 architecture — Three-tier web application |
 medallion — Survey dataset storage tiers |
 bar — Revenue by region, FY2025 |
 quadrant — Q3 initiatives by effort and impact |
All 28 grammars rendered — every one of these is a committed artifact that passes its linter with zero findings
 data-flow |
 dp-integration |
 dp-security-matrix |
 er |
 flowchart |
 high-level |
 it-state |
 layers |
 loop |
 nested |
 org-chart |
 process |
 product-flow |
 sequence |
 state |
 swimlane |
 tree |
 gantt |
 line |
 pyramid |
 radar |
 scatter |
 timeline |
 venn |
Choose the setup that matches your AI coding agent. Every option installs the same canonical skill (skills/design-os/SKILL.md) and /doctor diagnostic suite.
Run these two commands inside Claude Code:
/plugin marketplace add jangtrinh/design-os
/plugin install design-os@design-os
Install to your personal Codex skills directory:
git clone https://github.com/jangtrinh/design-os.git /tmp/design-os
mkdir -p ~/.codex/skills/design-os ~/.codex/prompts
cp -R /tmp/design-os/skills/design-os/* ~/.codex/skills/design-os/
cp /tmp/design-os/prompts/doctor.md ~/.codex/prompts/design-os-doctor.md
rm -rf /tmp/design-osgit clone https://github.com/jangtrinh/design-os.git /tmp/design-os
mkdir -p ~/.gemini/skills/design-os
cp -R /tmp/design-os/skills/design-os/* ~/.gemini/skills/design-os/
rm -rf /tmp/design-osgit clone https://github.com/jangtrinh/design-os.git /tmp/design-os
mkdir -p ~/.agents/skills/design-os
cp -R /tmp/design-os/skills/design-os/* ~/.agents/skills/design-os/
rm -rf /tmp/design-osVerify or uninstall
Verify installation:
python3 scripts/self_check.pyOr run /doctor inside Claude Code / Codex to inspect installed skill paths, plugin manifests, and gate health.
Uninstall:
rm -rf ~/.claude/skills/design-os ~/.codex/skills/design-os ~/.gemini/skills/design-os ~/.agents/skills/design-osOnce installed, skills/design-os/SKILL.md activates automatically when your prompt matches its domain triggers. Browse docs/cookbook.md for copy-paste prompt recipes.
Generate & Scaffold
"Initialize DESIGN:OS in this repo and wire our design tokens and verification gates for Claude Code and Codex."
Audit & Verify
"Audit our current frontend components against DESIGN:OS typography, spacing, and WCAG 2.2 AA contrast gates."
Export & Automate
"Scaffold a new product landing page with 2:1 axonometric B&W technical illustrations and SEO/AEO/GEO metadata."
| Command | Environment | What it does |
|---|---|---|
/doctor |
Claude Code / Codex | Read-only diagnostic check of installed skill paths, plugin manifests, and runtime prerequisites |
python3 scripts/self_check.py |
Terminal / CI | Deterministic offline verification of manifests, skill contracts, and governance files |
npx @easestart/design-os doctor |
Terminal / CI | Inspect active runtime adapters (Claude Code, Codex, Gemini CLI, Antigravity) and installed skill packs |
ui taste-lint — 35 absolute checks |
Terminal / CI | Generated-UI taste checks across motion, typography, spacing, and focus states |
ui validate-layout — 20 checks |
Terminal / CI | Structural and overflow safety checks (100vw traps, unclosed tags, sticky overflow) |
ui content-lint — 15 checks |
Terminal / CI | Honest copy checks (placeholder text, lorem-ipsum, generic names, all-caps shouting) |
ui tell-lint — 43 checks, any language |
Terminal / CI | Polyglot design-tell checks across HTML, JSX/TSX, Vue, Svelte, CSS, SwiftUI, and Flutter |
ui ds guidelines build | check | export --target figma |
Terminal / CI | Compile a cited guidelines pack from your sealed design system, rulings and glossary; lint it; export a ready-to-upload Figma library guidelines bundle |
skills/design-os-designer — build.mjs, validate.mjs, look.mjs |
Any agent runtime | The shipped design skill: build an evidence folder for a page, run ui gate plus file checks, render a three-width sheet with measured items; ui init wires it into Claude Code, Antigravity and Codex |
- Intent Routing — Your AI agent matches your request against
skills/design-os/SKILL.mdfrontmatter triggers. - Progressive Context Loading — Only the entry-point
SKILL.mdrouter loads initially; detailed recipes and scripts load on demand. - Deterministic Execution — Domain scripts and contracts execute locally with reproducible inputs and outputs.
- Gate Verification —
scripts/self_check.pyand domain verifiers (npx @easestart/design-os init) validate the artifact before completion. - Zero-Telemetry Delivery — Verified files are written directly to your workspace with zero external network calls.
User Prompt / CLI Invocation
│
▼
┌───────────────────────────────────────────────┐
│ skills/design-os/SKILL.md
│ Progressive disclosure router & hard rules │
└───────────┬───────────────────────────────────┘
│ loads on demand
▼
┌───────────────────────────────────────────────┐
│ npx @easestart/design-os doctor
│ Deterministic build & contract enforcement │
└───────────┬───────────────────────────────────┘
│ verifies
▼
┌───────────────────────────────────────────────┐
│ python3 scripts/self_check.py + /doctor │
│ Zero-telemetry local verification gate │
└───────────────────────────────────────────────┘
design-os uses progressive context disclosure so your agent's context window stays fast and focused:
| File | Loaded when |
|---|---|
skills/design-os/SKILL.md |
Always (~100–150 lines) — routing rules, triggers, and execution workflow |
docs/cookbook.md |
When selecting multi-step prompt recipes or domain patterns |
commands/doctor.md / prompts/doctor.md |
Only when running /doctor diagnostics |
llms.txt |
When indexed by AI answer engines, RAG pipelines, or repo crawlers |
design-os/
├── .claude-plugin/
│ ├── plugin.json # Claude Code plugin manifest
│ └── marketplace.json # Claude Code marketplace catalog
├── .codex-plugin/
│ └── plugin.json # OpenAI Codex plugin & interface manifest
├── .agents/
│ └── plugins/marketplace.json # Cross-agent local plugin registry
├── .factory-plugin/
│ ├── plugin.json # Factory / Droid plugin manifest
│ └── marketplace.json # Factory marketplace metadata
├── skills/
│ └── design-os/
│ └── SKILL.md # Canonical agent skill entrypoint
├── commands/
│ └── doctor.md # /doctor diagnostic slash command
├── prompts/
│ └── doctor.md # Codex diagnostic prompt mirror
├── scripts/
│ └── self_check.py # Deterministic repository & skill self-check
├── docs/
│ ├── cookbook.md # Copy-paste prompt recipes by job
│ └── llms.txt # GEO / LLM discovery index
├── PRIVACY.md # Zero-telemetry & local-first privacy commitment
├── SECURITY.md # Vulnerability disclosure & security policy
├── CONTRIBUTING.md # Contribution workflow & verification rules
├── CODE_OF_CONDUCT.md # Contributor Covenant code of conduct
├── CHANGELOG.md # Release history
├── llms.txt # Root LLM indexing manifest
├── LICENSE # PolyForm Shield 1.0.0
└── README.md
python3 scripts/self_check.pyprintsok: design-os self-check passed (design-os)and exits0.npx @easestart/design-os initpasses with zero contract or test regressions.- Your AI coding agent loads
skills/design-os/SKILL.mdautomatically when given any prompt fromdocs/cookbook.md. - Network traffic during execution remains zero (guaranteed by
PRIVACY.md).
- Throwaway backend scripts with zero user-facing UI, CLI, or documentation surface.
- Teams looking for a hosted cloud SaaS design tool rather than a local-first, git-tracked engineering toolchain.
- Replacing human brand strategy — DESIGN:OS enforces and scales design systems, not customer discovery.
DESIGN:OS is free and source-available (PolyForm Shield). The Autonomous Design Agent talks from the design:os by Jang channel (Vietnamese, with English terms):
- Xây dựng autonomous design agent cho doanh nghiệp? how? (7:06): How the DESIGN:OS CLI, plugin bridge and deterministic gates turn an ontology into an autonomous design agent for enterprise design systems. Vietnamese, with English terms.
- Ontology là gì? apply vào build AI design agent như thế nào? (7:33): What an ontology is and how it grounds an AI design agent: shared vocabulary, semantic RAG and reusable design decisions. Vietnamese, with English terms.
- Autonomous Design Agent · Tập 03: Model + Harness và 6 trụ cột DESIGN:OS (8:44): Agent = Model + Harness: the six pillars of DESIGN:OS (single source of truth, Ontology Graph RAG, dual-layer spec, 126 CI gates, 9-verb patch, watchers). Vietnamese, with English terms.
Autonomous Design Agent (ADA) short series, six short episodes (Vietnamese, with English terms; English-titled playlist: ADA (English), Vietnamese playlist: ADA · Tiếng Việt):
- Storybook + W3C DTCG Tokens: Hard Boundaries for Designers and AI Agents (3:10)
- Ontology Graph RAG: Why AI Picks the Wrong UI Component (3:02)
- Spec-Driven Development: Lock the Spec Before AI Writes Code (2:25)
- Harness 2.0: The Control Frame That Keeps AI Agents Out of Endless Loops (2:43)
- Two-way bridge: One Comment, the AI Fixes Design and Code (2:44)
- Watchers, a 2-Mac Cluster and PII Guard: Infrastructure for Long-Running AI Agents (2:06)
More: Autonomous Design Agent series.
design-os is one of the 13 source-available engines in the DESIGN:OS ecosystem:
- Contributing: See
CONTRIBUTING.mdfor pull request guidelines and verification gates (python3 scripts/self_check.py). - Code of Conduct: Governed by
CODE_OF_CONDUCT.md. - Security & Privacy: Zero telemetry — see
SECURITY.mdandPRIVACY.md. - Changelog & LLM Index: Track releases in
CHANGELOG.mdand machine-readable index inllms.txt. - License: PolyForm Shield 1.0.0 © Jang Trinh. Free to use and modify, including for paid client work; you may not resell it or build a competing product from it.
If design-os saves you time, a star on GitHub helps other engineers and AI agents find it.
