Skip to content

OpenClaw Carapace - Setup Guide

Source: openclaw/carapace (2.3K combined installs) GitHub: github.com/openclaw/carapace (4⭐) Category: Design System / UI Engineering First Seen: July 21, 2026 Security: Gen Agent Trust Hub Pass · Socket Pass (Warn on openclaw-design) · Snyk Pass

Carapace is the official OpenClaw design system, packaged as six focused skills. Instead of one monolithic design rulebook, each skill owns one surface. A router skill (openclaw-design) dispatches to the correct branch so the agent only loads what the task needs.


Installation

# Install all six skills from the repo
npx skills add https://github.com/openclaw/carapace

# Or install individually
npx skills add https://github.com/openclaw/carapace --skill openclaw-design
npx skills add https://github.com/openclaw/carapace --skill openclaw-brand
npx skills add https://github.com/openclaw/carapace --skill openclaw-carapace
npx skills add https://github.com/openclaw/carapace --skill openclaw-marketing-pages
npx skills add https://github.com/openclaw/carapace --skill openclaw-design-audit
npx skills add https://github.com/openclaw/carapace --skill openclaw-design-system

The Six Skills

Skill Installs Use For
openclaw-design 2,075 Router - chooses the focused branch before changing any interface
openclaw-brand 379 Identity: typography, logos, imagery, voice, non-product brand artifacts
openclaw-carapace 1,725 App UI: semantic tokens, themes, component reuse, framework adapters
openclaw-marketing-pages 2,072 Public pages: landing/content composition, navigation, SEO, responsive layout
openclaw-design-audit 379 Design drift, token misuse, component substitution, accessibility, recurring audits
openclaw-design-system 2,070 v0.1.x compatibility alias for projects upgrading an existing skill lock

Routing rule: for a public website change start with openclaw-marketing-pages, adding openclaw-brand only when the task changes identity, logo, imagery, typography, or voice. For a product application start with openclaw-carapace. Load multiple branches only when the task genuinely crosses them.


Working With openclaw-carapace

The core skill enforces a read-then-build workflow over five reference files:

  1. references/tokens.md - read before choosing colors, spacing, type, radii, or shadows
  2. references/consumer-adapters.md - current framework adapter conventions
  3. references/application-surfaces.md - shells, panes, settings, operational screens
  4. references/terminal-ui.md - terminal interface design and auditing
  5. references/embedded-surfaces.md - surfaces rendered inside host frames (MCP apps)

Core rules: use semantic tokens for UI intent and palette primitives only for documented exceptions; keep application behavior, routes, and information architecture unchanged unless the task says otherwise; validate affected routes with existing tests and real browser screenshots; inspect the consumer's existing shared primitives before creating a new component.


Prerequisites

Requirement Details
Node.js + npx For the skills.sh CLI install path
OpenClaw or Clawd runtime Skills target OpenClaw interfaces; guidelines are reusable in Hermes UI work
Git access For direct clone installs

Direct clone alternative:

git clone --depth 1 https://github.com/openclaw/carapace.git /tmp/carapace
cp -r /tmp/carapace/openclaw-* ~/.clawd/skills/

CorpusIQ Use Cases

Use Case How
UI surface audits Run openclaw-design-audit methodology against agent-built dashboards to catch token misuse and drift
Landing page composition openclaw-marketing-pages for SEO-aware, responsive public pages
Terminal UI polish terminal-ui.md reference when building or auditing TUI surfaces
Brand consistency openclaw-brand when identity, voice, or imagery is involved in generated copy
Framework adapter decisions consumer-adapters.md before choosing React/Vue/vanilla adapters in agent projects

Integration with CorpusIQ

CorpusIQ is the connector layer that lets AI assistants read the systems a business already runs on - accounting, payments, CRM, email, files and analytics - and answer with the source named. It is read-only by design: it retrieves data and never writes to, or acts in, a connected system, which is what makes it safe to point at live accounts.

This page is part of the CorpusIQ directory, the reference we maintain for the MCP server and agent-skill ecosystem. Run alongside CorpusIQ, a third-party server like this one can be combined with a company's own business data in a single answer.

See what CorpusIQ connects

Limitations / Verification

  • Low GitHub star count (4⭐) and first seen July 21, 2026 - young package, follow release cadence
  • openclaw-design-system is a compatibility alias, not new functionality - install only when upgrading old locks
  • Verify install: npx skills list | grep -i carapace should show six entries
  • Security posture is clean on all three scanners (Socket flags one Warn on the router skill)

Powered by CorpusIQ

AI answers on your data. See where every number comes from.

Connect your tools and ask in ChatGPT, Claude, or Perplexity. Enterprise-grade answers in minutes. From $29.95/month.

Start free, no credit card
Contact Terms Privacy Pricing