By cmj-hub
Grep-strength brand-voice scanner against a configurable JSON profile. Flags banned phrases, stale stat-claims, and missed reframes. Pure static analysis — no LLM, no paid APIs.
This skill should be used when auditing a Next.js or React project for design-system drift — CTA color violations, heading hierarchy mistakes, missing alt text, container pattern breaks, or semantic-color hygiene. Trigger phrases include "audit the design", "check design system", "design drift", "find CTA color violations", "heading hierarchy check", "design audit". Reads every page.tsx (or configurable glob) and reports findings by severity. Deterministic, no LLM, no paid APIs.
This skill should be used when checking a deployed or local web app for layout regressions, horizontal overflow, element clipping, or breakage across viewports. Trigger phrases include "visual QA", "check for overflow", "layout check", "responsive bug", "test mobile layout", "playwright overflow". Generates a Playwright spec that crawls a sitemap and checks every route across mobile (375px), tablet (768px), and desktop (1440px) viewports. Reports horizontal overflow, off-screen fixed elements, and zero-height containers. Deterministic, no LLM, requires Playwright as a peer dep.
This skill should be used when scanning a repo for tone violations, banned phrases, stale stat-claims, or missed brand-voice reframes against a configurable brand-profile JSON. Trigger phrases include "voice sweep", "tone check", "brand-voice audit", "find banned phrases", "stale stat claims", "find old positioning". Pure static scan; no LLM, no paid APIs. Operates on any text source — Markdown, MDX, TSX, JSX, TS, JS — and reports per-file findings classified by severity.
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
Cross-runtime agent skills for shipping things, not preparing to ship things.
Each skill in this monorepo is a single capability — a specific quality gate, audit, or sweep — packaged as a plain Node 18+ CLI first, with Claude Code (Agent Skills spec) and MCP as additional install paths. No LLM calls, no paid APIs, no remote dependencies by default.
| Skill | What it does | Status |
|---|---|---|
design-audit | Reads every page.tsx and checks it against a design-system ruleset — CTA colors, heading hierarchy, container patterns, semantic-color hygiene. | Wave 1 |
visual-qa | Crawls a sitemap and runs Playwright across 3 viewports (375 / 768 / 1440) to detect horizontal overflow, clipping, and layout breakage. | Wave 1 |
voice-sweep | Scans a repo for tone violations against a configurable brand-profile — flags banned phrases, stale stat-claims, and missed reframes. | Wave 1 |
All three are pure static / browser analysis. They do not call OpenAI, Anthropic, or any external paid API. The Voice Sweep brand-profile is a plain JSON file you own. Bring your own rules.
# Per-skill
npm i -D @operator-skills/design-audit
npm i -D @operator-skills/visual-qa
npm i -D @operator-skills/voice-sweep
Then in package.json:
{
"scripts": {
"audit:design": "operator-design-audit",
"audit:visual": "operator-visual-qa",
"audit:voice": "operator-voice-sweep"
}
}
Each CLI is zero-config out of the box (works on Next.js App Router by
default) and accepts a config file (<skill>.config.json) at the repo
root for project-specific overrides.
/plugin marketplace add cmj-hub/operator-skills
/plugin install design-audit
/plugin install visual-qa
/plugin install voice-sweep
This installs all three skills via the marketplace manifest in
.claude-plugin/marketplace.json.
Claude Code reads each skill's SKILL.md and surfaces them based on
their trigger phrases. The skill body shells out to the same CLI under
the hood.
Each skill ships an MCP server entrypoint via fastmcp.
Wire into your runtime's MCP config:
{
"mcpServers": {
"design-audit": { "command": "npx", "args": ["-y", "@operator-skills/design-audit", "mcp"] },
"visual-qa": { "command": "npx", "args": ["-y", "@operator-skills/visual-qa", "mcp"] },
"voice-sweep": { "command": "npx", "args": ["-y", "@operator-skills/voice-sweep", "mcp"] }
}
}
Each CLI accepts a --remote-config <url> flag (and a matching
<skill>RemoteConfig field in the config file) that fetches a JSON
ruleset from a URL on demand. Use cases:
design-rules.json at your org
URL; every project pulls the latest at audit time.These connections are opt-in, no-LLM, and the JSON schema is
documented in each skill's references/remote-config.md. Default
behavior is fully offline.
| Runtime | Status |
|---|---|
| Plain Node 18+ CLI | Verified Wave 1 |
| Claude Code (Agent Skills spec) | Verified Wave 1 |
| MCP (fastmcp) | Verified Wave 1 |
| Cursor | Planned Wave 2 |
| Gemini CLI | Planned Wave 2 |
| Codex | Planned Wave 2 |
Cross-runtime ports for Cursor / Gemini / Codex are gated on each runtime's plugin model stabilizing. The Claude Code and MCP variants work in those runtimes today via MCP bridges.
operator-skills/
├── .claude-plugin/marketplace.json # Claude Code plugin marketplace manifest
├── .github/workflows/ci.yml # build + validate + publish dry-run
├── packages/ # shared internal packages (validators, registry)
├── skills/
│ ├── design-audit/
│ │ ├── SKILL.md # Agent Skills spec entrypoint
│ │ ├── package.json # npm publishable
│ │ ├── src/ # TypeScript source — CLI + MCP shim
│ │ ├── bin/ # CLI shim (operator-design-audit)
│ │ └── references/ # rule docs (loaded into context as needed)
│ ├── visual-qa/
│ └── voice-sweep/
├── pnpm-workspace.yaml
└── tsconfig.base.json
npx claudepluginhub cmj-hub/operator-skills --plugin design-auditUltra-compressed communication mode. Cuts ~75% of tokens while keeping full technical accuracy by speaking like a caveman.
Frontend design skill for UI/UX implementation
Comprehensive UI/UX design plugin for mobile (iOS, Android, React Native) and web applications with design systems, accessibility, and modern patterns
Memory compression system for Claude Code - persist context across sessions
Marketing skills for AI agents — conversion optimization, copywriting, SEO, paid ads, ad creative, and growth
Standalone image generation plugin using Nano Banana MCP server. Generates and edits images, icons, diagrams, patterns, and visual assets via Gemini image models. No Gemini CLI dependency required.