Installable Claude skills that punch above their weight.
Senior-grade, reference-backed skills you can drop into any project — install the whole marketplace or copy one skill.
Install • Skills • What's inside • Contributing • License
ghost-skills is a marketplace of installable Claude skills. Each skill is self-contained, reference-backed, and follows the Agent Skills open standard — so it works in Claude Code (CLI, desktop, IDE) and any compatible tool.
Add the marketplace once and install any skill with a single command, or copy a skill folder straight into your project.
| Skill | What it does |
|---|---|
| senior-ux-audit | A principal-level UI/UX audit (web · iOS · Android · desktop) from your codebase — heuristics, accessibility, performance vitals, and a severity-rated, prioritized report with a 30/60/90 roadmap. |
| agentic-ai-backend-audit | A senior adversarial audit of an AI-agent-built / agentic backend — 18 lenses across access control, prompt & agent safety, PII & data leakage, reliability, observability/evals/governance, and AI code smells. Severity-rated, file:line-cited, ends in a Ship/Fix/Block verdict. |
More skills are on the way. See CONTRIBUTING.md to add one.
Pick whichever fits your setup.
Works with SkillFish or Vercel skills. One command detects the agents on your system (Claude Code, Cursor, Copilot, …) and installs the skill into each:
npx skillfish add CreatorGhost/ghost-skills
# or
npx skills add CreatorGhost/ghost-skillsList what the repo offers, or target one skill directly:
npx skills add CreatorGhost/ghost-skills --list
npx skills add https://github.com/CreatorGhost/ghost-skills/tree/main/plugins/senior-ux-audit/skills/senior-ux-audit/plugin marketplace add CreatorGhost/ghost-skills
/plugin install senior-ux-audit@ghost-skills
Or from a local clone:
/plugin marketplace add /path/to/ghost-skills.
cp -r plugins/senior-ux-audit/skills/senior-ux-audit ~/.claude/skills/ # all projects
cp -r plugins/senior-ux-audit/skills/senior-ux-audit .claude/skills/ # this project onlyInvoke it — directly, or in plain language (Claude auto-loads it from the description):
/senior-ux-audit # explicit
Audit the UI/UX of this app # natural language
Review my web app for accessibility (WCAG 2.2)
Why isn't this checkout flow converting?
Check the UX across all screens before launch
Claude reads your codebase, runs the app locally where it can, and returns a severity-rated report with a prioritized fix list.
- Run from the project root so the skill reads your real components, styles, and routes — it audits the codebase, not a guess.
- Start your app first (
npm run dev, a simulator, etc.). A running app lets it verify contrast, focus order, responsive reflow, and loading/empty/error states that code alone can't show. - Name the platform and depth — e.g. "deep audit of the iOS app" vs "quick scan of the landing page." Deep audit (default) covers everything; quick scan is faster and narrower.
- Point it at the flows that matter — "focus on onboarding and checkout." Findings are ranked by the business impact of the flow they sit in.
- Give context for sharper findings — who the users are (jobs-to-be-done) and what you're optimizing for (conversion, retention, accessibility).
- Ask for the output you want — a Markdown report (default) or the JSON variant for importing into a dashboard or tickets.
- Run it before launch and after building several screens to catch consistency drift across the app.
- It's read-only — it reports findings; ask it to fix them as a separate, explicit step.
ghost-skills/
├── .claude-plugin/
│ └── marketplace.json # marketplace registry
├── plugins/
│ └── senior-ux-audit/ # one plugin per skill
│ ├── .claude-plugin/plugin.json
│ ├── README.md
│ └── skills/senior-ux-audit/
│ ├── SKILL.md # the skill
│ └── references/
│ └── report-template.md
├── CONTRIBUTING.md
└── LICENSE
Each skill is one folder under plugins/ plus one entry in marketplace.json. The full recipe is in CONTRIBUTING.md.
Aditya Pratap Singh — @CreatorGhost