shadcn

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). The skill’s runtime workflow can ingest outsider-authored free text from community registries/GitHub via npx shadcn@latest search/view/docs URL fetching and then reads the resulting registry item file contents (“Review added components — always read the added files”).

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly instructs fetching component docs, preset URLs, and raw registry/item files at runtime (e.g. https://ui.shadcn.com and raw.githubusercontent.com paths, and examples like https://api.npoint.io/abc123), which are then used to drive the agent's context/installation behavior and can inject remote code or prompts.