shadcn

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

• Third-party content exposure detected (high risk: 0.75). The required workflow includes running npx shadcn@latest docs <component> and then fetching the returned example/docs URLs (often public web/GitHub raw content), which is outsider-authored free text that can be ingested into the agent’s LLM context via the fetched documentation/examples.

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 external docs/examples at runtime (e.g. URLs returned by npx shadcn@latest docs such as https://raw.githubusercontent.com/... and allows arbitrary install URLs like https://api.npoint.io/abc123 and https://ui.shadcn.com/...), and those fetched files are used as source content that directly drives what code/instructions the agent produces or installs.