shadcn

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.70). These URLs include third-party/custom registry endpoints and an arbitrary JSON hosting URL (api.npoint.io, acme.com, private.com) that can deliver source/registry metadata outside the official shadcn registry and could be used to inject malicious component code; the official ui.shadcn.com and raw.githubusercontent.com links are normal documentation/CDN URLs and not flagged.

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

• Third-party content exposure detected (high risk: 0.75). The required workflow includes fetching component documentation/examples and viewing registry items (e.g., npx shadcn@latest docs <component> then fetching the URLs, and view/add --view), which can ingest outsider-authored free text from public web sources or community registries into the agent’s LLM context.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill explicitly instructs the CLI/agent to fetch external docs, presets, and registry JSON at runtime (e.g. https://ui.shadcn.com, https://raw.githubusercontent.com/.../examples/..., https://api.npoint.io/abc123, https://ui.shadcn.com/r/registries.json, https://acme.com/r/{name}.json, https://private.com/r/{name}.json), which means remote content is fetched during runtime and can directly control the agent's instructions or install/overwrite code.