unicorn-studio

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

• Third-party content exposure detected (high risk: 0.75). The skill's embed pattern explicitly loads code from a public CDN (jsDelivr) and can fetch an arbitrary scene JSON from an external URL via the data-us-project-src="https://.../scene.json.txt" attribute, which causes the agent/page to ingest and render untrusted third‑party content.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill instructs loading the Unicorn Studio SDK from a CDN at runtime (e.g. https://cdn.jsdelivr.net) — which fetches and executes remote JavaScript — and also supports loading a runtime scene JSON via data-us-project-src (e.g. https://.../scene.json.txt), so external content is fetched during execution and directly controls behavior.