thatopen-impl-federation

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

• Third-party content exposure detected (high risk: 0.80). The skill explicitly ingests external IFC models via IfcLoader.load (and initializes a worker from a public URL via fragments.init(workerURL) — e.g., the example unpkg URL in references/examples.md), and then uses classifier.find results (model names, storeys, categories) to drive Hider/BoundingBoxer actions (isolate/set/toggle, camera fitting), so untrusted third-party model content can materially influence tool use and next actions.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill calls fragments.init("https://unpkg.com/@thatopen/fragments@3.3.6/dist/Worker/worker.mjs") at runtime which fetches and executes a remote worker JS module (remote code executed as a required runtime dependency), so the URL can directly control runtime behavior.