axiom-ai

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

• Third-party content exposure detected (high risk: 0.90). The skill explicitly documents and requires Tool protocol calls that fetch external public data (e.g., the GetFactTool example in skills/foundation-models-diag.md that mentions Wikipedia/news APIs and the GetWeatherTool example in skills/foundation-models-ref.md), and those tool outputs are inserted into the session transcript and used by the model to drive subsequent generation, so untrusted third‑party content can materially influence agent behavior.