risk-assessment-frameworks

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

• Third-party content exposure detected (high risk: 0.80). The skill's required code and workflows explicitly read and act on external/public content—e.g., queries against vdem_data and tables like document_content, media_article, and online_extremist_content (and integration of V‑Dem/Transparency International indices) are used at runtime to compute risk scores and trigger warnings, meaning untrusted public/user‑generated content can materially influence agent decisions.