brand-extract

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

• Third-party content exposure detected (medium risk: 0.65). The skill’s runtime path renders brand.html from a JSON payload (__OD_BRAND_PAYLOAD__) that is populated from the measured target website (outsider-authored page content such as logo/imagery URLs and extracted text like voice/tagline), and that payload is parsed and inserted into the LLM context via the agent’s own extraction/synthesis loop (i.e., untrusted outsider text can be fed back into the agent for further generation).