html-to-interaction-prompts

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

• Third-party content exposure detected (medium risk: 0.65). The required workflow includes “Inspect the real source first” and “Read the HTML, CSS, and scripts” from the user-provided HTML reference, which at runtime can include outsider-authored free text (e.g., arbitrary HTML from a URL/page the operating user didn’t author) that the agent then ingests into the LLM context for analysis.