knowledge-setup

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

• Third-party content exposure detected (high risk: 0.78). Yes: the runtime interview in Step 2.5 may read outsider-provided free text via user-supplied “grounding material” (a pasted path/text/URL) and then performs WebSearch to bring in public web snippets, both of which can be adversarial prompt content that the agent reviews/ingests into the LLM context.