ai-agent-procurement-and-questionnaire

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

• Third-party content exposure detected (high risk: 0.78). This skill’s runtime workflow explicitly ingests the “buyer questionnaire” as an input artefact (outsider-authored free text supplied by the buyer), which the agent then reads in Workflow steps 1–3 (“Read the questionnaire… Read the questionnaire” / “Inspect the evidence…”), creating an indirect prompt-injection surface.