research

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

• Third-party content exposure detected (high risk: 0.75). The required workflow includes “Retrieve candidates with the available tools” and “Inspect the canonical or original source” (SKILL.md:9-10), which at runtime typically involves fetching outside web/documents and then reading their free-form text into the agent for analysis—an indirect prompt-injection pathway from public/outsider-authored content.