augment

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

• Third-party content exposure detected (high risk: 0.85). The required workflow spawns a “dimension-analyst” agent that performs WebSearch/WebFetch and then reads retrieved web page text as part of its findings, which is outsider-authored free text injected into the LLM context via the analyst’s runtime research loop.