opposing-counsel-attack-en

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

• Third-party content exposure detected (high risk: 0.85). The required workflow explicitly takes “the submission…received FROM the other side” as user-provided material and then “reads the material in full” for analysis, so any outsider-authored draft/response text would be ingested into the LLM context at runtime (indirect prompt-injection risk via adversarial content in that submission).