saos-orzecznictwo

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

• Third-party content exposure detected (high risk: 0.78). At runtime the skill calls SAOS’s public API and ingests outsider-authored court text fields like textContent/summary from the JSON responses, which are then printed/returned to the agent and can be included in the LLM context (indirect prompt injection via untrusted textContent).