orca-impact-analysis

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

• Third-party content exposure detected (high risk: 0.72). High likelihood: the required runtime workflow fetches alert/related/attack-path/compliance/breakage data via MCP tools (e.g., get_alert, get_asset_related_alerts_summary, get_alert_attack_path_data, search_cdr_events) and then synthesizes them into the LLM context; those tool responses can include free-text fields originating from outside the operating user (e.g., alert titles/stories, embedded CloudTrail/user-agent/source details, and other alert narratives), creating an indirect prompt-injection surface.