exa-agent

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

• Third-party content exposure detected (high risk: 0.75). The required runtime workflow centers on Exa Agent runs and later calls agent_get_run_output, which can ingest Exa-generated “structured” and “grounding” text originating from web pages and other third-party/public sources, creating an indirect prompt-injection risk via outsider-authored content; the file itself does not show any user-authored-only restriction.