agent-abuse-pentest

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

• Third-party content exposure detected (high risk: 0.75). Outsider-authored free text can be ingested when the agent performs “Indirect (retrieved doc)” / “Malicious README/issue/PR” / “Malicious tool result” tests by fetching or reading repo/ticket/comment content or tool outputs at runtime, which the workflow treats as attacker-controlled prose and feeds into the LLM context.