evolving-ai-agents

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

• Third-party content exposure detected (medium risk: 0.30). The required workflow involves loading and evaluating agent tasks and trajectory outputs from benchmark datasets (such as SWE-bench, where external GitHub issue text and code patches are ingested), presenting prompt injection exposure.