zenodo

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

• Third-party content exposure detected (high risk: 0.75). The skill’s search and download commands fetch Zenodo public record JSON at runtime (outsider-authored metadata like titles/creators/files) via urllib.request.urlopen() in make_request(), then print/serialize it (e.g., cmd_search and cmd_download), which becomes readable text in the agent’s context.