citations

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

• Third-party content exposure detected (medium risk: 0.65). The required workflow (Zotero/MCP → CrossRef/arXiv fallback) ingests outsider-authored free text from public web APIs/HTML—e.g., scripts/extract_metadata.py fetches CrossRef/arXiv responses and formats their returned title/author strings into BibTeX that can then be included in the agent’s LLM context; no exclusion applies.