citation-network

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

• Third-party content exposure detected (high risk: 0.80). The skill's init.py directly calls semantic_scholar_get_paper and semantic_scholar_citations to fetch paper metadata and citation lists from public services (Semantic Scholar/arXiv), meaning the agent ingests and summarizes open third-party content that could contain user-generated text capable of influencing its actions.