novelty-check

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

• Third-party content exposure detected (high risk: 0.90). This skill's Phase B explicitly instructs using WebSearch to query arXiv/Google Scholar/Semantic Scholar and to "WebFetch" paper abstracts and related-work sections, and Phase C then feeds those third‑party contents into the reviewer model to decide novelty, so untrusted public web content is directly ingested and can materially influence the agent's decisions.