reference-audit-guide

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

• Third-party content exposure detected (high risk: 0.80). The skill's required workflow explicitly uses WebSearch and Google Scholar (references/README.md "主要工作流" and SKILL.md) and the included scripts (scripts/verify-citations.py and scripts/api-clients.py) call public APIs/URLs (CrossRef, arXiv, Semantic Scholar, DOI resolver), so the agent fetches and interprets open third‑party web content that can materially influence verification actions.