industrial-ai-research

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

• Third-party content exposure detected (high risk: 1.00). The skill's workflow (Phase 3 "Source Collection" and Source Strategy / references/source-priority.md) explicitly requires gathering and reading public third-party content such as arXiv streams, IEEE Xplore, and publisher/conference pages and then using those documents to assemble evidence packs and drive synthesis, so untrusted web/post content can directly influence agent decisions.