literature-report

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

• Third-party content exposure detected (high risk: 0.80). The SKILL.md explicitly instructs the agent to use Find Literature and Semantic Scholar Lookup and to run commands like "asta papers get CorpusId:" and "go through ALL highly relevant papers — extract relevance criteria, snippets, and citation contexts from each," which means the agent will fetch and read public third‑party content (Semantic Scholar / paper sources) that can influence its decisions and tool use.