baseline-selection-audit

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

• Third-party content exposure detected (high risk: 0.90). The Progressive Loading section of SKILL.md explicitly instructs verifying current SOTA with web search and public sources (e.g., OpenReview, arXiv, PMLR, ACL Anthology, CVF, DBLP, Semantic Scholar), so the agent will fetch and interpret untrusted third‑party content that can materially influence baseline selection and downstream actions.