mckinsey-consultant

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

• Third-party content exposure detected (high risk: 0.90). The skill explicitly instructs Claude to perform web_searches and ingest publicly sourced URLs as part of its workflow (e.g., SKILL.md and references/methodology.md: "STEP 3: 执行5-10次快速web_search" and "STEP 6-7: 对每页执行2-5次web_search / 记录完整URL" and Dummy/Excel specs require storing source URLs), meaning it fetches and interprets untrusted third‑party web content that can materially influence subsequent analysis and tool use.