create-a-skill

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

• Third-party content exposure detected (high risk: 0.80). The SKILL.md's Phase 1 "Research the domain" explicitly instructs the agent to "Use web search" and "research in parallel via subagents" (i.e., fetch/open public web docs and examples), and the eval viewer templates also load external CDN resources (Google Fonts and SheetJS), so the skill will consume untrusted public web content that could influence tool decisions.