research-executor

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

• Third-party content exposure detected (high risk: 1.00). This skill's SKILL.md and instructions.md explicitly require using WebSearch and WebFetch / mcp__web_reader__webReader to discover and extract content from public/untrusted sources (e.g., Reddit/user reviews, vendor blogs, arXiv, public websites) as part of the mandatory multi-agent retrieval and synthesis workflow, so third‑party content is read and can materially influence agent actions and decisions.