run-research

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly fetches and scrapes open web and Reddit content (see "Research tool surface" and the use of raw-web-search/smart-web-search and raw-scrape-links/smart-scrape-links in SKILL.md and references/tools.md), and it requires the agent to read and base syntheses/decisions on those untrusted, user-generated pages (e.g., Reddit threads, blogs, GitHub issues), which could allow indirect prompt injection via third‑party content.