researcher-role-skill

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

• Third-party content exposure detected (high risk: 0.90). The skill explicitly requires "Internet access for external research" and instructs the agent to consume community resources and public content (e.g., GitHub repos, Stack Overflow, Reddit, HackerNews, public websites and URLs) as part of its research workflow, which are untrusted user-generated third‑party sources the agent will read and interpret.