run-research

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

• Third-party content exposure detected (high risk: 1.00). The SKILL.md explicitly requires discovery and ingestion of public web pages and Reddit threads (e.g., "After search_google, scrape the best 3–5 URLs immediately using scrape_pages" and "After search_reddit, fetch the best 5–10 threads using fetch_reddit with fetch_comments=true"), so the agent will read untrusted, user-generated third‑party content that can influence its actions.