osint-investigator

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

• Third-party content exposure detected (high risk: 1.00). The SKILL.md explicitly instructs the agent to fetch and browse public third‑party sites (via agent-browser or web fetch/curl) — e.g., executing /dork and /recon against sites like twitter/reddit/pastebin, archive.org and WHOIS — and to read and act on those results (passive entity recognition, pivots, automated /full investigations), so untrusted user-generated web content can directly influence tool use and next actions.