reconnaissance

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

• Third-party content exposure detected (high risk: 0.90). The skill explicitly instructs agents to discover, fetch, clone, scrape and extract content from open public sources—e.g., "Phase 2: Source Discovery" and the Reconnaissance Methodology which list GitHub, arXiv, product pages, blogs, HackerNews/Reddit and "Selective Clone" in Phase 5—so the agent will read and interpret untrusted, user-generated third‑party content as part of its workflow, creating a clear avenue for indirect prompt injection.