deep-research

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly spawns a "web-research" agent (Phase 2 in SKILL.md and detailed in references/agent-roles.md) with WebSearch/WebFetch privileges (and the devil's-advocate agent also uses WebSearch/WebFetch) to fetch and ingest public web content whose findings are merged in Phase 3 and can materially influence recommendations and follow-up actions, so it consumes untrusted third‑party content that could enable indirect prompt injection.