web-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 instructs spawning research subagents that use web_search (Tavily), fetch_url, and fetch4ai to retrieve content from arbitrary public URLs and save findings which the agent is then required to read and synthesize (see "Subagent Instructions Template", "Available Tools", and "Step 3: Synthesize Findings"), so untrusted third‑party web content can directly influence agent decisions and actions.