expert-research

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

• Third-party content exposure detected (high risk: 1.00). The skill's Step 3 explicitly instructs the agent to use web search and third‑party sources (LinkedIn, podcast listings, conference pages, niche Slack communities, Reddit, Twitter/X) to build the target list, meaning it will fetch and interpret untrusted, user‑generated public content that directly drives outreach and decision-making.