keyword-deep-dive

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly instructs the agent to "Google the target keyword" and "read the top 10 results in detail, the top 3 in full" (SKILL.md Steps 1 and 5), which requires fetching and ingesting open/public third-party web pages whose user-generated/untrusted content could materially influence the agent's decisions.