language

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

• Third-party content exposure detected (high risk: 0.90). The SKILL.md explicitly instructs using "Sales Safari" search patterns and platform-specific searches to find and extract exact quotes from public, user-generated watering holes (e.g., Reddit, news.ycombinator.com, StackOverflow, Product Hunt, G2, Twitter/X) and requires ingesting those quotes into the workflow, which clearly exposes the agent to untrusted third-party content that could carry indirect prompt injections.