landing-page-design

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

• Third-party content exposure detected (medium risk: 0.30). The runtime path uses the belt CLI “research” example to send the user-provided query string (outsider-authored text) to a remote search assistant (belt app run tavily/search-assistant --input ...), which ingests that free text at execution time.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.80). The SKILL includes and documents belt CLI commands that call remote inference services to execute image-generation and search apps, so the inference provider URL https://inference.sh is used at runtime to process submitted prompts and execute remote work.