ai-automation-workflows

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

• Third-party content exposure detected (high risk: 0.90). The skill's Sequential Pipeline explicitly runs a search assistant (RESEARCH=$(belt app run tavily/search-assistant --input "...")) to fetch web search results and then feeds that untrusted third-party content directly into a downstream model prompt to generate an article, which lets external/web content materially influence agent behavior.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill calls external apps via the inference.sh platform (https://inference.sh) at runtime (e.g., using belt app run to get RESEARCH/ARTICLE) and then injects those results into subsequent prompts, so remote content from that URL can directly control agent instructions.