customer-persona

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

• Third-party content exposure detected (high risk: 0.90). The skill's Step 1 Research explicitly instructs running infsh apps (e.g., tavily/search-assistant, exa/search, exa/answer) to fetch open-web search results and answers from external/public websites, which the agent is expected to read and use to build personas—allowing untrusted third-party content to materially influence its outputs and actions.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill’s runtime examples call the inference.sh CLI to run remote apps (e.g., infsh app run tavily/search-assistant and falai/flux-dev-lora), so it depends on and executes remote code hosted via https://inference.sh during runtime.