customer-persona

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 running web-search apps (e.g., belt app run tavily/search-assistant and belt app run exa/search in "Step 1: Research") which fetch and ingest open/public, user-generated web content that the agent is expected to read and use to shape persona decisions, enabling indirect prompt injection.

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

• Potentially malicious external URL detected (high risk: 0.70). The skill explicitly requires and calls the inference.sh CLI (https://inference.sh) via belt app run (e.g., tavily/search-assistant, falai/flux-dev-lora), which runs remote apps/models at runtime and thus executes external code that the skill relies on for its outputs.