customer-persona

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

• Third-party content exposure detected (high risk: 0.90). This SKILL.md explicitly instructs using the belt CLI to run tavily/search-assistant, exa/search, and exa/answer (see "Step 1: Research" examples) which fetch untrusted public web/search results that the agent is expected to read and use to drive persona creation, allowing third‑party content to materially influence actions and enabling indirect prompt injection.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill explicitly invokes external apps via the inference.sh "belt" platform (https://inference.sh) using commands like "belt app run tavily/search-assistant" and "belt app run falai/flux-dev-lora", which at runtime execute remote apps/models and therefore fetch/execute external code that controls prompts.