provider-architecture

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

• Potentially malicious external URL detected (high risk: 0.70). The skill hardcodes and uses the API base URL "https://api.myprovider.com/v1" at runtime (config default and get_api_base) which the provider will call to obtain LLM responses that directly determine the agent's outputs, so this external endpoint is a runtime dependency controlling prompts.