tao-run-automl

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

• Potentially malicious external URL detected (high risk: 0.90). The skill calls an external LLM endpoint at runtime (https://inference-api.nvidia.com), used by the llm/hybrid/autoresearch modes and even queried (curl $AUTOML_LLM_ENDPOINT/models) so remote model responses directly influence AutoML recommendations and thus control agent decisions.