ai-llm-ops

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

• Potentially malicious external URL detected (high risk: 0.80). The skill config and runtime commands reference remote model repositories (e.g., meta-llama/Llama-3.1-70B-Instruct used in "vllm serve" and AutoModel.from_pretrained) and explicitly sets trust-remote-code: true, which means code from those remote repos will be fetched and can be executed at runtime — a required external dependency that can execute remote code.