vast-instance

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). Insecure: the prompt explicitly tells the agent to pass the live Hugging Face token inline (HF_TOKEN=... python train.py), which requires embedding the secret verbatim in commands/outputs and thus exposes it.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill includes a runtime pip install that uses the external index URL https://download.pytorch.org/whl/cu121 to fetch and install the torch wheel (remote code executed at install and required for the pinned stack), so this URL is a runtime dependency that executes remote code.