local-model-selection

Pass

Audited by Gen Agent Trust Hub on Aug 14, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: Fetches model configuration and metadata files from Hugging Face's official repositories to calculate VRAM requirements for hardware recommendations.
  • [REMOTE_CODE_EXECUTION]: Uses a Python command to parse downloaded JSON data during model size estimation; however, this is conducted via a hardcoded script processing data from a trusted source.
  • [COMMAND_EXECUTION]: Employs standard system utilities such as du, df, and rm, alongside local model management tools like lms and ollama, to manage and monitor disk usage.
  • [PRIVILEGE_ESCALATION]: Documents the process for uninstalling software utilizing administrative privileges (e.g., sudo), emphasizing user-mediated execution in a separate terminal and prohibiting the agent from handling or viewing credentials.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 14, 2026, 07:56 AM
Security Audit — agent-trust-hub — local-model-selection