hugging-face-community-evals

Pass

Audited by Gen Agent Trust Hub on Apr 26, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill includes Python scripts (inspect_eval_uv.py, inspect_vllm_uv.py, and lighteval_vllm_uv.py) that utilize subprocess.run to execute external CLI tools such as inspect and lighteval. These commands are constructed using argument lists, which is a secure method for executing shell commands.
  • [EXTERNAL_DOWNLOADS]: The Python scripts define their dependencies using PEP 723 inline script metadata. When executed via the uv tool as recommended in the instructions, these standard machine learning libraries (including transformers, vllm, torch, and accelerate) are downloaded from well-known package registries.
  • [COMMAND_EXECUTION]: The skill provides support for the --trust-remote-code flag. This is a standard feature in the Hugging Face ecosystem required for loading models with custom architectures, and it is appropriately documented within the skill's help text and instructions.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 26, 2026, 09:36 PM
Security Audit — agent-trust-hub — hugging-face-community-evals