skills/jakerains/agentskills/sam3/Gen Agent Trust Hub

sam3

Pass

Audited by Gen Agent Trust Hub on Sep 21, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill facilitates the installation of the SAM 3 repository from the official Meta (facebookresearch) GitHub repository. It also installs official machine learning libraries from the PyTorch registry and models via Hugging Face. These are well-known and trusted sources.
  • [COMMAND_EXECUTION]: The utility script scripts/create_inference_starter.py generates Python starter scripts and saves them to the local filesystem. It uses chmod 0o755 to make these generated scripts executable, which is an expected and functional behavior for a code-generation tool.
  • [DYNAMIC_EXECUTION]: The preflight check script (scripts/sam3_preflight_check.py) uses importlib.import_module() to dynamically verify that required dependencies like torch and sam3 are correctly installed in the user's environment. The module names are hardcoded within the script and do not accept untrusted external input.
  • [DATA_EXPOSURE]: The preflight check script inspects the environment for HF_TOKEN or HUGGINGFACE_HUB_TOKEN to confirm the user is authenticated with Hugging Face. This check is purely local for reporting purposes and does not exfiltrate or display the secret values.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 21, 2026, 08:30 PM