kermt-add-cmim-pretrain

Pass

Audited by Gen Agent Trust Hub on Aug 13, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes shell commands to establish run directories and configure environment variables, such as generating unique timestamps for run folders.
  • [COMMAND_EXECUTION]: It invokes several local Python scripts (check_checkpoint.py, prepare_data.py, upgrade_to_hybrid.py, and run_pretrain_local.py) located within the vendor's repository via a containerized wrapper script (kermt_container.sh).
  • [EXTERNAL_DOWNLOADS]: The workflow supports optional telemetry and logging through Weights & Biases (wandb), which is a well-known and standard third-party service for machine learning experiment tracking.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 13, 2026, 11:11 AM
Security Audit — agent-trust-hub — kermt-add-cmim-pretrain