code-exec-skills
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill contains extensive documentation of shell commands intended for setting up research environments and automating pipelines. Examples include Python environment management (
python -m venv,pip install,uv), system package installation (apt-get), and data operations (wget,unzip). These are standard educational instructions for the target research use-case. - [EXTERNAL_DOWNLOADS]: The guides instruct users and agents on how to fetch datasets, libraries, and tools from established services. References include official Kaggle API interactions, Hugging Face dataset loading, and fetching Nix packages from GitHub. These operations target well-known and reputable domains.
- [REMOTE_CODE_EXECUTION]: Several guides, particularly the
sandbox-execution-guide, describe methods for executing scripts programmatically. This includes the use ofsubprocess.runto execute Python code in isolated processes and the use of Docker to run research containers with resource limits. These patterns are presented as best practices for secure and reproducible research computing.
Audit Metadata