experiment-design

Warn

Audited by Gen Agent Trust Hub on Apr 6, 2026

Risk Level: MEDIUMCOMMAND_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill includes a shell script pattern within a Python code block that uses the source command to activate a virtual environment. This command references a hardcoded absolute path (/Users/zhangmingda/clawd/.venv/bin/activate), which exposes the developer's local username and directory structure.- [COMMAND_EXECUTION]: The power analysis logic is executed via a shell heredoc (python3 << 'EOF') instead of a standard Python script. This execution pattern is non-portable and relies on shell-level operations to run code, which is a common technique for hiding or wrapping more complex commands.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Apr 6, 2026, 11:30 PM
Security Audit — agent-trust-hub — experiment-design