skill-comply

This module does not contain clear self-contained malware, obfuscation, or direct exfiltration logic. However, it executes scenario-provided setup commands via subprocess using high-capability tools (git/npm/pip/tar/unzip) and invokes an external 'claude' CLI with 'Bash' enabled. If scenarios are untrusted or sandbox isolation is weak, this becomes a supply-chain and execution-risk vector (e.g., package-install script execution, network access from git/pip/npm, or arbitrary command execution via allowed tools). Home-path redaction is limited and may allow sensitive data to appear in serialized observations.