The Agent Skills Directory

[SAFE]: The skill provides deterministic local auditing tools designed to validate configurations and datasets without executing untrusted text or making unauthorized network requests.\n- [SAFE]: Data validation scripts (e.g., audit_dataset.py) implement robust safety checks including SHA-256 checksum verification, schema validation, and duplicate detection to prevent data leakage and ensure integrity.\n- [SAFE]: Configuration parsing uses a restricted YAML loader derived from yaml.SafeLoader that explicitly forbids aliases, anchors, and custom tags, mitigating common YAML-based injection vectors.\n- [SAFE]: The skill employs bounded I/O operations by enforcing root directory constraints and rejecting symbolic links or path traversal attempts in all provided scripts.\n- [SAFE]: Sensitive information is protected through checks that prevent credentials from being included in logs or reports, relying on environment variable presence verification rather than reading secret values.

hypogenic