The Agent Skills Directory

[COMMAND_EXECUTION] (SAFE): The skill uses subprocess.run to call the quantize binary. It provides examples of secure execution using environment isolation and path validation.
[DATA_EXFILTRATION] (SAFE): The automated scanner alert for logger.info is a false positive. No actual network connections to untrusted or blacklisted domains are made.
[SAFE] (SAFE): The skill demonstrates best practices by including a ModelVerifier for SHA256 integrity checks and a validate_model_path function to prevent directory traversal attacks.
[PROMPT_INJECTION] (LOW): There is a potential surface for indirect prompt injection in the QualityAnalyzer if test_prompts are sourced from untrusted users, but this is a standard risk for model testing tools.

model-quantization