training-machine-learning-models

SUSPICIOUS: The stated purpose is benign and ML-related, but the skill is under-specified and paired with unrestricted Bash access that exceeds the documented workflow. There is no evidence of credential theft, exfiltration, or malicious payloads, but the broad execution scope and missing dependency/install constraints make it a medium-risk agent skill.