The Agent Skills Directory

[SAFE]: No malicious patterns, obfuscation, or data exfiltration attempts were detected across the skill files.
[SAFE]: The skill includes strong security constraints in the main instruction file, such as mandates to avoid hardcoding credentials and to implement proper access controls when training on production data.
[COMMAND_EXECUTION]: The reference files contain templates for orchestrating machine learning workflows, which include executing scripts within containerized environments (Kubeflow Pipelines) and using shell scripts for distributed training. These patterns are standard for MLOps and do not involve unsafe command injection or privilege escalation.
[SAFE]: The code templates provided for model and feature pipeline serialization use standard libraries like pickle and joblib. While these are considered insecure if loading data from untrusted sources, they are utilized here as standard implementation templates for managing internal pipeline artifacts.

ml-pipeline