The Agent Skills Directory

[DYNAMIC_EXECUTION]: The pattern template for decision boundaries (patterns/P6-decision-boundary.md) utilizes np.load() with allow_pickle=True. This setting can lead to arbitrary code execution if the input .npz file originates from an untrusted source. While common in data science for handling object arrays (like class names), it is a known security risk that should be avoided by using safer data formats or ensuring the input source is verified.
[SAFE]: The skill uses established, well-known Python libraries (matplotlib, seaborn, numpy, pandas) for its primary operations. No unauthorized network operations, exfiltration patterns, or hardcoded credentials were detected.
[SAFE]: File system access is restricted to reading local data files and writing visualization outputs to a dedicated ./figures directory. The use of PEP 723 headers for dependency management via the uv runner is a secure practice for script execution.
[SAFE]: The instructions focus entirely on data mapping and aesthetic styling. There are no attempts to bypass safety filters, override system prompts, or establish persistence on the host system.

matplotlib