The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses Python scripts to perform statistical calculations. Analysis logic is implemented using standard libraries such as pandas and numpy. There is no evidence of arbitrary command execution or dangerous subprocess spawning.
[EXTERNAL_DOWNLOADS]: The skill does not perform any network requests or download external scripts. It attempts to use the optbinning library if present in the environment for optimal binning, which is standard for feature engineering tasks.
[DATA_EXFILTRATION]: All data processing is local to the agent's environment. The skill reads from specified data paths and writes results to a local output directory. No outbound network activity or exfiltration patterns were detected.
[PROMPT_INJECTION]: The skill's instructions in SKILL.md are focused on data exploration and feature analysis. There are no attempts to override system prompts, bypass safety guidelines, or extract sensitive model information.
[SAFE]: The code structure uses a portable _vendor pattern for shared utilities and implements robust data handling practices. The behavior aligns with the stated purpose of providing a univariate analysis tool.

univariate-analysis