The Agent Skills Directory

[DYNAMIC_EXECUTION]: The documentation in references/model_evaluation.md provides examples for using the pickle and joblib libraries to save and load models. These tools are unsafe for deserializing data from untrusted sources as they can be exploited to execute arbitrary code.
[INDIRECT_PROMPT_INJECTION]: The skill facilitates processing external data, which is a vector for indirect prompt injection.
Ingestion points: Data is ingested via pd.read_csv('data.csv') as shown in the SKILL.md workflows.
Boundary markers: The skill does not implement delimiters or provide instructions to the agent to ignore instructions embedded within the ingested data.
Capability inventory: The skill can execute Python scripts to analyze data, train models, and save results to the filesystem.
Sanitization: No evidence of data sanitization or schema validation was found in the skill's implementation.
[EXTERNAL_DOWNLOADS]: The skill recommends installing several well-known machine learning packages, such as scikit-learn, pandas, numpy, matplotlib, and seaborn, from official registries.

scikit-learn