The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection as it facilitates the processing of untrusted external data.
Ingestion points: Instructions in SKILL.md and references/quick_reference.md suggest using pd.read_csv('data.csv') to load datasets.
Boundary markers: None. There are no instructions to use delimiters or warnings to prevent the agent from interpreting instructions embedded within the processed data.
Capability inventory: The skill has access to the Bash, Write, and Edit tools, which provides a high-capability environment that could be exploited if an injection occurs.
Sanitization: No data validation or sanitization steps are included in the suggested workflows.
[EXTERNAL_DOWNLOADS]: The skill recommends installing several well-known and trusted Python packages from PyPI, such as scikit-learn, numpy, pandas, matplotlib, seaborn, umap-learn, category-encoders, and imbalanced-learn. These are standard tools for machine learning development and originate from recognized sources.
[COMMAND_EXECUTION]: The skill uses the Bash tool to perform routine tasks such as installing packages via uv and executing example scripts (e.g., scripts/classification_pipeline.py). This is the primary intended function of the skill and is documented clearly.
[SAFE]: The skill provides extensive, accurate, and professional-grade documentation for machine learning tasks. It includes comprehensive reference guides for supervised learning, clustering, model evaluation, and preprocessing. While the documentation in references/model_evaluation.md mentions model persistence using pickle and joblib (which involve unsafe deserialization of untrusted data), they are presented neutrally as standard tools for local data science workflows.

scikit-learn