The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of external data from files such as survival_data.csv and survival_data.arff. This represents a potential indirect prompt injection surface if the contents of these files are untrusted and contain instructions intended to influence the agent's behavior.
Ingestion points: Identified in references/data_competing_risks.md through methods like pd.read_csv and sksurv.io.loadarff.
Boundary markers: No explicit delimiters or warnings to ignore embedded instructions within data files are present.
Capability inventory: The skill enables file system writes (via matplotlib plot saving) and package installation.
Sanitization: No specific data sanitization or validation routines are described for the content of processed files.
[EXTERNAL_DOWNLOADS]: The documentation includes instructions for installing various Python libraries.
Evidence: pip install scikit-survival scikit-learn pandas numpy matplotlib.
Context: These are widely recognized and standard packages in the Python data science ecosystem.

scikit-survival-analysis