The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the multiprocessing and joblib libraries to perform parallel processing of statistical metrics (such as Population Stability Index and Information Value). This is a standard performance optimization for processing large-scale credit datasets on multi-core systems.
[EXTERNAL_DOWNLOADS]: The skill references several well-known third-party libraries including pandas, numpy, toad, lightgbm, scikit-learn, and openpyxl. These are standard tools for data analysis, machine learning, and Excel report generation.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external data files provided by the user.
Ingestion points: Data enters the environment via references/func.py using pandas readers (pd.read_parquet, pd.read_csv, etc.) based on user-supplied paths.
Boundary markers: The skill does not implement specific boundary delimiters or "ignore embedded instructions" warnings for the data it processes.
Capability inventory: The skill possesses the capability to read local files, write Excel reports to the filesystem, and spawn multiple processes for calculation.
Sanitization: The skill performs technical data cleaning by replacing specific numerical placeholders (e.g., -999) with NaN, though it does not perform security-centric sanitization of the input data.

datanalysis-credit-risk