The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external data which presents a theoretical indirect prompt injection surface.
Ingestion points: The script scripts/credit_analytics_validation.py reads data from a user-supplied CSV file via the input_csv argument.
Boundary markers: The instructions do not define specific delimiters for the processed data to prevent the agent from interpreting data contents as commands, though the script's output is limited to numeric aggregates.
Capability inventory: The skill uses pandas for file reading and standard Python file operations for writing results to a local JSON file.
Sanitization: The Python script implements type validation and error coercion using pd.to_numeric and pd.to_datetime, which ensures numeric and temporal consistency and limits the ability of malformed data to affect execution.

credit-analytics