The Agent Skills Directory

[SAFE]: The skill implements a deterministic diagnostic workflow for financial data. Analysis of the scripts and instructions reveals no evidence of prompt injection, data exfiltration, or unauthorized command execution.
[INDIRECT_PROMPT_INJECTION]: The skill contains a data ingestion surface for processing external CSV files.
Ingestion points: The scripts/products_and_markets_diagnostics.py script reads user-provided data from input.csv using pandas.read_csv.
Boundary markers: Not explicitly defined in the script's logic.
Capability inventory: The script's operations are restricted to statistical calculations (mean, standard deviation, quantiles) and writing the resulting summary to a local JSON file.
Sanitization: The script performs type validation using pd.to_numeric and pd.to_datetime with errors='coerce', which effectively filters out non-numeric or malformed data that could be used for injection before processing.

products-and-markets