The Agent Skills Directory

Indirect Prompt Injection (LOW): The skill processes external .xlsx files using pandas and openpyxl.
Ingestion points: pd.read_excel() and load_workbook() ingest data from files that could contain malicious instructions.
Boundary markers: None explicitly defined in the prompt instructions to warn the agent about ignoring embedded text in cells.
Capability inventory: The skill uses pandas and openpyxl for file reading and writing. No subprocess or eval calls are present.
Sanitization: No explicit sanitization of cell content is shown before use, though the primary use case is data analysis rather than instruction following.

xlsx