The Agent Skills Directory

[DYNAMIC_EXECUTION]: The create_excel_spreadsheet and modify_excel_spreadsheet tools execute Python code provided as a string via the python_code parameter. This allows the agent to dynamically generate and run logic at runtime using libraries like openpyxl, pandas, and numpy to manage spreadsheet data.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data from Excel spreadsheets while possessing powerful code execution capabilities, creating a vulnerability to indirect injection.
Ingestion points: External data enters the context through files specified in modify_excel_spreadsheet (via source_name), read_excel_spreadsheet, and preview_excel_sheets.
Boundary markers: No delimiters or instructions are provided to help the agent distinguish between spreadsheet data and potentially malicious instructions embedded within that data.
Capability inventory: The skill has the ability to execute arbitrary Python code and manipulate files using standard data science and spreadsheet libraries.
Sanitization: There is no evidence of input validation, filtering, or sanitization for the Python code generated by the agent before it is executed.

excel-spreadsheets