The Agent Skills Directory

[COMMAND_EXECUTION]: The create_excel_spreadsheet and modify_excel_spreadsheet tools accept a python_code parameter containing raw Python scripts. This allows the agent to execute code directly, which can lead to arbitrary command execution on the host system if isolation is insufficient.
[REMOTE_CODE_EXECUTION]: The skill utilizes a dynamic code execution model where scripts are generated and run at runtime. Although the documentation specifies the use of libraries such as openpyxl and pandas, there is no evidence of technical restrictions to prevent the use of sensitive modules like os or subprocess to perform unauthorized actions.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface because it processes data from external spreadsheet files.
Ingestion points: The read_excel_spreadsheet and modify_excel_spreadsheet tools read data from potentially untrusted .xlsx files into the agent's context.
Boundary markers: There are no markers or instructions provided to delimit spreadsheet content or to warn the agent against following instructions embedded within the data.
Capability inventory: The skill has the capability to list files in the workspace, read/write files, and execute arbitrary Python code.
Sanitization: The skill does not perform any validation or sanitization on the data retrieved from external spreadsheets before it is interpreted or used in subsequent operations.

excel-spreadsheets