The Agent Skills Directory

[COMMAND_EXECUTION]: The recalc.py script executes system commands via subprocess.run to interact with LibreOffice (soffice) and system utilities (timeout or gtimeout). This is documented as the method for recalculating formulas in Excel files and follows best practices by using argument lists to prevent shell injection.
[PROMPT_INJECTION]: The skill provides the agent with capabilities to read and process external Excel and CSV files, which constitutes a surface for indirect prompt injection.
Ingestion points: Untrusted data enters the agent context via pandas.read_excel() and openpyxl.load_workbook() calls in SKILL.md.
Boundary markers: None are specified in the instructions to separate external data from agent instructions.
Capability inventory: The skill allows file system writes (wb.save(), df.to_excel()) and local command execution via recalc.py.
Sanitization: There is no explicit logic to sanitize or validate the content of processed spreadsheets before the agent acts on them.

xlsx