The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's instructions mandate a 'Verification Loop' that requires the agent to execute a local Python script scripts/recalc.py. This script uses subprocess.run to call the soffice (LibreOffice) binary and system utilities like timeout or gtimeout to handle spreadsheet recalculation tasks.
[PERSISTENCE]: The verification script scripts/recalc.py automatically installs a LibreOffice macro (Module1.xba) into the user's local configuration directory (~/.config/libreoffice/ or ~/Library/Application Support/LibreOffice/). This macro, which contains code to calculate and store the workbook, persists across application sessions to enable headless operation.
[INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted spreadsheet data. Because the agent is instructed to 'fix all errors identified' in the output of the recalculation script, malicious data within a spreadsheet (e.g., in cell values or formulas) could be used to manipulate the agent's logic or downstream actions.
Ingestion points: User-provided data for modeling and the generated output.xlsx files analyzed by the recalculation script.
Boundary markers: None provided; the skill lacks explicit delimiters or instructions to ignore data-embedded commands.
Capability inventory: The skill can execute shell commands via subprocess.run and write files to the local filesystem for macro installation.
Sanitization: The skill does not implement sanitization or validation of the spreadsheet content before the agent attempts to fix errors reported by the recalculation tool.

excel