The Agent Skills Directory

Dynamic Execution (MEDIUM): The script recalc.py dynamically generates a LibreOffice Basic macro (Module1.xba) and writes it to the local user's configuration directory (e.g., ~/.config/libreoffice/). It then executes this macro using the soffice command-line interface. Script generation combined with execution is a sensitive pattern that could be abused if the generation logic is manipulated.
Command Execution (LOW): The skill uses subprocess.run to call soffice and timeout utilities. Although arguments are passed as a list (mitigating shell injection), the script relies on the presence and integrity of these system-level binaries.
Indirect Prompt Injection (LOW): The script processes untrusted Excel files by iterating through all sheets and cells. This exposes the agent to vulnerabilities in the openpyxl library or the LibreOffice calculation engine if a malicious file is provided.
Ingestion points: recalc.py opens user-provided Excel files using load_workbook.
Boundary markers: Absent. The script processes the raw content of the workbook without sanitization.
Capability inventory: The script can execute subprocesses (soffice) and write files to the user's home directory config folders.
Sanitization: None. The script reads and interprets cell values directly to identify error strings.

xlsx