The Agent Skills Directory

[Dynamic Execution] (MEDIUM): The script recalc.py generates a LibreOffice Basic macro (Module1.xba) at runtime and writes it to the user's application configuration directory. This code is then executed using soffice to automate the formula recalculation process.
[Persistence Mechanisms] (MEDIUM): The macro is written to a persistent location (~/.config/libreoffice/...) and is not removed after execution, modifying the local LibreOffice installation environment. This is downgraded from HIGH as it is core to the skill's primary purpose.
[Indirect Prompt Injection] (LOW): The script parses and extracts values from all cells within the Excel file to report errors. If the resulting JSON summary is returned to the agent, malicious cell content could attempt to influence the agent's behavior. Evidence: 1. Ingestion point: recalc.py using openpyxl.load_workbook. 2. Boundary markers: No explicit delimiters or warnings are added to the JSON output. 3. Capability inventory: Script can write files, read files, and execute subprocesses. 4. Sanitization: No sanitization is performed on the data extracted from the spreadsheet.

xlsx