The Agent Skills Directory

Dynamic Execution (LOW): The script recalc.py dynamically generates a LibreOffice Basic macro (Module1.xba) and writes it to the user's local configuration directory (~/.config/libreoffice or ~/Library/Application Support/LibreOffice). It then executes this macro via a specific URI scheme to perform Excel recalculations. While this is a functional requirement for the skill, writing and triggering execution of dynamic scripts is a sensitive operation.
Indirect Prompt Injection (LOW): The skill processes untrusted Excel files and extracts sheet names and cell coordinates to report errors. A maliciously crafted Excel file could use sheet names to inject instructions into the agent's context when it reviews the output JSON.
Ingestion points: recalc.py reads data from user-provided Excel files using the openpyxl library.
Boundary markers: The output JSON does not use delimiters or explicit warnings to separate untrusted content (like sheet names) from the tool's summary.
Capability inventory: The script has the capability to execute shell commands (soffice, timeout) and write to the local filesystem.
Sanitization: No sanitization or validation is performed on the metadata or strings extracted from the workbook before reporting them.

xlsx