The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/recalc_bridge.py executes external processes to perform spreadsheet recalculations.
It uses subprocess.run to invoke the soffice (LibreOffice) binary for macro execution and file conversion.
It attempts to locate and execute a specific Python script (recalc.py) from a related xlsx skill directory if present in the user's home directory (~/.claude/skills/xlsx/scripts/recalc.py) to provide interoperability between skills.
[DYNAMIC_EXECUTION]: The skill employs dynamic script generation to interact with the LibreOffice spreadsheet engine.
In scripts/recalc_bridge.py, a LibreOffice Basic macro is constructed from a hardcoded template at runtime, written to a temporary .bas file, and executed to trigger a full workbook recalculation, which is a necessary step for financial modeling as standard Python libraries do not evaluate formulas.
[INDIRECT_PROMPT_INJECTION]: The skill acts as an operational layer that processes external Excel files, presenting an attack surface for indirect prompt injection.
Ingestion points: The scripts idfa_ops.py and idfa_audit.py parse .xlsx files using the openpyxl library to retrieve values, formulas, and comments.
Boundary markers: The skill does not implement explicit delimiters or instruction-ignore warnings when extracting content from the workbook to be processed by the agent.
Capability inventory: The skill has the capability to write to the filesystem, execute system commands (LibreOffice), and run local Python scripts.
Sanitization: While the audit script identifies coordinate references and hardcoded values for compliance checking, the skill does not perform general sanitization or escaping of cell data before it is presented in JSON output for agent consumption.

idfa-ops