The Agent Skills Directory

[COMMAND_EXECUTION]: The recalc.py script uses subprocess.run and subprocess.Popen to execute system binaries like soffice and Xvfb. This provides a mechanism for the skill to interact with the operating system, which could be exploited if file paths or parameters are controlled by an attacker.\n- [REMOTE_CODE_EXECUTION]: The script recalc.py dynamically writes a LibreOffice Basic macro to the user's persistent application configuration folder (~/.config/libreoffice/ or ~/Library/Application Support/LibreOffice/). This macro is then automatically loaded and executed by LibreOffice, establishing a form of persistence and local code execution.\n- [PROMPT_INJECTION]: The skill is designed to ingest and process data from external spreadsheets, creating a vulnerability to indirect prompt injection.\n
Ingestion points: Files are read using pandas and openpyxl in recalc.py and through examples provided in SKILL.md.\n
Boundary markers: There are no markers or instructions implemented to prevent the agent from following commands embedded within spreadsheet cells.\n
Capability inventory: The skill possesses the ability to execute shell commands and modify the local filesystem.\n
Sanitization: No input validation or content filtering is performed on the spreadsheet data before it is processed.\n- [EXTERNAL_DOWNLOADS]: The skill requires LibreOffice to be installed on the host and uses the pandas and openpyxl Python packages.

xlsx