The Agent Skills Directory

COMMAND_EXECUTION (SAFE): The pack.py script executes soffice (LibreOffice/OpenOffice) via subprocess.run to validate document integrity after repacking. This is implemented securely by passing arguments as a list and avoiding shell=True, which prevents command injection vulnerabilities.\n- DATA_EXFILTRATION (SAFE): The skill handles XML data using defusedxml in pack.py and unpack.py, mitigating risks from XML External Entity (XXE) and recursive entity expansion (Billion Laughs) attacks. While docx.py uses lxml, the overall design indicates a strong emphasis on secure file processing in a document-handling context.\n- REMOTE_CODE_EXECUTION (SAFE): No remote code execution patterns, external script downloads, or dynamic execution of untrusted input were detected. The skill relies on established libraries and standard system utilities for its document processing tasks.

pptx