The Agent Skills Directory

[COMMAND_EXECUTION] (SAFE): The pack.py script uses subprocess.run to invoke soffice (LibreOffice) for document validation via headless conversion. This is a local operation with fixed arguments, posing no risk of command injection.
[DATA_EXFILTRATION] (SAFE): No evidence of hardcoded credentials, sensitive file access, or unauthorized network communication was found.
[PROMPT_INJECTION] (SAFE): The skill templates and scripts contain only structural and functional logic for document processing, with no attempts to influence or override agent instructions.
[REMOTE_CODE_EXECUTION] (SAFE): All processing is performed locally. The skill explicitly uses the defusedxml library to mitigate XML External Entity (XXE) vulnerabilities during document parsing.

docx