docx
Pass
Audited by Gen Agent Trust Hub on Mar 24, 2026
Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses
subprocess.runto execute various system utilities, includingsoffice(LibreOffice) for document conversion,pandocfor text extraction, andpdftoppmfor image generation. - [COMMAND_EXECUTION]: In
scripts/office/soffice.py, the skill embeds C source code that is compiled at runtime usinggccto create a shared object (lo_socket_shim.so). This library is utilized withLD_PRELOADto emulate AF_UNIX socket behavior, which is necessary for LibreOffice to operate in certain sandboxed or restricted environments. - [SAFE]: The skill implements a significant attack surface by processing untrusted document data, but provides appropriate mitigations:
- Ingestion points: External .docx files are processed through
scripts/office/unpack.pyand theDocumentclass inscripts/document.py. - Boundary markers: Data is handled within structured XML formats, maintaining a clear separation between document content and agent instructions.
- Capability inventory: The skill possesses capabilities to execute shell commands and compile code, which are used solely for document processing tasks.
- Sanitization: All XML parsing is performed using the
defusedxmllibrary to prevent XML External Entity (XXE) vulnerabilities.
Audit Metadata