docx

Warn

Audited by Gen Agent Trust Hub on Apr 1, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill performs runtime compilation of C source code. In scripts/office/soffice.py, the script writes a socket shim (lo_socket_shim.c) to a temporary directory and executes gcc to compile it into a shared library. This library is then dynamically loaded via LD_PRELOAD to modify system call behavior for LibreOffice.
  • [COMMAND_EXECUTION]: The skill generates and executes application macros. In scripts/accept_changes.py, it writes a StarBasic macro to the filesystem and directs the soffice binary to execute it, which allows for automated document manipulation.
  • [COMMAND_EXECUTION]: The skill extensively utilizes the subprocess module to execute system binaries including soffice, pandoc, pdftoppm, git, and gcc to perform core document management and environment setup tasks.
  • [EXTERNAL_DOWNLOADS]: The skill instructions guide the user to install external dependencies such as the docx JavaScript library from the global npm registry.
  • [PROMPT_INJECTION]: The skill has a high surface area for indirect prompt injection. It ingests and processes untrusted data from .docx files without implementing boundary markers or specific instructions to the agent to ignore potentially malicious embedded commands within the document XML.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Apr 1, 2026, 05:08 PM