skills/timblazing/skills/docx/Gen Agent Trust Hub

docx

Warn

Audited by Gen Agent Trust Hub on Jun 16, 2026

Risk Level: MEDIUMREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The script scripts/office/soffice.py dynamically compiles a C compatibility shim using gcc and injects it into the environment via the LD_PRELOAD mechanism. This is designed to bypass AF_UNIX socket restrictions in sandboxed environments but involves runtime code generation and process manipulation.
  • [REMOTE_CODE_EXECUTION]: The script scripts/accept_changes.py generates and writes a LibreOffice Basic macro file (Module1.xba) at runtime to automate the 'Accept All Tracked Changes' functionality.
  • [COMMAND_EXECUTION]: Several components use the subprocess module to execute system commands, including gcc for compilation, git diff for word-level document comparison in scripts/office/validators/redlining.py, and soffice for core document conversion tasks.
  • [EXTERNAL_DOWNLOADS]: Documentation in SKILL.md instructs the installation of the docx library via npm install -g docx as a necessary dependency for generating new documents.
  • [PROMPT_INJECTION]: The skill processes untrusted external data in the form of .docx files (via scripts/office/unpack.py). While it uses defusedxml for secure XML parsing, the ingestion of untrusted content alongside the skill's command execution capabilities constitutes a surface for indirect prompt injection.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jun 16, 2026, 05:54 PM