docx
Warn
Audited by Gen Agent Trust Hub on Sep 10, 2026
Risk Level: MEDIUMDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [DYNAMIC_EXECUTION]: The script
scripts/office/soffice.pycontains an embedded C source code string that is written to a temporary file and compiled into a shared library usinggccat runtime. It then executes the LibreOffice binary (soffice) with theLD_PRELOADenvironment variable pointing to this library. This mechanism is used to intercept and redirectsocketsystem calls to bypass environment-specific socket restrictions. - [DYNAMIC_EXECUTION]:
scripts/accept_changes.pydynamically generates a LibreOffice Basic macro file (Module1.xba) in a temporary directory and executes it via the command line to automate document cleaning and the acceptance of tracked changes. - [INDIRECT_PROMPT_INJECTION]: The skill's primary function involves unpacking
.docxfiles and extracting their internal XML content. This content enters the agent's context to guide modifications or analysis, creating an ingestion surface where instructions hidden within untrusted documents could potentially override agent behavior. The skill lacks explicit boundary markers to segregate this external content from instructions. - [COMMAND_EXECUTION]: The skill invokes several system utilities via
subprocess.run, includingsoffice(LibreOffice) for file conversion,gccfor the runtime compilation of helper shims, andgitfor performing word-level diffs during document validation.
Audit Metadata