xlsx

Warn

Audited by Gen Agent Trust Hub on Jun 15, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [DYNAMIC_EXECUTION]: The file scripts/office/soffice.py contains an embedded C source string (_SHIM_SOURCE) which is written to a temporary file and compiled into a shared library at runtime using gcc. This library is then injected into the LibreOffice (soffice) process via the LD_PRELOAD environment variable to modify its socket behavior.
  • [DYNAMIC_EXECUTION]: The file scripts/recalc.py dynamically generates a LibreOffice Basic macro and writes it to the user's application configuration directory (e.g., ~/.config/libreoffice/4/user/basic/Standard/Module1.xba). This macro is subsequently invoked via a vnd.sun.star.script URI to automate formula recalculation.
  • [COMMAND_EXECUTION]: Several components execute shell commands using the subprocess module:
  • scripts/office/soffice.py runs gcc for compilation and executes the soffice binary.
  • scripts/recalc.py executes soffice, as well as timeout or gtimeout for process control.
  • scripts/office/validators/redlining.py executes git diff to compare text content within documents.
  • [PERSISTENCE_MECHANISMS]: scripts/recalc.py installs a macro into the host's LibreOffice configuration directory. This modification persists on the filesystem and affects the LibreOffice environment across different sessions.
  • [INDIRECT_PROMPT_INJECTION]: The skill possesses a significant attack surface for indirect prompt injection as it processes untrusted tabular data.
  • Ingestion points: External spreadsheet files (.xlsx, .csv, etc.) are read into the agent context using pandas and openpyxl as described in SKILL.md.
  • Boundary markers: Absent. The instructions do not specify any delimiters or warnings to ignore instructions embedded within the spreadsheet cells.
  • Capability inventory: The skill can execute arbitrary commands (via recalc.py and soffice.py), write to the filesystem, and modify application configuration files.
  • Sanitization: While the skill uses defusedxml to mitigate some XML-related attacks (like XXE), it does not sanitize or filter the actual data content from cells before it is processed by the agent.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jun 15, 2026, 09:37 AM
Security Audit — agent-trust-hub — xlsx