xlsx
Warn
Audited by Gen Agent Trust Hub on Feb 28, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTIONSAFE
Full Analysis
- [DYNAMIC_EXECUTION]: The script
scripts/office/soffice.pyhardcodes C source code and compiles it at runtime into a shared object library usinggcc. This library is stored in the system's temporary directory.- [COMMAND_EXECUTION]: The skill utilizes theLD_PRELOADenvironment variable to perform library injection into thesoffice(LibreOffice) process, overriding standard socket behaviors to bypass sandbox restrictions.- [COMMAND_EXECUTION]: The skill invokes several system binaries via subprocesses, includingsofficefor spreadsheet processing,gccfor runtime compilation, andgitfor document content comparison.- [PERSISTENCE]: Thescripts/recalc.pyscript writes a persistent StarBasic macro to the application's configuration directory on the host system. This macro remains present in the LibreOffice environment across sessions to facilitate therecalcfeature.- [INDIRECT_PROMPT_INJECTION]: The skill presents a surface for indirect prompt injection as it ingests and processes untrusted data from external spreadsheet files (.xlsx, .csv, .tsv) and evaluates formulas within them. Evidence: 1. Ingestion points:pandas.read_excelandopenpyxl.load_workbookinSKILL.mdandscripts/recalc.py. 2. Boundary markers: Absent. 3. Capability inventory: Subprocess calls tosofficeandgcc, file write operations across multiple scripts. 4. Sanitization: Absent.
Audit Metadata