xlsx
Warn
Audited by Gen Agent Trust Hub on Mar 4, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The
recalc.pyscript executes the system binarysoffice(LibreOffice) to perform headless formula recalculations. - [COMMAND_EXECUTION]: To support its recalculation feature, the script dynamically generates a StarBasic macro and writes it to the user's local LibreOffice configuration directory (e.g.,
~/.config/libreoffice/4/user/basic/Standard/Module1.xba). This creates a persistent executable script on the filesystem. - [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection as it processes external spreadsheet data with high-privilege capabilities.
- Ingestion points: Files are loaded using
openpyxlandpandasas described insub-skills/data-analysis-with-pandas.mdand used byrecalc.py. - Boundary markers: No boundary markers or instructions are provided to the agent to disregard instructions embedded in spreadsheet cells or formulas.
- Capability inventory: The skill possesses the ability to execute system commands, write to configuration directories, and modify local files.
- Sanitization: No evidence of data validation or sanitization of spreadsheet contents was found before processing.
- [PROMPT_INJECTION]: The
LICENSE.txtfile claims copyright by Anthropic, PBC, which contradicts the author provided in the skill context (Dokhacgiakhoa), potentially misleading users about the origin and safety of the skill.
Audit Metadata