docx
Pass
Audited by Gen Agent Trust Hub on Mar 5, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The instructions contain steering directives such as 'MANDATORY
- READ ENTIRE FILE' and 'NEVER set any range limits.' These are functional requirements intended to ensure the agent maintains full context for complex document manipulation and are considered safe within the primary use-case.
- [COMMAND_EXECUTION]: The skill relies on local binaries including
pandocfor text extraction,sofficefor PDF conversion, andgitfor generating tracked change diffs. These operations are executed on local files using controlled parameters. - [COMMAND_EXECUTION]: Programmable document creation is performed by the agent generating and executing JavaScript code using the
docxlibrary. This is the intended design for document generation within the skill. - [SAFE]: The skill correctly mitigates XML-based vulnerabilities, such as XML External Entity (XXE) attacks, by using the
defusedxmllibrary for all document parsing and editing tasks.
Audit Metadata