The Agent Skills Directory

[SAFE]: The skill consistently employs the defusedxml library for parsing and manipulating XML content across its Python scripts. This measure effectively mitigates risks associated with XML External Entity (XXE) attacks and other XML-based vulnerabilities.
[COMMAND_EXECUTION]: The skill utilizes the subprocess module to execute external utilities including pandoc, soffice (LibreOffice), pdftoppm, and git. These tools are used for legitimate document processing tasks such as structure extraction, PDF/image conversion, and change validation. Execution is performed on local temporary files, minimizing risk.
[PROMPT_INJECTION]: Instructions in the markdown documentation include directives such as 'MANDATORY
READ ENTIRE FILE' and 'NEVER set any range limits'. These are technical instructions designed to ensure the AI agent has full context for complex document structure rules, rather than attempts to bypass safety filters or ignore prior instructions.
[SAFE]: As a tool that ingests and processes user-provided Word documents, the skill has an inherent surface for indirect prompt injection via document content. However, the skill implements structural validation and metadata sanitization, and its logic is focused on container manipulation rather than autonomous execution of document-based commands.

docx