The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The scripts ooxml/scripts/unpack.py, ooxml/scripts/validation/base.py, and ooxml/scripts/validation/redlining.py use the zipfile.ZipFile.extractall() function to unpack presentation archives. This method is vulnerable to Zip Slip path traversal attacks, allowing a malicious PPTX file to overwrite arbitrary files on the host system if they contain filenames with directory traversal sequences.
[REMOTE_CODE_EXECUTION]: Unsafe XML parsing is present in several files including ooxml/scripts/validation/base.py and ooxml/scripts/validation/docx.py via lxml.etree.parse() and xml.etree.ElementTree.parse() without explicitly disabling entity resolution. This exposes the skill to XML External Entity (XXE) attacks when processing untrusted PowerPoint XML components.
[PROMPT_INJECTION]: The skill extracts slide text using scripts/inventory.py and markitdown for the agent to analyze. It does not employ boundary markers or specific instructions for the agent to disregard instructions within that text, creating an indirect prompt injection surface where a presentation's content could hijack the agent's behavior.
[COMMAND_EXECUTION]: Scripts such as scripts/thumbnail.py and ooxml/scripts/pack.py execute external binaries including soffice (LibreOffice) and pdftoppm via subprocess.run(). While used with argument lists, the execution of complex document renderers on potentially hostile files carries inherent exploitation risks.
[EXTERNAL_DOWNLOADS]: The skill requires and provides instructions for installing multiple external libraries and system utilities from public registries and repositories, including markitdown, playwright, sharp, and LibreOffice.

pptx