The Agent Skills Directory

[REMOTE_CODE_EXECUTION] (HIGH): The skill is vulnerable to 'Zip Slip' (directory traversal during extraction) in ooxml/scripts/unpack.py and ooxml/scripts/validation/docx.py. The use of zipfile.ZipFile.extractall() on untrusted input files without validating member paths allows an attacker to create a crafted Office document that overwrites arbitrary files on the system when unpacked.
[COMMAND_EXECUTION] (MEDIUM): In ooxml/scripts/pack.py, the script executes the soffice (LibreOffice) binary using subprocess.run. While it uses an argument list, the inclusion of a user-controllable file path (doc_path) without strict validation creates a surface for argument injection if the filename is maliciously crafted (e.g., starting with a dash).
[DATA_EXFILTRATION] (MEDIUM): The DOCXSchemaValidator in ooxml/scripts/validation/docx.py uses lxml.etree.parse() to process XML components. Unlike the defusedxml used elsewhere in the skill, lxml can be susceptible to XML External Entity (XXE) attacks depending on the environment's default settings, potentially allowing an attacker to read local files via a malicious .docx file.
[SAFE] (INFO): The skill correctly uses defusedxml for its primary XML formatting tasks in unpack.py and pack.py, mitigating standard XXE risks in those specific operations.

pptx