The Agent Skills Directory

[COMMAND_EXECUTION] (HIGH): A Zip Slip vulnerability exists in ooxml/scripts/unpack.py where zipfile.extractall() is used on the input Office file without path validation. A malicious archive containing path traversal sequences (e.g., ../../) could overwrite arbitrary files on the system.
[COMMAND_EXECUTION] (HIGH): Unsafe XML parsing in ooxml/scripts/validation/docx.py using lxml.etree.parse() without explicitly disabling external entities. This makes the skill vulnerable to XML External Entity (XXE) attacks, which can be used to read local system files or conduct SSRF when processing a crafted document.
[COMMAND_EXECUTION] (MEDIUM): In ooxml/scripts/pack.py, the skill executes the soffice system binary via subprocess.run() to validate documents. This exposes the system to the large attack surface of the LibreOffice conversion engine using potentially attacker-controlled document data.
[EXTERNAL_DOWNLOADS] (MEDIUM): Several internal modules referenced in ooxml/scripts/validation/__init__.py (including pptx.py and redlining.py) are missing from the skill package, which prevents a full security verification of the document validation logic.
[EXTERNAL_DOWNLOADS] (LOW): The skill relies on external Python packages including lxml, defusedxml, python-pptx, and six. Per [TRUST-SCOPE-RULE], these dependencies are classified as LOW severity because the skill originates from a trusted source (Anthropic).

pptx