The Agent Skills Directory

[Indirect Prompt Injection] (LOW): The ooxml/scripts/unpack.py script uses zipfile.extractall() to process user-provided Office documents. This lacks path validation for ZIP entries, exposing a 'Zip Slip' vulnerability surface where malicious archives could attempt to write files outside the intended output directory.
[Unverifiable Dependencies & Remote Code Execution] (LOW): The ooxml/scripts/pack.py script executes the soffice command via subprocess.run for document validation. This introduces a dependency on external system binaries.
[Data Exposure & Exfiltration] (SAFE): No hardcoded credentials or external data exfiltration patterns were detected. The skill primarily performs local file manipulations.
[Dynamic Execution] (SAFE): The skill uses defusedxml to safely handle XML parsing in several components, though standard lxml is used in docx.py.

pptx