The Agent Skills Directory

[SAFE]: The skill implements presentation processing workflows using established libraries and system utilities. No malicious behaviors, prompt injections, or data exfiltration patterns were detected.
[COMMAND_EXECUTION]: Several scripts utilize subprocess.run to call system tools like LibreOffice (soffice), Poppler (pdftoppm), and git for legitimate document processing tasks.
ooxml/scripts/pack.py uses soffice to validate document integrity by attempting conversion to HTML.
scripts/thumbnail.py uses soffice and pdftoppm to generate visual slide previews.
ooxml/scripts/validation/redlining.py uses git diff for precise text comparison when validating tracked changes.
These calls are implemented using argument lists rather than shell strings and operate on files within the local environment.
[DATA_EXPOSURE]: The skill processes user-provided PowerPoint files, which are a form of untrusted data.
It mitigates risks by using the defusedxml library for XML parsing across its core scripts (unpack.py, pack.py), protecting against XML External Entity (XXE) and other XML-based denial-of-service attacks.
Structured data extraction is performed through specialized scripts like inventory.py that target specific document properties.

pptx