The Agent Skills Directory

[REMOTE_CODE_EXECUTION] (CRITICAL): The file ooxml/scripts/unpack.py uses zipfile.ZipFile.extractall() to unpack Office documents, which is vulnerable to Zip Slip. An attacker can use path traversal sequences (e.g., ../../) in filenames to overwrite arbitrary files on the system.
[REMOTE_CODE_EXECUTION] (HIGH): The file ooxml/scripts/validation/docx.py uses lxml.etree.parse() to read XML from documents. Without explicit security settings (such as resolve_entities=False), this is vulnerable to XML External Entity (XXE) attacks, which can lead to local file disclosure or SSRF.
[COMMAND_EXECUTION] (HIGH): The script ooxml/scripts/pack.py executes the soffice binary (LibreOffice) via subprocess.run(). This creates an attack surface where a maliciously crafted document could exploit the office suite's rendering or conversion engine during the validation process.
[PROMPT_INJECTION] (HIGH): The skill ingests and processes external content from Word and PowerPoint files. This creates a significant Indirect Prompt Injection surface where instructions embedded in the documents could be interpreted by the agent, potentially leading to unauthorized actions.

pptx