The Agent Skills Directory

[COMMAND_EXECUTION]: The utility scripts 'ooxml/scripts/pack.py' and 'ooxml/scripts/validation/redlining.py' utilize the 'subprocess' module to execute system tools like 'soffice' and 'git'. While list-based arguments are used to mitigate shell injection, the scripts process user-provided paths which could be vulnerable to argument injection if file names are not carefully restricted.- [COMMAND_EXECUTION]: The 'ooxml/scripts/unpack.py' script uses 'zipfile.ZipFile.extractall()' without path validation, creating a 'Zip Slip' vulnerability surface. A maliciously crafted document archive with relative path components (e.g., '../../') could attempt to overwrite files outside the intended working directory.- [PROMPT_INJECTION]: The skill processes untrusted '.docx' files by extracting text and XML for agent analysis, creating an indirect prompt injection surface. Evidence Chain: (1) Ingestion points: 'ooxml/scripts/unpack.py' (XML extraction) and 'pandoc' (markdown extraction); (2) Boundary markers: Absent; extracted content is placed directly into the agent context; (3) Capability inventory: Subprocess calls to system utilities and full file system write access; (4) Sanitization: Absent for extracted text content.- [EXTERNAL_DOWNLOADS]: The documentation references instructions for installing necessary system and library dependencies such as 'pandoc', 'LibreOffice', and 'defusedxml' through well-known official package managers and registries.

docx