The Agent Skills Directory

COMMAND_EXECUTION (SAFE): The pack.py utility invokes the soffice command to validate that the repacked Office files are not corrupt. This is handled via subprocess.run with a list of arguments, which is a secure way to execute system commands without shell injection risks.\n- INDIRECT_PROMPT_INJECTION (SAFE): The skill processes external Office documents, creating a surface for indirect prompt injection. However, it mitigates common XML-based attacks by using defusedxml for parsing.\n
Ingestion points: Office documents (.docx, .pptx, .xlsx) are ingested via unpack.py, pack.py, and rearrange.py.\n
Boundary markers: Not explicitly defined in the scripts.\n
Capability inventory: The skill possesses file system read/write capabilities and the ability to execute the soffice validation command.\n
Sanitization: The skill employs defusedxml to parse and format XML content, significantly reducing the risk of XXE and related XML vulnerabilities. Note that while docx.py uses lxml.etree, the primary extraction and formatting logic in unpack.py and pack.py utilizes safe parsers.

pptx