The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run within scripts/office.py to perform document conversion via LibreOffice (soffice). The implementation follows security best practices by using list-based argument passing (avoiding shell command injection), resolving the binary path through verified locations, and utilizing a temporary directory for the execution profile (-env:UserInstallation) to ensure process isolation.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external documents (PDFs and Office files) which represents a vulnerability surface for indirect prompt injection. Ingestion points: Files are read via scripts/pdf.py and scripts/office.py for manipulation and rendering. Boundary markers: The skill documentation (SKILL.md) explicitly instructs the agent to render documents to PNG images for visual inspection and warns against auto-opening links or executing embedded content. Capability inventory: The skill can write files to the local system and execute the LibreOffice binary for conversion. Sanitization: The code uses defusedxml for XML processing and leverages standard library parsing logic in pypdf to handle complex document structures.
[EXTERNAL_DOWNLOADS]: The skill relies on uv run to dynamically manage standard Python dependencies including pypdf, reportlab, pypdfium2, defusedxml, and pillow. These are well-established libraries in the document processing ecosystem and are used here for their intended purpose.

pdf