The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes system utilities to handle presentation file formats and validation tasks. evidence: \n
scripts/thumbnail.py executes soffice for PDF conversion and pdftoppm for imaging.\n
ooxml/scripts/pack.py uses soffice to verify document integrity during repacking.\n
ooxml/scripts/validation/redlining.py calls git for diff analysis in tracked changes validation.\n- [INDIRECT_PROMPT_INJECTION]: The skill has a processing surface for untrusted data by parsing user presentations and rendering HTML slides. evidence: \n
Ingestion points: scripts/inventory.py extracts text from .pptx files, and scripts/html2pptx.js renders user-provided HTML.\n
Boundary markers: Instructions do not specify delimiters for content extracted from user presentations, which is a potential surface for indirect injection.\n
Capability inventory: The skill can execute shell commands and write to the file system.\n
Sanitization: Implements defusedxml for secure XML parsing in most scripts to prevent XXE, though some validation modules use lxml.\n- [EXTERNAL_DOWNLOADS]: The skill requires several well-known libraries and tools for its operation, all sourced from established registries or official repositories. evidence: \n
Python dependencies: markitdown, defusedxml, python-pptx, Pillow, lxml, six.\n
Node.js dependencies: pptxgenjs, playwright, sharp, react-icons, react, react-dom.\n
System tools: LibreOffice, Poppler, and git.

pptx