The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes external command-line utilities to perform its core functions. It uses soffice (LibreOffice) for converting and rendering presentations and standard PDF tools such as pdftoppm, mutool, or magick (ImageMagick) to generate slide images. These tools are invoked with validated paths and are necessary for the skill's stated purpose.
[EXTERNAL_DOWNLOADS]: The skill initialization process in scripts/pptx.sh uses npm ci to install its Node.js dependencies (jszip and @xmldom/xmldom) from the official NPM registry into a local cache directory. This is standard dependency management from a well-known service.
[INDIRECT_PROMPT_INJECTION]: The skill extracts text and speaker notes from PPTX slides to provide context to the agent, creating a surface where malicious presentation content could attempt to inject instructions.
Ingestion points: Text extraction occurs in scripts/lib/ooxml.mjs via the readPptxFacts and presentationSlides functions.
Boundary markers: None. The extracted text is provided directly to the agent without specific delimiters or warnings.
Capability inventory: The skill possesses capabilities for file system read/write and execution of external binaries (LibreOffice, ImageMagick).
Sanitization: The skill performs standard XML parsing and whitespace normalization but does not implement specific sanitization to filter potential prompt injection instructions from the slide content.
[SAFE]: The skill implements strong security controls, including ZIP entry validation in scripts/lib/ooxml.mjs to prevent path traversal (Zip Slip) and strict path ownership checks in scripts/lib/paths.mjs ensuring files stay within the designated work directory.

pptx