The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes content from external PowerPoint (.pptx) files and templates, creating a surface for indirect prompt injection where instructions embedded in a document could influence the agent's behavior.
Ingestion points: Office Open XML files (slide XMLs, notes, and comments) are unpacked and read by the agent via scripts/inventory.py and ooxml/scripts/unpack.py.
Boundary markers: Absent; the skill does not explicitly instruct the agent to use delimiters or ignore instructions found within the PowerPoint data.
Capability inventory: The skill can execute shell commands via the subprocess module (soffice, pdftoppm, git) and has file write access to the workspace.
Sanitization: The skill utilizes the defusedxml library in ooxml/scripts/unpack.py and ooxml/scripts/pack.py to securely handle XML parsing and mitigate XML-based attacks.
[COMMAND_EXECUTION]: Several utility scripts use the subprocess module to call system binaries for document conversion, validation, and comparison.
soffice (LibreOffice) is executed in ooxml/scripts/pack.py and scripts/thumbnail.py for format conversion and visual validation.
pdftoppm is used in scripts/thumbnail.py to generate slide images from PDF intermediates.
git is used in ooxml/scripts/validation/redlining.py to perform character-level differences between document versions.
[EXTERNAL_DOWNLOADS]: The skill documentation directs the user to install several packages from official registries to support its functionality.
Python packages including markitdown and defusedxml are installed via pip.
Node.js packages including pptxgenjs, playwright, and sharp are installed via npm.
System utilities including libreoffice and poppler-utils are installed via the apt package manager.

pptx