The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The script scripts/office/soffice.py performs runtime compilation of a C source file into a shared library using gcc. This library is then injected into the LibreOffice (soffice) process via LD_PRELOAD. This technique is used to shim AF_UNIX socket calls for platform compatibility but represents a highly sophisticated execution pattern.
[COMMAND_EXECUTION]: Several components execute shell commands using subprocess.run. This includes:
scripts/thumbnail.py: Executes soffice for PDF conversion and pdftoppm for image generation.
scripts/office/validators/redlining.py: Executes git diff to perform revision comparisons between document versions.
scripts/office/soffice.py: Executes gcc to compile the compatibility shim.
[PROMPT_INJECTION]: The skill processes untrusted input from user-provided .pptx files, creating a surface for indirect prompt injection.
Ingestion points: Presentation content is extracted via markitdown and visualized through generated thumbnail images in scripts/thumbnail.py.
Boundary markers: Instructions do not specify explicit delimiters or isolation for extracted slide content, and the agent is encouraged to blindly report all 'issues' found in the slides.
Capability inventory: The skill possesses extensive capabilities including shell command execution and runtime code generation.
Sanitization: No sanitization or filtering of presentation text is performed before it enters the agent's context.
[EXTERNAL_DOWNLOADS]: The skill requires several third-party dependencies from public registries. This includes Python packages like markitdown and Pillow, and Node.js packages such as pptxgenjs, sharp, and react-icons. It also relies on system-level binaries including soffice (LibreOffice) and pdftoppm (Poppler).

pptx