The Agent Skills Directory

[COMMAND_EXECUTION]: The helper script scripts/office/soffice.py performs runtime compilation of embedded C source code using gcc. The resulting shared library is injected into the soffice (LibreOffice) process using the LD_PRELOAD environment variable to shim system calls related to Unix sockets. This is used to enable document conversion functionality in restricted sandbox environments.
[COMMAND_EXECUTION]: The skill invokes several external command-line tools via subprocess.run, including git (for revision diffing), soffice (for PDF generation), and pdftoppm (for rendering slide thumbnails).
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it parses and processes untrusted text data extracted from .pptx files.
Ingestion points: Content is extracted using scripts/inventory.py and the markitdown library and then presented to the agent.
Boundary markers: Absent. The skill instructions do not implement markers or defensive prompts to isolate extracted data.
Capability inventory: The agent has access to tools for shell command execution, file system modification, and a headless browser via Playwright.
Sanitization: No specific sanitization or filtering is applied to the extracted text content.
[EXTERNAL_DOWNLOADS]: The skill requires the installation of several packages from established registries, including markitdown, Pillow, pptxgenjs, playwright, and sharp.

pptx