The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes external binaries to perform its primary functions. The script scripts/render.py attempts to locate and run a headless web browser (Google Chrome, Chromium, Microsoft Edge, or Brave) or weasyprint to render HTML as PDF. It also utilizes the pypandoc library, which executes the pandoc binary to convert Markdown into HTML and DOCX formats. Additionally, scripts/build_reference_docx.py executes pandoc to extract default styling data.
[INDIRECT_PROMPT_INJECTION]: The skill acts as a processing pipeline for external Markdown content, which creates a potential surface for indirect injection attacks targeting the rendering engine (e.g., Cross-Site Scripting within the generated HTML or PDF).
Ingestion points: Markdown files and directories provided as input arguments to the scripts/render.py script.
Boundary markers: No explicit delimiters or security instructions are used to isolate user-provided content from the processing instructions.
Capability inventory: The skill possesses file system read/write access and the ability to execute subprocesses for pandoc and web browsers.
Sanitization: The skill relies on the internal parsing of pypandoc and employs standard security flags for the headless browser execution, such as --headless=new, --disable-gpu, and --disable-extensions.

render