They may not expose the actual PDF to you at all, just some software rendering of it. In that case I’d focus on making screenshotting efficient. Find a program that lets you save the whole screen to file automatically at once (one button press), or use Firefox ctrl+shift+s -> click on the page area -> save -> enter.
Feedback: to see an example one has to click through to another file in the repo.
Is it a subset of Markdown or YAML? It is a type of decision that it would be good to be upfront with to the users. It also gives you a framework for further thinking and development, and some out of the box parsability.