Add a working directory selector: the user would select the tutorial folder, something like this and all the inputs would be taken from there. This folder will be in a version-controlled repository. For now, we can generate all the outputs in this folder as well, later we may want to separate large binary assets (captured screenshots, annotated screenshots) to prevent the repository from growing too big.
Instead "Run and annotate" and "Test translation" we would need:
"Capture screenshots" (run tutorial and save screenshots)
"Edit annotations" (open the editor)
"Generate output" (generate markdown file and screnshots with burnt-in annotations)