The taskgrader tool manages every step of grading a contest task, from the generation of test data to the grading of a solution output.
It allows for a wide variety of contest task types and languages to be evaluated, and is meant to be used both locally for tests and in contest evaluation settings.
It uses isolate as a sandbox to run solutions to limit execution time and memory, as well as their access to the environment.
The full documentation is avaliable on GitHub pages or in the docs/ folder.