I wrote an Agent-assisted homework grading system.

After entering the grading standards and the assignments,
the Agent quickly lists a score table.
Each evaluation criterion’s weight is well-ordered,
including the parts where students may be cheating—these are all marked and require manual review.
It’s also encouraging that students use Ai tools to help with learning, but it looks like their tool-usage skills still haven’t surpassed mine.
Next, it will take a few days to review them one by one.