AI Services Market › Use cases › Work & Operations
Work & OperationsDone with Claude Mar 12, 2026
Grade student submissions from Canvas against a rubric
An instructor's pipeline pulls student files from Canvas, has Claude score each per criterion with comments, and posts grades back after human review. · Mar 12, 2026
- What They Did
- Vaibhav Bhandari, who teaches a hacking course, used a personal Canvas API token to download submissions. For each one, a script extracts text, or renders scanned PDFs as images. It sends that and a YAML rubric to Claude, which returns per-criterion scores, a first-person comment and a flag. He reviews a CSV, then confirms the upload to Canvas.
- How It Went
- On Lab 5, 13 of 22 submissions were flagged, mostly for suspected AI-generated content or thin evidence. Scoring was consistent but not calibrated to his own grading. He still read every capstone submission and heard every presentation.
- Worth Knowing
- Add examples of what a 90 looks like in the rubric's extra instructions, since scores weren't calibrated to his grading. Responses aren't cached, so re-runs call the API again.
Try it yourself with Claude
Read the student submissions in [folder path of downloaded Canvas files] and grade each against the rubric in [rubric file]. For every student, write a score and a short comment per criterion into a single spreadsheet at [output file], and flag any submission you're unsure about. Do not post anything back to Canvas; stop once the spreadsheet is complete so I can review it first.
Do this with a skill.
These AI Services Market playbooks do this exact job.