Agent.index

Work & Operations

Revise courses from student evaluation feedback

A professor had Claude Code summarize course-evaluation PDFs, filed themes as GitHub issues, and used agents to revise his security course and two others.

Done with Claude Code

What they did

Nick Feamster gave Claude Code a course-evaluation PDF, had it summarize the themes, and filed them as a GitHub issue checklist. Agents then rebuilt assignments with rubrics, standardized exam rules, and converted about two dozen slide decks to Quarto, one agent per deck, about six at a time. He approved a hand-built reference deck first. He then wrote playbook issues so fresh sessions could repeat this in two other courses.

How it went

The security course got rubrics, a syllabus communication section, consistent exam rules and rebuilt slides. The same approach was repeated in the ML systems and censorship courses. An agent also found and reconstructed a past final with empty source files.

Worth knowing

He reviewed diffs, required confirmation before any commit or push, and approved the reference deck before the fan-out. Web-sourced claims were verified rather than trusted.

Try it yourself with Claude Code

Read the course evaluation PDFs in [folder] for [course name] and summarize the recurring themes in student feedback, grouped as praise, complaints, and concrete suggestions. Then propose specific revisions to my course materials in [folder or repo], and make them on a new branch so I can review the diff. Keep any student-identifying details out of the summaries, and show me the plan before changing files.

Source

Read the original