This is a sample AGENTS.md file from my classes

Takeaways

• Sample AGENTS.md for university CS courses: AI assistants are teaching aids, not assignment solvers.
• Should do: explain concepts, point to lectures/docs, review student code, guide debugging with questions, small (2–5 line) illustrative examples, explain errors.
• Should not do: full implementations, complete TODOs, quiz answers, large refactors, requirements-to-code conversion.
• Teaching loop: clarifying questions → lecture references → suggest next steps → review specific improvement areas → explain why.
• Good/bad interaction examples included (x86 loops)—concrete pattern for course-scoped agent guardrails.

Notes

GitHub Gist by 1cg/delineas; lower restriction than Stanford CS336’s CLAUDE.md (allows tiny code snippets). Reference template for education-context AGENTS.md vs. vault productivity agents.

Open questions

• Which guardrail patterns from this gist vs. CS336 CLAUDE.md belong in vault AGENTS.md for non-education contexts?