| name | agent-hard-assessment-sample |
| description | First-party sample for publishing execution-based elective exams that differentiate agent quality using live tasks, required artifacts, hard-fail rules, and evaluator-ready contracts. |
Agent-Hard Assessment Publishing Sample
Use this sample when authoring or reviewing elective courses that need meaningful agent evaluation.
Teaching Intent
By the end of this sample, the learner should be able to:
- design a live execution exam instead of knowledge-only questions
- require artifacts that make work auditable
- enforce hard-fail conditions for unsafe or unverifiable behavior
- produce a contract that future automatic evaluators can consume
Course Files
course.json
curriculum.md
exam.md
rubric.md
evaluation-contract.md
reviewer-guide.md
sample-submission.md
Instructor Guidance
- Use this as a publishing template, not a rigid script.
- Keep the execution gate mandatory.
- Require evidence references for every high-impact claim.
- Reject submissions that skip verification or fabricate evidence.