Rubric

Evaluation Rubric

1. Technical Completeness (12 points)

Score	Criteria
10–12 pts	Agent pipeline runs stably, all requirements met, full error handling
7–9 pts	Core functionality works, some edge cases unhandled
4–6 pts	Basic functionality only, multiple bugs present
1–3 pts	Partial implementation, demo fails
0 pts	No submission or completely non-functional

2. Application of Ralph Loop Philosophy (6 points — included in Technical Completeness)

Evaluation criteria:

• HOTL governance layer implemented
• Harness script (with backpressure)
• Context management (Context Rot prevention)
• Evidence of instruction tuning applied
• State tracking file utilized

3. Problem Fit (6 points)

Score	Criteria
5–6 pts	Genuinely useful problem; solution is effective and creative
3–4 pts	Appropriate problem; standard solution
1–2 pts	Problem is too trivial or unrealistic

4. Presentation Quality (6 points)

Item	Points
Clarity of problem explanation	2 pts
Architecture explanation (with diagrams)	2 pts
Live demo or demo video	2 pts

5. Peer Evaluation (6 points)

Individual project format (2026): With no teammates, peer evaluation is aggregated from mutual evaluation of other students’ presentations and deliverables.

• Each student evaluates all other students’ presentations (5 criteria, normalized to 100 points)
• The average becomes your peer evaluation score (out of 6)
• Evaluation form distributed during Week 16 class

---

Capstone Total Score

Item	Points
Technical Completeness	12 pts
Problem Fit	6 pts
Presentation Quality	6 pts
Peer Evaluation	6 pts
Total	30 pts

30 points = 30% of the final grade