| name | assess |
| description | Assess the quality of KANDy model results by examining generated plots. Use after training a model, running rollouts, or generating comparison plots. |
| argument-hint | ["system-name"] |
Assess Model Quality
Assess the quality of results for $ARGUMENTS.
Current Results
!ls results/ 2>/dev/null || echo "No results yet"
Instructions
Use the model-quality-assessor agent to:
- Find all plots in
results/$ARGUMENTS/ (or search results/ if the exact directory name differs)
- Read and visually analyze each plot (PNG files)
- Apply system-specific assessment criteria:
- ODEs (Lorenz, Hopf): trajectory tracking, attractor geometry, divergence time
- PDEs (Burgers, KS): space-time rollout accuracy, shock locations, spectral content
- Maps (Henon, Ikeda): attractor structure, iterate accuracy
- Check for common failure modes (blowup, phase drift, amplitude mismatch, overfitting)
- Rate overall quality: Excellent / Good / Fair / Poor
- Provide specific recommendations for improvement if needed
Output Format
- Plots reviewed (list each file)
- Per-plot analysis
- System-specific checks (pass/concern/fail)
- Overall rating with justification
- Actionable recommendations
