Evaluation stage

The Evaluation stage validates selected outputs against policy expectations and produces review evidence.

What happens in this stage

• Detection logic runs on selected output content.
• Supervisor prompt evaluations provide semantic quality review.
• Results are grouped into actionable findings and exportable evidence.

Tabs and how to use them

• Detections

- Rule/detection-based checks over selected outputs.

- Use for repeatable policy-alignment verification.

• Supervisor

- Prompt-based evaluator for deeper semantic review.

- Use to assess completeness, clarity, and implementation quality.

• Results / Evidence

- Review matched issues, rationale, and suggested fixes.

- Export for governance/audit trail as needed.

Practical workflow

1. Run Detections for objective checks.

2. Run Supervisor for qualitative review.

3. Address high-priority issues, then re-run both.

4. Export evidence when results are acceptable.