Overview
Assessments happen on traces, not review tasks. When a trace arrives, policies evaluate it and produce assessments. These assessments determine whether a workflow triggers and whether a review task is created.Assessment Flow
Assessment Generated
Each policy produces an assessment with risk level, compliance, and recommendations
What Policies Evaluate
Policies examine trace data:| Data | Description |
|---|---|
| Function Name | The action being attempted |
| Arguments | Input data for the action |
| Metadata | Additional context |
| Source | Where the trace came from |
| LLM Explanation | If provided, the AI’s reasoning |
Assessment Output
Each policy produces:| Field | Description |
|---|---|
| Compliant | Does the trace comply with this policy? |
| Risk Level | Low, Medium, High, or Critical |
| Confidence | How certain is the assessment (0-1) |
| Reason | Explanation of the assessment |
| Recommendation | Suggested action |
| Tags | Categorization labels |
How Assessments Lead to Review Tasks
- Policies create assessments on the trace
- Workflow triggers check these assessments
- If a trigger matches, the workflow executes
- If the workflow has a human intervention node, a review task is created
- The review task links to the trace with all its assessments
Viewing Assessments
Assessments can be viewed:- On the trace - Each trace shows its policy assessments
- On the review task - If created, shows the linked trace’s assessments
- In the session - All traces and their assessments in context