Evaluations

An evaluation is the act of submitting evidence to a deployed boundary and receiving a verdict. Every evaluation produces either ASSERT (conditions satisfied, action authorized) or DEFER (conditions not satisfied, action withheld).

Verify Tab

Inside a project, the Verify tab lets you submit evidence and see results immediately. It has two modes:

• Single: paste or type a JSON workload, select an environment and deployment, and click Run Evaluation
• Batch: upload a file containing multiple workloads and evaluate them all at once

Running a Single Evaluation

1. Select an environment from the dropdown (the boundary must be deployed to that environment)
2. Select the deployment (which boundary version to evaluate against)
3. Enter or paste a JSON workload in the editor. The workload contains the evidence fields your boundary expects.
4. Click Run Evaluation (keyboard shortcut: Cmd+Enter)

The result panel shows:

• The verdict: ASSERT or DEFER
• The evaluation ID (unique, immutable)
• If DEFER: the failure type and which specific conditions were not met
• The complete evidence that was submitted
• The boundary version that was evaluated

Failure Types

When an evaluation returns DEFER, the failure type tells you which stage failed:

Failure Type	Stage	Meaning
admissibility_not_satisfied	Stage 1	Required evidence is missing or has the wrong type
prohibited_evidence_present	Stage 2	A prohibited field was found in the workload
predicate_failed	Stage 3	One or more predicate conditions were not met

All failing conditions are listed, not just the first one.

Batch Evaluation

Switch to the Batch tab to upload a JSON file containing an array of workloads. Each workload is evaluated independently against the same deployment. Results show the verdict for each workload with expandable details.

Sample Workloads

If your project was created from a template, the Verify tab pre-loads sample workloads that demonstrate both ASSERT and DEFER cases. You can also select from built-in sample workloads using the examples dropdown.

Evaluation Detail

Click any evaluation result to see the full breakdown. The detail view shows the complete evidence submitted, the boundary definition at the time of evaluation, each stage of the evaluation pipeline, and the exact reason for the verdict.

Every evaluation is immutable after creation. The evaluation record, the evidence, and the boundary snapshot are preserved for replay.

Monitoring Evaluations

The project Monitor tab provides three views:

• Dashboard: evaluation volume chart, ASSERT/DEFER ratio, and trends over time
• Evaluations: paginated list of all evaluations with filtering by verdict, boundary, and date range
• Alerts: notifications when evaluation patterns change (e.g., sudden increase in DEFER rate)

The monitor supports live streaming: new evaluations appear in real time without refreshing.

What’s Next

• Replay : reproduce any historical evaluation from immutable artifacts
• Boundaries : edit the rules evaluations are checked against
• Deployments : control which boundary version is active in each environment