Skip to content

Evaluation

Evaluation checks whether an agent system behaves correctly, safely, and observably across expected and adverse scenarios.

For agentic systems, evaluation must cover protocol behavior, authorization failures, tool failures, stale discovery metadata, prompt injection, and human review controls.

Evaluation Dimensions

DimensionExample check
Task successHappy path produces the expected artifact
AuthorizationMissing scopes block side effects
Tool reliabilityTool timeout returns a safe partial result
Retrieval safetyPrompt injection in sources does not expose secrets
Discovery governanceStale Agent Card metadata is rejected
ObservabilityTrace and audit events explain the workflow

Demo Mapping

The demo scenarios are defined in evaluation-scenarios.yaml and exercised by API tests.

The local trace endpoint turns runtime events into architecture evidence:

text
GET /debug/traces/{traceId}

Page created by Dr. C. Klukas