Synthetic persona
A labeled artificial participant used as a repeatable test fixture rather than represented as a real person.
Source: TrustSafetyComplianceGlossary
Scenario pack
A reusable bundle of a use case, persona constraints, expected conditions, and review criteria.
Source: ProductFeatureSetSimulation trace
The recorded sequence of interactions produced during a synthetic test run.
Source: ArtifactCaptureStrategyEvaluation rubric
A set of stated criteria used to review behavior and outcomes consistently.
Source: ProductFeatureSetRisk flag
A recorded signal that a reviewer should inspect possible bias, manipulation, policy, or safety concerns.
Source: CoreAgentOrAutomationReplay log
A saved record that lets a reviewer revisit how a simulated exchange unfolded.
Source: ProductFeatureSetEvaluation set
A saved scenario, evidence, and criteria that can be rerun against later system versions.
Source: AgentWorkflowPersona catalog
The planned library from which labeled synthetic test participants can be selected.
Source: ProductFeatureSetHappy path
A cooperative, expected flow that omits many objections, failures, or unusual conditions.
Source: CorePainPointRegression test
A repeatable check used to see whether a later system version lost previously reviewed behavior.
Source: SelfImprovementLoopTrace provenance
Information showing where a simulation record came from and how it was produced.
Source: TrustSafetyComplianceSensitive pack
A scenario collection whose subject or behavior requires extra policy review before use.
Source: AgenticAutonomyAndSafetyModelRubric agreement
The degree to which evaluators reach consistent judgments when applying stated criteria.
Source: AgentEvalMetricsVersion comparison
A review of results from the same evaluation set across different agent releases.
Source: AgentWorkflowPolicy control
A rule that limits, blocks, or escalates a proposed simulation or use.
Source: ProductFeatureSet