Skip to main content
INFORMATIVEDRAFTDocumentation Governance

Golden Flows — Informative (Non-Normative)

Golden Flows are evaluation scenarios that describe expected behavior for MPLP implementations.

How to Use This Section

Your GoalStart Here
Verify single-agent implementationFLOW-01: Single Agent – Happy Path
Test tool integrationFLOW-03: Single Agent – With Tools
Validate confirmation flowsFLOW-05: Single Agent with Confirm Required
Understand SA profileSA-FLOW-01: SA Basic Execution
Understand MAP coordinationMAP-FLOW-01: MAP Turn-Taking Session

Core Flows (v1.0 Compliance Boundary)

FlowNamePurposeModules Involved
FLOW-01Single Agent – Happy PathBaseline 2-step plan executionContext, Plan, Trace
FLOW-02Single Agent – Large PlanVolumetric 20+ step validationContext, Plan, Trace
FLOW-03Single Agent – With ToolsTool integration via agent_roleContext, Plan, Trace
FLOW-04Single Agent with LLM EnrichmentAEL reasoning integrationContext, Plan, Trace, Core
FLOW-05Single Agent with Confirm RequiredMulti-round approvalContext, Plan, Confirm, Trace

SA Profile Flows

FlowNamePurposeModules Involved
SA-FLOW-01SA Basic ExecutionSingle-step SA lifecycleContext, Plan
SA-FLOW-02SA Multi-Step EvaluationMulti-step SA validationContext, Plan

MAP Profile Flows

FlowNamePurposeModules Involved
MAP-FLOW-01MAP Turn-Taking SessionSequential multi-agent coordinationContext, Plan, Collab, Role
MAP-FLOW-02MAP Broadcast Fan-outParallel multi-agent dispatchContext, Plan, Collab, Role

Traceability

All Golden Flows are mapped to:

  • Modules they validate
  • Cross-Cutting Duties they exercise
  • Test files in tests/golden/flows/

Change Policy

Golden Flows are Normative:

  • Semantic changes require MPGC approval
  • Test files in tests/golden/flows/ are the source of truth
  • All downstream content must align with source of truth

Document Status: Normative (Validation Scenarios)
Total Flows: 9 (5 Core + 2 SA + 2 MAP)
Source of Truth: tests/golden/flows/