0% read on this device
Browse the curriculum

Start Here

Vision

First Principles

Operating Model

Domain Model

Agent Factory

Runtime Architecture

AI Engineering

Autonomous Workflows

Verification & Delivery

Factory Platform

Quality Engineering

Security & Governance

Case Studies

Labs

Interview Practice

Research Journal

Reference

Curriculum/Start Here/Complete source chapter
Start Here5 min readchapterQuick Read

Detailed Architecture Coverage Matrix

Give every material factory responsibility one canonical name, owner, chapter, control boundary, and validation path.

Status: Review readyRisk: variableLifecycle: intent · plan · execute · verify · deliver · learnContent reviewed 2026-08-30Maturity guide →
Claim boundaryThis is curriculum guidance. It does not by itself prove a production implementation.
Quick Read

The chapter in one pass.

~2 min
  • Purpose: Give every material factory responsibility one canonical name, owner, chapter, control boundary, and validation path.
  • Use it when: Reviewing scope, assigning an architecture question, or checking whether a diagram creates a competing taxonomy.
  • Core rule: One responsibility may have many implementations, but it has one accountable curriculum location.
  • Evidence boundary: Coverage records documentation ownership. It does not prove that an implementation operates as described.

1. The problem

Broad architecture maps are good orientation aids and poor accountability systems. The same idea is often labeled differently in lifecycle, runtime, security, and operations diagrams. Gaps hide behind overlapping labels while attractive visuals make partial coverage look complete.

This matrix is the normalization layer. Detailed chapters own specifications; the matrix owns traceability among capability, lifecycle, plane, role, risk, maturity, evidence boundary, and validation.

2. Canonical coverage matrix

CapabilityLifecycleOwning planeAccountable roleRiskCanonical specificationDocumentation maturityImplementation evidence boundaryValidation
Business intent and accepted outcomeIntent, learnHuman governanceBusiness ownerVariableIntent-to-delivery lifecycleReview readyNo complete production outcome chain is claimedIntent and outcome trace review
Plan and executable specificationPlanControlEngineering ownerHighSpecification engineeringReview readyRepresentative plan-assurance evidence is incompleteAmbiguity and acceptance test
Factory system inventoryIntent, operate, retireControl/dataSystem ownerHighInventory and classificationReview readyNo production-wide inventory completeness claimRequired-field and lifecycle audit
Capability registry and lifecycleDefine, learnCapabilityCapability ownerHighCapability supply chainReview readyUnified production registry is unprovenCertification and revocation lab
Tool, skill, and integration contractsDefine, executeCapability/executionCapability ownerHighCapability contract referenceReview readyExisting tools are not asserted to meet every fieldContract conformance suite
Architecture and autonomy selectionPlanControlArchitectHighPattern selection ladderReview readyArchitecture decisions require local evidenceSimpler-alternative review
Orchestration and durable workflowPlan, executeControl/executionRuntime ownerCriticalOrchestration contractsReview readyFull component conformance is unprovenState, stop, recovery, and replay tests
Agent, model, prompt, and tool compositionExecuteExecutionAI engineering ownerHighAgent architectureReview readyVersioned case evidence has bounded gapsFrozen-configuration replay
Knowledge ingestion and retrievalPlan, executeKnowledgeKnowledge ownerHighKnowledge pipeline specificationReview readyProduction source registry and benchmark are unprovenPermission, freshness, deletion, and retrieval tests
Multi-agent collaborationPlan, execute, verifyExecution/qualityWorkflow ownerHighMulti-agent topologiesReview readyIndependence and benefit must be proven per workflowDelegation, disagreement, and correlation tests
Sandbox and compute isolationExecuteExecution/securityPlatform ownerCriticalSandboxed executionDraft for studyProduction escape resistance is unprovenEscape, teardown, and residue tests
Verification and evidenceVerifyQualityQuality ownerCriticalQuality and evidence architectureReview readyComplete independent proof path is unprovenFault-sensitivity and provenance tests
Human review and authorityIntent, verify, deliverGovernanceNamed decision ownerCriticalAuthority and emergency controlReview readyTested response time is not claimedOverride, dual-control, and containment lab
Governance control frameworkAllSecurity/governanceControl ownerCriticalAgentic governance controlsReview readyControl design is not operating effectivenessControl evidence inspection
Organizational decision rightsAllHuman governanceExecutive sponsorCriticalGovernance operating modelReview readyLocal assignments remain organization-specificDecision simulation and RACI review
Delivery and rollbackDeliverDeliveryRelease ownerCriticalProgressive deliveryReview readyComplete production delivery path is unprovenCanary and rollback lab
Scheduling, capacity, and costExecutePlatformPlatform operationsHighOperations and FinOps referenceReview readyFleet-scale behavior is unprovenLoad, fairness, budget, and attribution tests
Observability and forensicsExecute, learnObservability/dataReliability ownerHighObservability semanticsReview readyComplete production semantic conformance is unprovenTrace completeness and forensic reconstruction
Monitoring, detection, and responseOperate, learnOperationsIncident ownerCriticalControl tower responseReview readyResponse effectiveness needs exercised incidentsDetection-to-closure simulation
Continual improvementLearnControl/qualityChange ownerHighGoverned continual learningReview readyAutomated promotion is neither required nor claimedBaseline, candidate, approval, rollback test
AI systems foundationsSupportingCross-cuttingArchitectVariableAI systems primerReview readyEducational background onlyDecision-focused teach-back

3. Canonical relationships

The lifecycle is Intent -> Plan -> Select -> Execute -> Verify -> Evidence -> Decide -> Deliver -> Observe -> Learn. Logical planes are responsibility boundaries, not mandated services. The governed inventory points to—but never duplicates—the service, capability, model, policy, and evidence registries. Architecture patterns select the minimum sufficient autonomy; maturity labels describe documentation or scoped evidence, not how impressive a pattern is.

4. Classification rules

  1. A new label must map to an existing capability or justify a new owner.
  2. Product and vendor names are examples, never canonical components.
  3. General AI background is supporting material unless it changes a factory architecture decision.
  4. A diagram without a text or table equivalent is incomplete.
  5. Review-ready documentation never advances a capability to operationally proven status.
  6. A shared concern may cross planes, but one role owns the decision and one record is authoritative.

5. Standards baseline

ReferenceState used hereUse
NIST AI RMF 1.0 and Generative AI ProfilePublishedRisk, governance, measurement, and management framing
NIST SSDF publicationsPublished baseline with later revision work tracked separatelySecure software lifecycle controls
OWASP Agentic AI threats and mitigationsCurrent community guidanceAgent, tool, context, memory, identity, and autonomy threats
SPIFFE Workload APIPublished specificationWorkload identity and credential delivery boundary
SLSA 1.2PublishedBuild provenance and supply-chain integrity
OpenTelemetry generative-AI semantic conventionsDevelopingTelemetry vocabulary; pin exact versions
WCAG 2.2W3C RecommendationAccessible interaction and diagram equivalents

6. Review exercise

Choose one production change and one failure. Trace both across the matrix. For each transition name the actor, identity, authoritative record, policy, evidence, stop condition, recovery action, and human decision. Record any responsibility with two owners or no owner as a taxonomy defect.

7. Explicit nonclaims

This matrix does not certify implementations, prescribe an organization chart, or require separate deployments for each plane. It is a review-ready ownership map awaiting external architecture, security, operations, and usability review.

External review

Review this chapter.

Challenge a claim, boundary, missing failure mode, unclear term, or unsupported evidence statement.

  • Claim
  • Boundary
  • Failure
  • Evidence