ADPSAgent Design Patterns Society

ADPS Chronicle

Chronicle

Source dates record when a workshop, case, or study entered ADPS. First-public dates record the first appearance of each page. Prior work for established methods remains in the references on each page.

Patterns and module guides

First publicPageRecorded source
A1 · Tool DispatchA1 · Tool DispatchADPS pattern white paper; prior work and references are listed in the article
A2 · Plan-and-ExecuteA2 · Plan-and-ExecuteADPS pattern white paper; prior work and references are listed in the article
A3 · Prompt ChainingA3 · Prompt ChainingADPS pattern white paper; prior work and references are listed in the article
C1 · Hierarchical DelegationC1 · Hierarchical DelegationADPS pattern white paper; prior work and references are listed in the article
C2 · Fan-out / GatherC2 · Fan-out / GatherADPS pattern white paper; prior work and references are listed in the article
C3 · Adversarial ReviewC3 · Adversarial ReviewADPS pattern white paper; prior work and references are listed in the article
C4 · Handoff ChainC4 · Handoff ChainADPS pattern white paper; prior work and references are listed in the article
C5 · Sub-Agent IsolationC5 · Sub-Agent IsolationADPS pattern white paper; prior work and references are listed in the article
C6 · ChoreographyC6 · ChoreographyADPS pattern white paper; prior work and references are listed in the article
F1 · Generator-CriticF1 · Generator-CriticADPS pattern white paper; prior work and references are listed in the article
F2 · Skill PackageF2 · Skill PackageADPS pattern white paper; prior work and references are listed in the article
F3 · Experience ReplayF3 · Experience ReplayADPS pattern white paper; prior work and references are listed in the article
F4 · Self-Heal LoopF4 · Self-Heal LoopADPS pattern white paper; prior work and references are listed in the article
G1 · Approval GateG1 · Approval GateADPS pattern white paper; prior work and references are listed in the article
G2 · Blast-Radius ControlG2 · Blast-Radius ControlADPS pattern white paper; prior work and references are listed in the article
G3 · Progressive CommitmentG3 · Progressive CommitmentADPS pattern white paper; prior work and references are listed in the article
M1 · Hierarchical RetentionM1 · Hierarchical RetentionADPS pattern white paper; prior work and references are listed in the article
M2 · RAG Pipeline · Retrieval-Augmented GenerationM2 · RAG Pipeline · Retrieval-Augmented GenerationADPS pattern white paper; prior work and references are listed in the article
M3 · Progress TrackingM3 · Progress TrackingADPS pattern white paper; prior work and references are listed in the article
M4 · Failure JournalsM4 · Failure JournalsADPS pattern white paper; prior work and references are listed in the article
M5 · Procedural MemoryM5 · Procedural MemoryADPS pattern white paper; prior work and references are listed in the article
P1 · Context TriageP1 · Context TriageADPS pattern white paper; prior work and references are listed in the article
P2 · Semantic CompactionP2 · Semantic CompactionADPS pattern white paper; prior work and references are listed in the article
P3 · Progressive DiscoveryP3 · Progressive DiscoveryADPS pattern white paper; prior work and references are listed in the article
P4 · Multi-Modal FusionP4 · Multi-Modal FusionADPS pattern white paper; prior work and references are listed in the article
R1 · Chain-of-ThoughtR1 · Chain-of-ThoughtADPS pattern white paper; prior work and references are listed in the article
R2 · Complexity-Based RoutingR2 · Complexity-Based RoutingADPS pattern white paper; prior work and references are listed in the article
R3 · Parallel ExplorationR3 · Parallel ExplorationADPS pattern white paper; prior work and references are listed in the article
R4 · Iterative Hypothesis TestingR4 · Iterative Hypothesis TestingADPS pattern white paper; prior work and references are listed in the article
R5 · Talker-Reasoner · Dual-Process ArchitectureR5 · Talker-Reasoner · Dual-Process ArchitectureADPS pattern white paper; prior work and references are listed in the article
A4 · Guardrail SandwichA4 · Guardrail SandwichADPS pattern white paper; prior work and references are listed in the article
A5 · Minimal Tool SetA5 · Minimal Tool SetADPS pattern white paper; prior work and references are listed in the article
G4 · Hooks PipelineG4 · Hooks PipelineADPS pattern white paper; prior work and references are listed in the article
Memory Module · Turning Past Work into a Governed Runtime AssetMemory Module · Turning Past Work into a Governed Runtime AssetFirst Memory Module Workshop (5 August 2026); Dongfang Yiteng Execution Agent case
Reflection Module: Turning Feedback into Controlled ChangeReflection Module: Turning Feedback into Controlled ChangeFirst Reflection Module Workshop (12 August 2026); Deep Agents dynamic collaboration study
Action Module · Turn a Decision into a Verifiable External ChangeAction Module · Turn a Decision into a Verifiable External ChangeFirst Action Module Workshop (6 August 2026); Dongfang Yiteng Execution Agent case
Perception Module · Control What Enters the Current DecisionPerception Module · Control What Enters the Current DecisionFirst Perception Module Workshop (13 August 2026)
Governance Module: Making Agent Autonomy ManageableGovernance Module: Making Agent Autonomy ManageableFirst Governance Module Workshop (18 August 2026); DeerFlow Guardrail architecture evolution
X1 · ObservabilityX1 · ObservabilityADPS pattern white paper; prior work and references are listed in the article
X2 · Evaluation & ValidationX2 · Evaluation & ValidationADPS pattern white paper; prior work and references are listed in the article
X3 · Security & IdentityX3 · Security & IdentityADPS pattern white paper; prior work and references are listed in the article
Collaboration · How multiple participants complete one body of workCollaboration · How multiple participants complete one body of workFirst Collaboration Module Workshop (25 August 2026); Deep Agents dynamic collaboration study
Reasoning · Compiling Evidence into Reviewable DecisionsReasoning · Compiling Evidence into Reviewable DecisionsFirst Reasoning Module Workshop (26 August 2026); published ADPS Reasoning pattern specifications

Concepts

First publicPageRecorded source
Activity Events and TimelineActivity Events and TimelineDongfang Yiteng execution-agent case ()
Anchor, Ledger, CollectionAnchor, Ledger, CollectionDongfang Yiteng execution-agent case ()
Control and Narrative PlanesControl and Narrative PlanesDongfang Yiteng execution-agent case ()
Execution vs Content AgentsExecution vs Content AgentsDongfang Yiteng execution-agent case ()
HITL Block and ResumeHITL Block and ResumeDongfang Yiteng execution-agent case ()
Intent as CompilationIntent as CompilationDongfang Yiteng execution-agent case ()
Layered Memory L1/L2/L3Layered Memory L1/L2/L3Dongfang Yiteng execution-agent case ()
Mechanical State PlaneMechanical State PlaneDongfang Yiteng execution-agent case ()
Memory EnvelopeMemory EnvelopeDongfang Yiteng execution-agent case ()
Responsibility Boundary Between the Orchestrator and MessageHandlerResponsibility Boundary Between the Orchestrator and MessageHandlerDongfang Yiteng execution-agent case ()
Scratchpad Board: A Short-Lived Work Surface for an AgentScratchpad Board: A Short-Lived Work Surface for an AgentDongfang Yiteng execution-agent case ()
Task DAG and State MachineTask DAG and State MachineDongfang Yiteng execution-agent case ()
Three Session-State PlanesThree Session-State PlanesDongfang Yiteng execution-agent case ()
Capability Certification Lifecycle: How a Capability Earns Automatic Execution RightsCapability Certification Lifecycle: How a Capability Earns Automatic Execution RightsXuanxu GIS publishing-agent case ()
Disk Fact Plane: Connect Runtime Stages through External ContractsDisk Fact Plane: Connect Runtime Stages through External ContractsXuanxu GIS publishing-agent case ()
External Acceptance Probe: Let the User-Side Result Decide SuccessExternal Acceptance Probe: Let the User-Side Result Decide SuccessXuanxu GIS publishing-agent case ()
Failure-to-Rule Loop: Make an Incident Change the Next RunFailure-to-Rule Loop: Make an Incident Change the Next RunXuanxu GIS publishing-agent case ()
One-Diagram Unit of Work: Commit One Verifiable Modeling IncrementOne-Diagram Unit of Work: Commit One Verifiable Modeling IncrementAI4MBSE modeling-agent project ()
Reasoning Assetization: Move Repeated Decisions into Runtime AssetsReasoning Assetization: Move Repeated Decisions into Runtime AssetsXuanxu GIS publishing-agent case ()
Resumable Clarification Job: Pause the Same Task When Information Is MissingResumable Clarification Job: Pause the Same Task When Information Is MissingAI4MBSE modeling-agent project ()
Structural Write-Back Gate: Check Engineering Invariants before Side EffectsStructural Write-Back Gate: Check Engineering Invariants before Side EffectsAI4MBSE modeling-agent project ()
Typed Intermediate Representation Chain: Narrow Language into Verifiable StructureTyped Intermediate Representation Chain: Narrow Language into Verifiable StructureAI4MBSE modeling-agent project ()
Vocabulary Equivalence Layer: Align Planning Terms with the Host MetamodelVocabulary Equivalence Layer: Align Planning Terms with the Host MetamodelAI4MBSE modeling-agent project ()
Abstraction-Reconstruction LoopAbstraction-Reconstruction LoopFirst Collaboration workshop ()
Context ContractContext ContractFirst Perception workshop ()
Dual Feedback ClockDual Feedback ClockFirst Reflection workshop ()
Dual-Scale GovernanceDual-Scale GovernanceFirst Governance workshop ()
Durable IntentDurable IntentFirst Governance workshop ()
Handoff ContractHandoff ContractFirst Collaboration workshop ()
Hook CompositionHook CompositionFirst Collaboration workshop ()
Judge MigrationJudge MigrationFirst Perception workshop ()
Knowledge CompilationKnowledge CompilationFirst Memory workshop ()
Plan CompilationPlan CompilationFirst Action workshop ()
Probabilistic Core, Deterministic ShellProbabilistic Core, Deterministic ShellFirst Collaboration workshop ()
Three Collaboration PlanesThree Collaboration PlanesFirst Collaboration workshop ()
Three Powers of AgencyThree Powers of AgencyFirst Governance workshop ()
Topology Governance MatrixTopology Governance MatrixFirst Collaboration workshop ()
Topology LoweringTopology LoweringFirst Collaboration workshop ()
Write-Conflict DomainWrite-Conflict DomainFirst Action workshop ()
Business Evaluation SetBusiness Evaluation SetFirst Reasoning workshop ()
CodeAct: Code as ActionCodeAct: Code as ActionExecutable Code Actions Elicit Better LLM Agents, ICML 2024
Programmatic Tool CallingProgrammatic Tool CallingAnthropic Programmatic Tool Calling documentation
ReAct Reasoning-Action LoopReAct Reasoning-Action LoopReAct: Synergizing Reasoning and Acting in Language Models, 2022
Scenario–Benchmark ContractScenario–Benchmark ContractFirst Reasoning workshop ()
Unified Convergence NodeUnified Convergence NodeFirst Reasoning workshop ()

Cases and research reports

First publicPageRecorded source
Dongfang Yiteng's Execution Agent: Preserve Business State Across a WorkflowDongfang Yiteng's Execution Agent: Preserve Business State Across a WorkflowDongfang Yiteng Execution Agent case; contributed by Bo Liang
AI4MBSE Modeling Agent: Qualifying Natural-Language Plans for Model Write-BackAI4MBSE Modeling Agent: Qualifying Natural-Language Plans for Model Write-BackAI4MBSE modeling agent project; authored by Liangding Yuan
Xuanxu Technology's GIS Publishing Agent: Turn Runtime Experience into Verifiable PipelinesXuanxu Technology's GIS Publishing Agent: Turn Runtime Experience into Verifiable PipelinesXuanxu Technology GIS publishing agent case; contributed by Yuke Xiong
DeerFlow Guardrails: From Pre-Call Interception to Two-Layer AuthorizationDeerFlow Guardrails: From Pre-Call Interception to Two-Layer AuthorizationWillem Jiang's DeerFlow Guardrail architecture presentation at the Governance workshop; checked by ADPS against public pull requests, source code, and documentation
Deep Agents: From Fixed Graphs to Code-Generated CollaborationDeep Agents: From Fixed Graphs to Code-Generated CollaborationHaili Zhang's LangGraph and Deep Agents analysis at the Collaboration workshop; checked by ADPS against public documentation and source code

Workshops

First publicPageRecorded source
ADPS Design Pattern Workshop Series · First Action Module WorkshopADPS Design Pattern Workshop Series · First Action Module WorkshopADPS Design Pattern Workshop Series · First Action Module Workshop; workshop held on
ADPS Design Pattern Workshop Series · First Memory Module WorkshopADPS Design Pattern Workshop Series · First Memory Module WorkshopADPS Design Pattern Workshop Series · First Memory Module Workshop; workshop held on
ADPS Design Pattern Series: First Reflection Module WorkshopADPS Design Pattern Series: First Reflection Module WorkshopADPS Design Pattern Series: First Reflection Module Workshop; workshop held on
ADPS Design Pattern Workshop · First Perception WorkshopADPS Design Pattern Workshop · First Perception WorkshopADPS Design Pattern Workshop · First Perception Workshop; workshop held on
First Governance Module WorkshopFirst Governance Module WorkshopFirst Governance Module Workshop; workshop held on
First Collaboration Module WorkshopFirst Collaboration Module WorkshopFirst Collaboration Module Workshop; workshop held on
First Reasoning Module WorkshopFirst Reasoning Module WorkshopFirst Reasoning Module Workshop; workshop held on
Agent Evaluation and Validation WorkshopAgent Evaluation and Validation Workshop; workshop held on

Topics

First publicPageRecorded source
AI-Driven Software Engineering: Specifications, Context, Execution, and FeedbackAI-Driven Software Engineering: Specifications, Context, Execution, and FeedbackWorkshop and case records cited in the article: First ADPS Perception Module Workshop ()
Agent Evaluation and Validation: From Output Scores to System AcceptanceAgent Evaluation and Validation: From Output Scores to System AcceptanceWorkshop and case records cited in the article: Dongfang Yiteng execution agent (); Xuanxu Technology GIS publishing agent (); First ADPS Action Module Workshop (); First ADPS Reflection Module Workshop ()
Enterprise Agent Evolution and Operating Model: Research AgendaEnterprise Agent Evolution and Operating Model: Research AgendaWorkshop and case records cited in the article: First Governance Module Workshop ()
Observability-Driven Agent Evolution: From Runtime Timelines to Verifiable ChangesObservability-Driven Agent Evolution: From Runtime Timelines to Verifiable ChangesWorkshop and case records cited in the article: Dongfang Yiteng execution-agent case (); First ADPS Action Module Workshop (); First ADPS Reflection Module Workshop (); First ADPS Governance Module Workshop ()
Agent Design Lifecycle: From Capability Registration to Evolution and RetirementAgent Design Lifecycle: From Capability Registration to Evolution and RetirementWorkshop and case records cited in the article: First Governance Module Workshop ()
Common Pattern Compositions: From Task Terrain to Runtime ArchitectureCommon Pattern Compositions: From Task Terrain to Runtime ArchitectureWorkshop and case records cited in the article: Bo Liang execution-agent case (); DeerFlow Guardrail case (); DeerFlow Guardrail code and architecture evolution ()
Human-Agent Interaction: Intent, Authority, Intervention, and AccountabilityHuman-Agent Interaction: Intent, Authority, Intervention, and AccountabilityWorkshop and case records cited in the article: Bo Liang execution-agent case (); First Governance Module Workshop (); Collaboration workshop ()
Abstraction and Reconstruction · From field evidence to patterns and backAbstraction and Reconstruction · From field evidence to patterns and backWorkshop and case records cited in the article: Collaboration workshop ()
Agent OS · From analogy to engineering checklistAgent OS · From analogy to engineering checklistWorkshop and case records cited in the article: Collaboration workshop ()
Hook Composition · Make hidden callbacks a readable control structureHook Composition · Make hidden callbacks a readable control structureWorkshop and case records cited in the article: Collaboration workshop ()
Pattern Selection Card: Turning Constraints into an Agent DesignPattern Selection Card: Turning Constraints into an Agent DesignADPS topic study; evidence and references are listed in the article
Six-Step Selection Method: From Business Problem to Pattern CompositionSix-Step Selection Method: From Business Problem to Pattern CompositionADPS topic study; evidence and references are listed in the article

Technical positions

First publicPageRecorded source
Evaluation as Engineering: From Design Constraints to Production EvidenceEvaluation as Engineering: From Design Constraints to Production EvidenceADPS technical position; arguments and references are listed in the article
Governance and Safety: Engineering Runtime ControlGovernance and Safety: Engineering Runtime ControlADPS technical position; arguments and references are listed in the article
Multi-Agent Systems: Adoption Criteria and Engineering BoundariesMulti-Agent Systems: Adoption Criteria and Engineering BoundariesADPS technical position; arguments and references are listed in the article
Where Prompt Engineering Fits in an Agent SystemWhere Prompt Engineering Fits in an Agent SystemADPS technical position; arguments and references are listed in the article