How we work · 02
How we work
Legacy Modernization
Safe, incremental, AI-assisted evolution
Specialised agents
- 1System DiscoveryUnderstand what you actually have
- 2Incremental PathsStrangler-pattern evolution
- 3Risk AnalysisSequence the work by risk
- 4Domain OntologyCanonicalise the business vocabulary
- 5Process ReconstructionPreserve the operational knowledge
1.
System Discovery Agent
$claude /system-discovery
- 0ReconnaissanceStack, versions, project layout, entry points — ground truth first
- 1ArchitectureReconstruct layers, components, and external dependencies from the code
- 2CouplingInter-module dependency graph, cycles, god classes ranked by size
- 3Dead codeUnreferenced classes, disabled tests, dormant features, deprecated APIs
- 4API surfaceEvery endpoint enumerated — auth gaps, undocumented operations called out
- 5Operational painCI/CD, committed secrets, build friction, observability — or their absence
Agent 1 · System Discovery
Architecture Mapped by the Agent
Agent 1 · System Discovery
Legacy Profile
2.
Modernization Pathways Agent
$claude /modernization-pathways
- 0Current stateEstablish today's posture from the discovery report — coupling, scale signals, operational pain
- 1Decision contextInfer organizational factors from the repo: release cadence, team shape, ops capability, compliance signals
- 2Confirm contextValidate non-code factors with the team — team size, on-call maturity, desired release cadence
- 3Three pathwaysA: Stabilized Modular Monolith · B: Hybrid Services · C: Event-Driven — concrete moves, real costs, fit conditions
- 4Tradeoff matrixScore each pathway: migration risk, operational burden, scaling ceiling, team fit, time to value
- 5Fit assessmentMap the confirmed context to pathways — a contextual starting point, not a verdict
Agent 2 · Modernization Pathways
Decision context — repo-inferred and user-confirmed factors shape the destination
Agent 2 · Modernization Pathways
Agent 2 · Modernization Pathways
Pathway B · Hybrid Service Architecture 
Pathway C · Event-Driven Evolution
3.
Risk Analysis
- Modernization risk scoring and migration sequencing
- Critical dependency identification
- Operational hotspots and deployment sensitivity
Agent 3 · Risk Analysis
Top Risk Drivers
Agent 3 · Risk Analysis
Risk Scoring
Agent 3 · Risk Analysis
Migration Sequence
4.
Domain Ontology Agent
$claude /domain-ontology
- 0Current stateBuild on the discovery report — entities, DTOs, controllers already mapped
- 1Term inventoryCollect every business term from entities, DTOs, endpoints, events, and frontend types
- 2Concept clusteringGroup terms that refer to the same business idea across modules
- 3SynonymsResolve synonym sets — Customer / User / Member / Account across the codebase
- 4PolysemesSurface where the same word means different things in different contexts
- 5Canonical vocabularyOne agreed term per concept, chosen for business clarity and API contract compatibility
Agent 4 · Domain Ontology
Canonical terms with definitions 
Synonym sets resolved across modules
Canonical Vocabulary and Synonym Sets
Agent 4 · Domain Ontology
Concept Cluster
5.
Process Reconstruction Agent
$claude /process-reconstruction
- 0Current stateBuild on discovery and ontology reports — entry points and actors already mapped
- 1Trigger inventoryEvery way a process starts: HTTP, message queues, scheduled jobs, admin actions, webhooks
- 2Process candidatesGroup triggers into end-to-end flows using entity lifecycles, verb-noun patterns, and saga classes
- 3ReconstructionTrace each process step by step — actor, action, side effects, branch points, each cited to file:line
- 4Human-in-the-loopSurface every manual step: approvals, data entry, customer waits — and flag automation candidates
- 5Bottlenecks & exceptionsSync external calls in hot paths, silent catch blocks that change flow, missing retry and idempotency
Agent 5 · Process Reconstruction
P-1 · Customer Registration 
P-3 · Order Status Lifecycle
Customer Registration and Order Lifecycle
Agent 5 · Process Reconstruction
Order Placement
More agents in the programme
The five shown cover discovery and planning. Several more exist.
- Modernisation Planning — sequenced delivery plan
- Ballpark Estimation — effort and cost signals
- Consistency Check Agent — keeps outputs coherent, resolves conflicts with human input
- Each agent's output is refined by a human architect with AI