Agent Roster¶

AgentArmy runs two agent armies. This document covers the Claude Code army — specialist sub-agents invoked from the local CLI. For the GitHub Copilot army (PR review, coding agent, @board-manager extension), see copilot.md.

📊 Agent Counts — See AGENT_COUNTS.md for an auto-updated inventory of agents per category.

When to use Claude Code vs Copilot¶

Signal	Use
Issue Size XS/S + Type Bug/Story, clear acceptance criteria	Copilot coding agent (`copilot-task` label)
Issue Size M+ or Type Feature/Enabler/Epic	Claude Code
PR < 200 lines	Copilot auto-review is sufficient
PR > 200 lines (`needs-deep-review` label)	Run `/review-pr` in Claude Code
Architecture decision needed	Claude Code `architect-reviewer`
Security-sensitive change	Claude Code `/security-review`

How to Use Claude Code Agents¶

Claude Code automatically routes to the right agent based on task description. You can also invoke explicitly:

Use the architect-reviewer agent to evaluate this design.

Or for parallel work:

Run the code-reviewer and security-auditor agents in parallel on this PR.

Design-Time Agents¶

Use these during PI planning, sprint planning, and feature refinement.

Agent	When to use
`product-manager`	Feature prioritisation, roadmap decisions, OKR alignment
`business-analyst`	Writing user stories, acceptance criteria, process mapping
`architect-reviewer`	Architecture decisions, tech stack evaluation, ADRs
`agent-distinctiveness-advocate`	MECE audit governance: validate new agents for routing clarity, resolve overlap disputes, maintain roster distinctiveness
`scrum-master`	Sprint ceremonies, retrospectives, impediment removal, velocity
`ui-designer`	Component design, design systems, visual hierarchy, accessibility
`ux-researcher`	Usability analysis, persona development, user journey mapping
`api-designer`	REST/GraphQL API design, OpenAPI specs, versioning strategy
`data-scientist`	Analytics requirements, ML feasibility, data modelling
`market-researcher`	Competitive analysis, market sizing, customer discovery
`spike-researcher`	Time-boxed technical spikes: library eval, runnable PoC code, build-vs-buy recommendation

Build-Time Agents¶

Use these during sprint execution.

Frontend¶

Agent	Speciality
`frontend-developer`	Multi-framework (React, Vue, Angular) full-stack integration
`react-specialist`	React 18+, hooks, state management, performance
`nextjs-developer`	Next.js 14+ App Router, server components, SEO
`typescript-pro`	Advanced type system, generics, type-level programming
`vue-expert`	Vue 3, Composition API, Nuxt 3
`angular-architect`	Angular 15+, RxJS, micro-frontends
`ui-designer`	Visual polish, design system implementation

Backend¶

Agent	Speciality
`backend-developer`	APIs, microservices, scalability, production architecture
`async-messaging-engineer`	Event-driven messaging: brokers (Kafka/RabbitMQ/SQS-SNS/NATS), AsyncAPI schemas, DLQ, consumer groups
`python-pro`	FastAPI, async Python, type-safe production code
`node-specialist`	Node.js APIs, CLIs, microservices
`golang-pro`	Concurrent Go systems, microservices, cloud-native
`java-architect`	Spring Boot, enterprise Java, microservices
`rust-engineer`	Systems programming, performance-critical code
`fastapi-developer`	Python async APIs, Pydantic v2
`django-developer`	Django 4+, REST APIs, async views

Data & AI¶

Agent	Speciality
`dlt-engineer`	dlt pipelines, source connectors, incremental loading, DuckDB/BigQuery/Snowflake
`data-engineer`	Pipelines, ETL/ELT, data platforms
`machine-learning-engineer`	Model serving, training pipelines, MLOps
`ai-engineer`	End-to-end AI systems, RAG, fine-tuning
`llm-architect`	LLM system design, inference, multi-model deployments
`nlp-engineer`	NLP pipelines, text processing, domain-specific models
`database-administrator`	High-availability, backup, disaster recovery
`schema-migration-engineer`	DB schema versioning & migration orchestration (Flyway/Liquibase/Alembic), zero-downtime evolution
`sql-pro`	Query optimisation, schema design, multi-database

Mobile & Desktop¶

Agent	Speciality
`mobile-web-specialist`	Responsive web, touch UX, mobile media queries, canvas/WebGL sizing, iOS Safari quirks
`react-native` / `expo-react-native-expert`	Cross-platform mobile, native modules
`flutter-expert`	Flutter 3+, custom UI, iOS/Android/Web
`swift-expert`	Native iOS/macOS, SwiftUI, async/await
`electron-pro`	Desktop apps, native OS integration, distribution

Quality Agents¶

Run these before merging or releasing.

Agent	When to use
`code-reviewer`	Code quality, security vulnerabilities, best practices
`security-auditor`	Compliance assessments, systematic vulnerability analysis
`security-engineer`	Threat modelling, zero-trust design, security automation
`penetration-tester`	Authorised offensive testing, vulnerability validation
`qa-expert`	Test strategy, quality metrics, test planning
`test-automator`	Automated test frameworks, CI/CD test integration
`contract-test-engineer`	Consumer-driven contract testing (Pact), provider verification, cross-spoke schema-drift detection
`performance-engineer`	Bottleneck identification, profiling, optimisation
`accessibility-tester`	WCAG compliance, assistive technology support

Skill shortcuts (run in current session without spawning an agent): - /review-pr — runs code-reviewer + security-auditor on the current PR - /security-review — security audit of current branch changes

Operations Agents¶

Use these for infrastructure, deployment, and reliability work.

Agent	When to use
`devops-engineer`	CI/CD pipelines, containerisation, deployment workflows
`deployment-engineer`	Pipeline design, deployment automation strategies
`sre-engineer`	SLOs, error budgets, reliability, incident response
`cloud-architect`	Multi-cloud strategy, migration, cost optimisation (use provider-specific agents for implementation)
`gcp-infra-engineer`	GCP Cloud Run, GKE, Cloud SQL, Vertex AI, IAM, Terraform for GCP
`aws-infra-engineer`	ECS/Fargate, RDS/Aurora, Bedrock, EKS, CDK/CloudFormation, IAM/SCP
`vercel-engineer`	Vercel Functions (serverless + edge), Postgres/KV/Blob, edge middleware, preview deploys, AI SDK
`kubernetes-specialist`	K8s cluster design, workload management
`terraform-engineer`	Infrastructure as code, multi-cloud IaC
`docker-expert`	Container images, orchestration, security hardening
`network-engineer`	Cloud network design, hybrid connectivity
`database-optimizer`	Query optimisation, indexing, execution plans
`observability-engineer`	Telemetry production: OpenTelemetry instrumentation, metrics/logs/traces pipelines, Grafana/Prometheus dashboards
`finops-engineer`	Cloud cost engineering: cost visibility, unit economics, rightsizing, commitments, showback/chargeback
`api-gateway-engineer`	API gateway config & policy: rate limiting, edge authN/Z, routing across spokes, developer portal
`feature-flag-engineer`	Feature flags & progressive delivery: targeting, lifecycle, kill switches, flag-debt cleanup
`release-manager`	Cross-spoke release trains: dependency-order cut & tagging, cross-repo changelog aggregation, semver

Management & Planning Agents¶

Agent	When to use
`project-manager`	Project plans, risk management, stakeholder coordination
`scrum-master`	Agile ceremonies, velocity, impediment removal
`risk-manager`	Risk identification, quantification, control frameworks
`compliance-auditor`	GDPR, HIPAA, PCI DSS, SOC 2 compliance
`technical-writer`	API docs, user guides, SDK documentation
`documentation-engineer`	Documentation systems, architecture docs
`developer-advocate`	DevRel: sample apps, external tutorials, community, developer feedback loops

Enterprise Architecture Agents¶

TOGAF ADM-aligned EA specialists for US commercial and federal contexts. All agents are designed around NIST frameworks, FedRAMP, FISMA, and US regulatory requirements — no UK-specific framework dependencies.

EA Routing¶

Engagement	Start with	Then involve
New technology investment	`enterprise-architect`	All EA specialists in ADM phase sequence
Strategic direction question	`wardley-strategist` (`/wardley`)	`capability-planner`, `enterprise-architect`
Capability model / investment	`business-architect` (`/capability-map`)	`capability-planner`
Platform / IDP architecture	`platform-architect`	`security-architect`, `integration-architect`
FedRAMP / FISMA / CMMC	`us-regulatory-architect`	`security-architect`
HIPAA data system	`us-regulatory-architect`	`information-architect`, `security-architect`
Architecture Decision	`/ea-adr`	(standalone skill)

EA Agents¶

Agent	TOGAF Phase	Purpose
`enterprise-architect`	All phases	TOGAF ADM orchestrator, Architecture Vision, Repository governance
`togaf-adm-advisor`	All phases	Phase deliverable templates, ADM tailoring, artifact guidance
`wardley-strategist`	A, B, E	Wardley Maps: value chain, evolution, doctrine, climate, gameplay
`business-architect`	B	Business capabilities, value streams, operating models (BIZBOK)
`solution-architect`	E, F	ABB→SBB, solution docs, vendor evaluation, transition architecture
`information-architect`	C (Data)	Conceptual/logical data models, MDM, data governance (DAMA DMBOK)
`capability-planner`	B, E, F	WSJF prioritization, investment heat maps, portfolio backlog
`integration-architect`	C (App), D	API governance, EDA, canonical data model, ESB modernization
`security-architect`	Cross-cutting	Zero Trust (NIST SP 800-207), NIST CSF 2.0, FedRAMP, CMMC
`platform-architect`	D	IDP, Team Topologies, Backstage, golden paths, DORA metrics
`us-regulatory-architect`	Cross-cutting	FISMA/RMF, HIPAA, CMMC 2.0, PCI DSS v4, SOX ITGC, CCPA/CPRA

EA Skills¶

Skill	What it does
`/wardley [domain]`	Full 5-stage Wardley pipeline → OWM syntax for create.wardleymaps.ai
`/ea-adr [decision]`	MADR v4.0 Architecture Decision Record
`/capability-map [domain]`	Business capability model + investment heat map

Full EA Program (TOGAF ADM)¶

1. enterprise-architect    → Architecture Vision (Phase A), Architecture Principles
2. wardley-strategist      → strategic landscape map (/wardley)
3. business-architect      → capabilities + value streams (Phase B)
4. capability-planner      → WSJF investment priorities
5. information-architect   → data architecture (Phase C)
6. integration-architect   → application integration (Phase C)
7. security-architect      → security by design (cross-cutting)
8. us-regulatory-architect → compliance constraints (cross-cutting)
9. platform-architect      → technology standards (Phase D)
10. solution-architect     → solution building blocks + transitions (Phase E/F)
11. enterprise-architect   → Architecture Contract (Phase G)

Knowledge & Ontology Agents¶

The semantic layer — foundational/applied ontologies, vocabularies, and knowledge graphs. Backs the ontology-derived generative pipeline (UFO authoring + BFO/CCO interop projection). Arranged on a formality gradient for unambiguous routing. Full README: .claude/agents/categories/12-knowledge-ontology/README.md.

Agent	Speciality	Skill	Model
`ontologist-ufo`	UFO/OntoUML/gUFO conceptual modeling — stereotypes, relators, anti-patterns (primary authoring discipline)	`ufo-ontology`	opus
`ontologist-bfo`	BFO 2020 / OBO / CCO realist ontology + BFO/CCO interop projection	`bfo-ontology`	opus
`ontologist-generalist`	Foundation-agnostic OWL/RDFS/SHACL, reuse & alignment, competency questions; cluster router	both (lightly)	sonnet
`knowledge-engineer`	KG construction & population, rules, reasoners, SPARQL, KB lifecycle	both	sonnet
`taxonomist`	Taxonomies, thesauri, SKOS, controlled vocabularies, facets (non-axiomatized)	—	sonnet

Boundaries: information-architect owns enterprise data architecture/governance (not formal semantics); knowledge-synthesizer learns from agent interactions (not domain KGs); nlp-engineer extracts from text (knowledge-engineer integrates & reasons).

Dual upper-ontology projection ("offer both")¶

1. ontologist-ufo        → OntoUML/gUFO source model (the design discipline)
2. ontologist-bfo        → BFO/CCO realist sidecar (the interop projection)
   ↳ ship mapping table + divergence list (best-effort, not a round-trip)
3. knowledge-engineer    → populate & reason over both projections

Agent Chaining Patterns¶

Feature implementation (SAFE Story)¶

1. business-analyst     → write acceptance criteria
2. architect-reviewer   → confirm technical approach
3. frontend-developer   → implement UI
4. backend-developer    → implement API
5. code-reviewer        → review PR
6. security-auditor     → security check
7. qa-expert            → test plan

PI Planning support¶

1. product-manager      → prioritise Feature backlog
2. business-analyst     → decompose Features into Stories
3. scrum-master         → estimate capacity, assign to Iterations
4. architect-reviewer   → identify Enabler stories
5. risk-manager         → flag programme-level risks

Production incident¶

1. devops-incident-responder  → triage and stabilise
2. sre-engineer               → root cause analysis
3. backend-developer          → implement fix
4. code-reviewer              → fast-track review
5. deployment-engineer        → deploy hotfix