---

name: senior-system-architecture description: | Guides senior system and solution architecture—cross-service boundaries, integration patterns, non-functional requirements (scale, reliability, security, cost), ADRs, C4-style modeling, architecture review, build-vs-buy, and phased migration (strangler, dual-write). Use when designing multi-service systems, evaluating platform or vendor choices, writing or reviewing architecture decision records, defining standards and principles, or assessing technical risk across domains—not for single-service RFCs and module design (senior-software-engineer), data platform or mesh decisions (data-architect), cloud landing zone, Well-Architected, and migration architecture (cloud-architect), cloud/IaC implementation (infrastructure-engineer, cloud-engineer), internal developer platform product (platform-engineer), or program tracking (technical-program-manager). For business strategy and cases, use business-consultant; for applied AI (RAG, agents, copilots), use applied-ai-architect-commercial-enterprise.

Senior System Architecture

When to Use

• Define or review cross-service/system boundaries and contracts
• Compare architectural options with explicit trade-offs and NFR impact
• Author or review ADRs for decisions that are hard to reverse
• Run architecture review before major launches or vendor commitments
• Plan migration off legacy systems (strangler, parallel run, cutover)
• Establish architecture principles, standards, and exception process

When NOT to Use

• Single-team service RFC and implementation slices → senior-software-engineer
• Lakehouse, mesh, or enterprise data modeling → data-architect
• VPC, Kubernetes, Terraform, or CI/CD build → infrastructure-engineer, devops
• IDP, golden paths, Backstage → platform-engineer
• Milestones, RAID, steering status → technical-program-manager
• Release cutover tactics only → deployment-strategist
• Security control catalog and enterprise GRC → cybersecurity
• LLM/RAG system design → ai-engineer
• Business strategy, issue trees, steerCo cases → business-consultant
• Applied AI / LLM solution architecture → applied-ai-architect-commercial-enterprise
• Customer-facing solution design, RFP/RFI, PoC scope, deal integration → solutions-architect

Related skills

Need	Skill
Service-level RFC and code review	senior-software-engineer
Data domain and governance architecture	data-architect
Cloud solution and migration architecture	cloud-architect
Enterprise cloud governance and landing zones	enterprise-cloud-architect
Cloud/network/IaC delivery	infrastructure-engineer, cloud-engineer
Platform-as-product and golden paths	platform-engineer
Multi-team launch coordination	technical-program-manager
Rollout and rollback planning	deployment-strategist
Security architecture and controls	cybersecurity, information-security-engineer
Requirements and business constraints	business-analyst
Strategy, business case, operating model	business-consultant
AI/LLM solution patterns	ai-engineer
Commercial/enterprise applied AI architecture	applied-ai-architect-commercial-enterprise
Customer deal solution, RFP, PoC handoff	solutions-architect

Core Workflows

1. Frame the decision

Capture before drawing boxes:

• Business outcome and measurable success criteria
• Constraints: budget, timeline, compliance, existing estate
• Non-goals (explicit scope cuts)
• Stakeholders and decision owner
• Reversibility: one-way door vs two-way door

One-way doors require ADR + architecture review. Two-way doors can stay in team RFC.

2. Model the system (C4-lite)

Minimum views for reviewers:

1. Context — actors, external systems, trust zones
2. Containers — deployable units, data stores, queues, who owns each
3. Critical path — sequence diagram for highest-risk flows only

Label every arrow: sync/async, protocol, auth model, and failure behavior.

See references/integration_patterns.md.

3. Define NFRs and quality attributes

For each capability, specify targets (not vague "high availability"):

Attribute	Example target	Verification
Availability	99.9% monthly	SLO, error budget
Latency	p99 < 300ms read	Load test + prod SLO
Throughput	5k RPS peak	Capacity model
Durability	RPO 1h, RTO 4h	DR drill
Security	mTLS east-west, OIDC	Threat model link
Cost	<$X / 1M requests	FinOps estimate

See references/nfr_quality_attributes.md.

4. Evaluate options

Present at least two viable options plus "do nothing / minimal change":

Criterion	Weight	Option A	Option B
Time to value
Operational burden
Scalability headroom
Team skill fit
Vendor lock-in
Security/compliance fit

Recommend one; document rejected options and why.

See references/adr_template.md.

5. Architecture review

Before build or contract signature:

1. Problem and constraints restated in one paragraph
2. Diagrams current; contracts versioned (OpenAPI, event schema)
3. Failure modes: partial outage, dependency down, poison messages
4. Data: ownership, retention, PII, migration path
5. Observability: golden signals per container
6. Security: authz boundaries, secrets, blast radius
7. Rollout and rollback linked to deployment-strategist if multi-phase

See references/architecture_review.md.

6. Evolution and migration

For legacy replacement:

1. Identify capability slices that can move independently
2. Prefer strangler over big-bang when risk is high
3. Define parity criteria before cutover
4. Plan dual-write / dual-read duration and reconciliation
5. Deprecate old path with telemetry proving zero traffic

See references/migration_evolution.md.

When to load references

• ADR format and bar → references/adr_template.md
• Review checklist → references/architecture_review.md
• Sync, events, sagas, APIs → references/integration_patterns.md
• SLOs, capacity, DR, cost → references/nfr_quality_attributes.md
• Strangler and cutover → references/migration_evolution.md