senior-architect
Use when the user needs system design, architecture decision records, scalability analysis, trade-off evaluation, or non-functional requirements planning. Triggers: new system design, technology selection, scaling strategy, ADR creation, infrastructure topology, service boundary definition.
- Installs
- 80
- Last 24h
- —
- Trending rank
- —
- Security
- Not audited
Install
Works with every agent the skills CLI supports.
npx skills add https://github.com/pixel-process-ug/superkit-agents --skill senior-architect- 01
Install
Run the command in your project. The CLI asks which agents to install for.
- 02
Describe the task
Ask your agent for something this skill covers. It loads senior-architect automatically when relevant.
- 03
Review the result
The agent follows the SKILL.md steps, scripts and references bundled with the skill.
SKILL.md
The instructions your agent reads when the skill is triggered.
Senior Architect
Overview
Provide architecture-level guidance for system design decisions. This skill produces Architecture Decision Records (ADRs), trade-off analyses, scalability blueprints, and non-functional requirements specifications. Every recommendation includes explicit trade-offs and is grounded in proven patterns.
Announce at start: "I'm using the senior-architect skill for system design and architecture decisions."
Phase 1: Requirements Analysis
Goal: Capture all functional and non-functional requirements before designing.
Actions
- Identify functional requirements (capabilities)
- Define non-functional requirements (quality attributes)
- Identify constraints (budget, team, timeline, compliance)
- Map integration points with existing systems
- Establish success criteria and SLOs
STOP — Do NOT proceed to Phase 2 until:
- Functional requirements are listed
- Non-functional requirements are quantified (not vague)
- Constraints are explicit
- Success criteria are measurable
Phase 2: Architecture Design
Goal: Evaluate options and select the approach with the best trade-off profile.
Actions
- Evaluate architectural styles (monolith, microservices, event-driven)
- Design component boundaries and interfaces
- Define data architecture (storage, flow, consistency)
- Plan infrastructure and deployment topology
- Address cross-cutting concerns (auth, logging, monitoring)
Architecture Style Decision Table
| Factor | Monolith | Modular Monolith | Microservices | Serverless |
|---|---|---|---|---|
| Team size < 10 | Preferred | Strong fit | Overkill | Good for bursty |
| Team size > 30 | Challenging | Good | Preferred | Depends |
| Domain well-understood | Good fit | Good fit | Not needed yet | Good fit |
| Domain evolving rapidly | Fine to start | Good fit | Too early | Good fit |
| Need independent deployment | Not possible | Limited | Key benefit | Built-in |
| Operational maturity low | Good fit | Good fit | High risk | Managed risk |
| Variable/bursty load | Over-provisioned | Over-provisioned | Possible | Strong fit |
Default Recommendation
Start with Modular Monolith: clear module boundaries, single deployment. Extract to microservices only when you have proven need for independent scaling, deployment, or team autonomy.
Trade-Off Analysis: Common Pairs
| Improving | May Degrade |
|---|---|
| Consistency | Availability, Latency |
| Performance | Maintainability, Cost |
| Security | Usability, Performance |
| Scalability | Simplicity, Cost |
| Flexibility | Performance, Complexity |
| Time to Market | Quality, Scalability |
Decision Matrix Template
Weight each quality attribute (1-5), score each option (1-5), multiply and sum.
| Quality Attribute | Weight | Option A | Option B | Option C |
|--------------------|--------|----------|----------|----------|
| Performance | 4 | 4 (16) | 3 (12) | 5 (20) |
| Maintainability | 5 | 5 (25) | 4 (20) | 2 (10) |
| Scalability | 3 | 3 (9) | 5 (15) | 4 (12) |
| Cost | 4 | 4 (16) | 2 (8) | 3 (12) |
| Total | | 66 | 55 | 54 |
STOP — Do NOT proceed to Phase 3 until:
- At least 2 architectural options have been evaluated
- Trade-offs are explicitly documented
- Decision matrix scores support the recommendation
- Data architecture is defined
Phase 3: Documentation and Validation
Goal: Record decisions and validate against requirements.
Actions
- Write Architecture Decision Records for key decisions
- Create system context and container diagrams (C4 model)
- Validate against non-functional requirements
- Identify risks and mitigation strategies
- Define evolutionary architecture guardrails
ADR Format
# ADR-{number}: {Title}
## Status
Proposed | Accepted | Deprecated | Superseded by ADR-{number}
## Context
What is the issue motivating this decision?
## Decision
What change are we proposing?
## Consequences
### Positive
- [Benefit 1]
- [Benefit 2]
### Negative
- [Trade-off 1]
- [Trade-off 2]
### Risks
- [Risk and mitigation]
## Alternatives Considered
| Option | Pros | Cons | Verdict |
|--------|------|------|---------|
| Option A | ... | ... | Chosen |
| Option B | ... | ... | Rejected because... |
C4 Model Levels
| Level | Shows | When to Use |
|---|---|---|
| Level 1: System Context | Users and external systems | Always |
| Level 2: Container | Major technical building blocks | Always |
| Level 3: Component | Components within containers | Complex services |
| Level 4: Code | Class-level detail | Critical/complex areas only |
STOP — Documentation complete when:
- ADRs written for all key decisions
- System context diagram created
- NFRs validated against design
- Risks documented with mitigations
Scalability Patterns
Horizontal Scaling Decision Table
| Pattern | Use When | Implementation |
|---|---|---|
| Load Balancing | Multiple instances of same service | Round-robin, least connections, IP hash |
| Stateless Services | Need to add/remove instances freely | JWT/external session store |
| Auto-scaling | Variable load patterns | CPU/memory/request-rate triggers |
| Read Replicas | Read-heavy workloads | Route reads to replicas, writes to primary |
Sharding Strategy Decision Table
| Strategy | How | Good For |
|---|---|---|
| Hash-based | Consistent hash of key | Even distribution |
| Range-based | Date range, ID range | Time-series data |
| Geographic | By region/country | Data locality |
| Tenant-based | Per customer | Multi-tenant SaaS |
Caching Layers
Client Cache (browser) -> CDN Cache -> API Gateway Cache ->
Application Cache (Redis) -> Database Query Cache -> Database
Non-Functional Requirements Template
## Performance
- Response time: p95 < 200ms, p99 < 500ms for API calls
- Throughput: 1000 RPS sustained, 5000 RPS peak
- Batch processing: 1M records/hour
## Availability
- Target: 99.9% (8.76h downtime/year)
- RTO (Recovery Time Objective): < 15 minutes
- RPO (Recovery Point Objective): < 5 minutes
## Scalability
- Current: 10K DAU
- 12-month target: 100K DAU
- Scale dimension: users, data volume, request rate
## Security
- Authentication: OAuth 2.0 / OIDC
- Authorization: RBAC with resource-level permissions
- Data encryption: at rest (AES-256) and in transit (TLS 1.3)
## Observability
- Logging: structured JSON, 30-day retention
- Metrics: RED method, custom business metrics
- Tracing: distributed tracing across all services
- Alerting: PagerDuty integration, tiered severity
SLO/SLA/SLI Framework
| Term | Definition | Example |
|---|---|---|
| SLI | Measurable metric | Request latency, error rate |
| SLO | Target value | 99.9% availability |
| SLA | Contractual commitment | 99.5% with penalty clause |
| Error Budget | 1 - SLO | 0.1% = 8.76h/year |
Anti-Patterns / Common Mistakes
| Anti-Pattern | Why It Is Wrong | Correct Approach |
|---|---|---|
| Resume-driven architecture | Complexity without benefit | Choose simplest solution that works |
| Distributed monolith | All downsides of both | Either true monolith or true microservices |
| Premature optimization | Scaling for 1M users with 100 | Design for current + 10x, not 1000x |
| Golden hammer | One technology for everything | Right tool for each problem |
| Architecture without validation | Untested assumptions | Load test, failure test, validate |
| Big upfront design without iteration | Requirements change | Evolutionary architecture with guardrails |
| Vague NFRs ("fast", "scalable") | Cannot be validated | Quantified targets with measurement |
Subagent Dispatch Opportunities
| Task Pattern | Dispatch To | When |
|---|---|---|
| Analyzing different architecture layers | Agent tool with subagent_type="Explore" (one per layer) | When reviewing frontend, backend, and infra independently |
| Security assessment of architecture | Agent tool invoking security-review skill | When architecture involves auth, data flow, or external APIs |
| Performance implications analysis | Agent tool invoking performance-optimization skill | When architecture decisions affect latency or throughput |
| Code quality review of existing patterns | Agent tool dispatching code-reviewer agent | When evaluating current codebase for refactoring |
Follow the dispatching-parallel-agents skill protocol when dispatching.
Integration Points
| Skill | Relationship |
|---|---|
senior-backend | Backend implementation follows architecture decisions |
senior-fullstack | Full-stack architecture follows service boundaries |
security-review | Security is a cross-cutting architectural concern |
performance-optimization | Performance NFRs drive optimization targets |
planning | Architecture decisions inform implementation planning |
code-review | Review validates architectural consistency |
acceptance-testing | NFRs become acceptance criteria |
Key Principles
- Start with the simplest architecture that could work
- Make decisions reversible when possible
- Design for failure (everything will fail eventually)
- Optimize for team cognitive load, not technical elegance
- Document decisions, not just outcomes
- Prefer boring technology for critical paths
- Every architectural decision has a cost — make it explicit
Skill Type
FLEXIBLE — Adapt architecture recommendations to the specific context. ADRs are strongly recommended for all significant decisions. Trade-off analysis is mandatory. NFRs must be quantified, not described vaguely.
FAQ
What does the senior-architect skill do?
Use when the user needs system design, architecture decision records, scalability analysis, trade-off evaluation, or non-functional requirements planning. Triggers: new system design, technology selection, scaling strategy, ADR creation, infrastructure topology, service boundary definition.
How do I install senior-architect?
Run npx skills add https://github.com/pixel-process-ug/superkit-agents --skill senior-architect in your project. The skills CLI detects the agents you use and copies the skill into the right folder. For Claude Code you can also place the skill folder at ~/.claude/skills/senior-architect/ (personal) or .claude/skills/senior-architect/ (project).
Which AI agents can use senior-architect?
Any agent that supports the open Agent Skills format (a folder with a SKILL.md file), including Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, Windsurf, Cline, OpenCode and more.
Is senior-architect safe to install?
No partner security audit is available yet. Skills can run commands and scripts, so read the SKILL.md and any bundled files before installing.
Who maintains senior-architect?
It is published by pixel-process-ug in pixel-process-ug/superkit-agents. It has 80 recorded installs.
More from pixel-process-ug/superkit-agents
ui-ux-pro-max
pixel-process-ug/superkit-agents
Use when the user needs full UI/UX design intelligence — styles, palettes, fonts, UX guidelines, chart selection, and accessible, performant implementation across any supported stack. Triggers: user says "design", "UI", "UX", "color palette", "typography", "accessibility", "responsive design", "chart type", "style guide", building any user-facing interface.
frontend-ui-design
pixel-process-ug/superkit-agents
Use when designing UI components, creating component architectures, implementing responsive layouts, setting up design systems, or selecting state management solutions for frontend applications
reverse-engineering-specs
pixel-process-ug/superkit-agents
Open the skill to read its SKILL.md instructions.
canvas-design
pixel-process-ug/superkit-agents
Use when the user needs HTML Canvas, SVG graphics, data visualization, generative art, or interactive visual elements with D3.js or similar libraries. Triggers: user says "canvas", "SVG", "chart", "data visualization", "D3", "generative art", "animation", "interactive graphics", drawing on screen.
autonomous-loop
pixel-process-ug/superkit-agents
Open the skill to read its SKILL.md instructions.
laravel-boost
pixel-process-ug/superkit-agents
Open the skill to read its SKILL.md instructions.
Popular code quality & workflow skills
grill-me
mattpocock/skills
A relentless interview to sharpen a plan or design.
improve-codebase-architecture
mattpocock/skills
Scan a codebase for deepening opportunities, present them as a visual HTML report, then grill through whichever one you pick.
handoff
mattpocock/skills
Compact the current conversation into a handoff document for another agent to pick up.
domain-modeling
mattpocock/skills
Build and sharpen a project's domain model. Use when discussing codebase terminology, writing or editing a GLOSSARY.md, or recording or editing an ADR.
codebase-design
mattpocock/skills
Shared vocabulary for designing deep modules. Use when the user wants to design or improve a module's interface, find deepening opportunities, decide where a seam goes, make code more testable or AI-navigable, or when another skill needs the deep-module vocabulary.
diagnosing-bugs
mattpocock/skills
Diagnosis loop for hard bugs and performance regressions. Use when the user says "diagnose"/"debug this", or reports something broken/throwing/failing/slow.