Install any skill in seconds. Free to start, no credit card required.
Get Started Free →System architecture design and review. Use when designing architecture, evaluating microservices vs monolith, writing ADRs, choosing a database, planning for scalability, reviewing system design, or generating architecture diagrams.
.claude/skills/borghei-senior-architect/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | 66% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 125% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 213% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 367% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -14% | 0% |
Architecture design and analysis tools for making informed technical decisions: visualize system structure, analyze dependencies and coupling, detect architectural patterns, and run decision workflows for databases, patterns, and monolith-vs-microservices trade-offs.
Before generating diagrams or an assessment, confirm these inputs. If any is unknown or vague, ASK — do not assume:
Stop rule: ask only the 2-3 that most change the output. If the user says "just draft it," proceed and list your assumptions at the top of the artifact.
| Tool | Purpose | Command | |------|---------|---------| | architecture_diagram_generator.py | Generate component/layer/deployment diagrams from project structure | python scripts/architecture_diagram_generator.py ./project --format mermaid --type component | | dependency_analyzer.py | Score coupling and find circular dependencies across package managers | python scripts/dependency_analyzer.py ./project --output json --check circular | | project_architect.py | Detect architecture pattern, layer violations, and code smells | python scripts/project_architect.py ./project --check layers --verbose |
Run any script with --help for full flags.
Load the reference that matches the task — keep this file lean and pull detail on demand:
Covers: system-level architecture analysis (pattern detection, layer validation, component diagramming) for existing codebases; technology-agnostic dependency analysis across npm, pip, Poetry, Go modules, and Cargo; architecture decision workflows (database, pattern, monolith-vs-microservices); diagram generation in Mermaid, PlantUML, and ASCII.
Does NOT cover:
senior-devops (capacity planning) and senior-qa (performance test harnesses).senior-security or senior-secops for CVE/SAST/DAST.senior-frontend and design-auditor.senior-devops and release-orchestrator.| Skill | Integration | Data Flow | |-------|-------------|-----------| | senior-backend | Architecture patterns inform backend service boundaries and API contract design | Architect assessment output (detected pattern, layer assignments) feeds into backend module scaffolding | | senior-devops | Deployment diagrams and technology detection drive infrastructure-as-code decisions | Deployment diagram type output + detected technologies list consumed by DevOps for Terraform/K8s config | | senior-security | Dependency analysis surfaces packages that need security review | Dependency list JSON (--output json) passed to security scanning for CVE correlation | | senior-fullstack | Architecture pattern selection determines which fullstack scaffold template to use | Pattern selection workflow result (e.g., modular monolith) maps to project_scaffolder.py --type flag | | code-reviewer | Layer violation and god-class findings become review checklist items | project_architect.py --output json issues array integrated into code review checklists | | tech-stack-evaluator | Technology detection results feed tech stack evaluation for upgrade/migration decisions | Detected technologies list and dependency versions inform stack evaluation decision matrices |
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | pass→fail | 14,825 | 25,159 | +70% | 1 | 1 | 0% | 2,265 | 5,724 | +153% | 0 | 0 | — |
case-02 | fail→fail | 27,138 | 27,944 | +3% | 1 | 1 | 0% | 3,990 | 5,716 | +43% | 0 | 0 | — |
case-03 | fail→fail | 19,383 | 25,025 | +29% | 1 | 1 | 0% | 2,940 | 5,150 | +75% | 0 | 0 | — |
case-04 | fail→pass | 9,762 | 7,836 | -20% | 1 | 1 | 0% | 1,649 | 2,734 | +66% | 0 | 0 | — |
case-05 | fail→pass | 6,334 | 6,790 | +7% | 1 | 1 | 0% | 1,114 | 2,502 | +125% | 0 | 0 | — |
case-06 | fail→pass | 6,413 | 9,041 | +41% | 1 | 1 | 0% | 866 | 2,713 | +213% | 0 | 0 | — |
case-07 | fail→pass | 6,385 | 20,660 | +224% | 1 | 1 | 0% | 1,006 | 4,696 | +367% | 0 | 0 | — |
case-08 | fail→pass | 15,539 | 3,947 | -75% | 1 | 1 | 0% | 2,281 | 1,956 | -14% | 0 | 0 | — |
case-09 | fail→pass | 13,049 | 5,168 | -60% | 1 | 1 | 0% | 1,951 | 2,216 | +14% | 0 | 0 | — |
case-10 | fail→pass | 11,265 | 2,555 | -77% | 1 | 1 | 0% | 1,687 | 1,795 | +6% | 0 | 0 | — |
case-11 | fail→pass | 11,765 | 3,452 | -71% | 1 | 1 | 0% | 1,661 | 1,942 | +17% | 0 | 0 | — |
case-12 | fail→pass | 12,589 | 4,520 | -64% | 1 | 1 | 0% | 1,866 | 2,137 | +15% | 0 | 0 | — |
case-13 | fail→pass | 7,840 | 2,227 | -72% | 1 | 1 | 0% | 1,210 | 1,711 | +41% | 0 | 0 | — |
case-14 | fail→pass | 10,316 | 2,396 | -77% | 1 | 1 | 0% | 1,449 | 1,747 | +21% | 0 | 0 | — |
case-15 | fail→pass | 8,586 | 2,176 | -75% | 1 | 1 | 0% | 1,207 | 1,694 | +40% | 0 | 0 | — |
case-16 | fail→pass | 11,248 | 1,760 | -84% | 1 | 1 | 0% | 1,641 | 1,599 | -3% | 0 | 0 | — |
case-17 | pass→pass | 9,219 | 1,949 | -79% | 1 | 1 | 0% | 1,297 | 1,674 | +29% | 0 | 0 | — |
case-18 | fail→pass | 10,767 | 2,181 | -80% | 1 | 1 | 0% | 1,512 | 1,747 | +16% | 0 | 0 | — |
case-19 | pass→pass | 11,674 | 14,703 | +26% | 1 | 1 | 0% | 1,735 | 3,571 | +106% | 0 | 0 | — |
case-20 | pass→pass | 11,754 | 5,719 | -51% | 1 | 1 | 0% | 1,784 | 2,171 | +22% | 0 | 0 | — |
case-21 | fail→pass | 8,142 | 9,857 | +21% | 1 | 1 | 0% | 1,178 | 1,875 | +59% | 0 | 0 | — |
case-22 | fail→pass | 17,104 | 18,012 | +5% | 1 | 1 | 0% | 2,961 | 4,581 | +55% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +68 percentage points is the difference between those two pass rates over the 22 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.