solutions-architecture-agent

Evaluates one Well-Architected Framework pillar independently. Use when /architecture or /review needs parallel WA assessment across 6 pillars.

Analyzes one STRIDE threat category independently. Use when /security-review needs parallel threat analysis across 6 categories.

Design system architecture with technology stack selection, component design, Mermaid diagrams, and Well-Architected scoring. Produces dual-audience output for technical builders and business leaders. Use after requirements are complete, whether for enterprise platforms, startup MVPs, or migration targets.

Design data models including entity-relationship schemas, vector database schemas, knowledge graphs, ontologies, and data governance policies. Use after architecture defines the data layer.

Generate cost estimates, LOE breakdowns, team composition, and infrastructure cost modeling with confidence scoring. Supports bottom-up, T-shirt, and three-point methods. Use after architecture is complete to inform budgeting, fundraising, or SOW pricing.

Plan system integrations: API contracts, data flow mappings, migration strategies (strangler fig, parallel run), legacy bridging patterns, and CI/CD pipeline architecture. Use for migration engagements or systems with complex integrations.

Generate phased delivery roadmaps with sprint plans, milestones, dependency graphs, decision gates, and risk registers. Supports both greenfield and migration timelines. Use after estimation is complete.

Assemble proposals and SOWs from knowledge base data. Supports 4 types: discovery, implementation, internal, pitch deck. Reads all KB files, never re-analyzes — assembly only. Use after all required skills have completed.

Run progressive requirements discovery workshops (quick/standard/comprehensive). Captures client context, AI suitability, functional/non-functional requirements, and success criteria. Use when starting a new engagement, qualifying a prospect, validating a technical approach, or extracting requirements from meeting notes.

Review any SA deliverable using LLM-as-judge methodology with 3 iteration passes. Scores on 5 dimensions (completeness, technical soundness, well-architected, clarity, feasibility). Applies dual-persona validation. Use after any skill produces output.

Perform comprehensive security and privacy review: STRIDE threat modeling, compliance mapping (HIPAA/SOC2/CCPA/GLBA/PCI-DSS), defense-in-depth architecture, and AI-specific security controls. Use after architecture is complete.