Stack Options & Evaluation (Orchestrator)

Generate 2-5 coherent stack candidates, evaluate them across multiple dimensions (including local dev experience), and present a comparable decision matrix. Different stacks can be optimize for specific vendor like AWS, GCP, Vercel.

Prerequisites

Requires architecture brief from architecture-refinement skill. If not available, invoke that skill first.

Process Overview

This skill orchestrates the following standalone skills (each can also be invoked directly):

1. candidate-generator - Generate 2-5 stack candidates based on requirements
2. decision-matrix - Build weighted comparison matrix
3. Cost evaluators - Estimate costs per provider:

- cost-aws - AWS infrastructure costs - cost-gcp - GCP infrastructure costs - cost-supabase - Supabase platform costs - cost-firebase - Firebase platform costs - cost-vercel - Vercel platform costs

1. local-dev-evaluator - Assess local development experience
2. offline-impact-evaluator - Validate stack against offline requirement
3. aspire-evaluator - Check .NET Aspire fit (when applicable)
4. custom-explorer - Evaluate user-proposed combinations

All sub-skills are in the stack-evaluation/ directory and can be triggered independently. For example: "Estimate AWS costs for EKS + Aurora" triggers cost-aws directly.

Evaluation Factors

Each candidate stack must be evaluated across:

• Monthly cost (baseline + scale curve)
• Maintainability / ops overhead
• Elasticity
• Reliability & DR posture
• Observability
• Security posture
• Vendor lock-in
• Local Dev DX (mandatory):

- Time-to-first-run - Local dependency orchestration complexity - Prod parity (auth/data/async/config) - Local debugging + dev-time observability

• Offline Capability (when required):

- Offline strategy (cache-only, IndexedDB/SQLite, event queue, CRDTs, background sync) - Data-loss prevention guarantees (none, ephemeral, session-durable, strong) - Consistency model (none, eventual, strong) - Sync complexity (1-5)

Output Contract

stack_evaluation:
  architecture_brief_id: "<reference>"
  candidates:
    - id: "<unique_id>"
      name: "<name>"
      summary: "<one-line>"
      components:
        frontend: { technology: "", hosting: "" }
        backend: { technology: "", hosting: "" }
        database: { technology: "", hosting: "" }
        auth: { provider: "" }
        async: { technology: "" }
        storage: { technology: "" }
      local_dev_story: "<how to run locally>"
      offline_profile:
        strategy: "<cache_only|indexeddb|sqlite|crdt|event_queue|none>"
        data_loss_guarantee: "<none|ephemeral|session_durable|strong>"
        consistency_model: "<none|eventual|strong>"
        sync_complexity: <1-5>
      cost_estimate:
        baseline: "<$/month>"
        at_scale: "<$/month>"
      scores:
        cost: <1-5>
        ops_overhead: <1-5>
        reliability: <1-5>
        scalability: <1-5>
        security: <1-5>
        vendor_lock_in: <1-5>
        local_dev_dx: <1-5>
        offline_capability: <1-5>
  decision_matrix:
    criteria_weights:
      cost: <1-5>
      ops_overhead: <1-5>
      reliability: <1-5>
      scalability: <1-5>
      security: <1-5>
      vendor_lock_in: <1-5>
      local_dev_dx: <1-5>
      offline_capability: <1-5>
    ranked_options:
      - candidate_id: "<id>"
        weighted_score: <score>
        rank: 1
  recommendation:
    top_choice: "<candidate_id>"
    rationale: "<why>"
    tradeoffs: "<key tradeoffs>"

Presentation

Present as comparison table:

Next Step

After user reviews candidates, proceed to stack-selection skill to confirm the chosen stack.