{
  "context": "Tests whether the agent performs a Rails architecture review following the rails-architecture-review skill: covering review areas in the correct order, classifying findings by severity, and producing output with all four required fields per finding.",
  "type": "weighted_checklist",
  "checklist": [
    {
      "name": "Entry points identified first",
      "description": "The review begins by identifying the entry points of the application (routes, controllers, or public API surface) before examining domain logic or models",
      "max_score": 8
    },
    {
      "name": "Domain logic layer checked",
      "description": "The review explicitly examines where domain/business logic lives — whether it's in models, services, callbacks, or controllers",
      "max_score": 8
    },
    {
      "name": "Models and callbacks inspected",
      "description": "The review addresses model associations, callbacks, or ActiveRecord concerns",
      "max_score": 8
    },
    {
      "name": "High: business logic in callbacks",
      "description": "Business logic embedded in ActiveRecord callbacks is classified as High severity (not Medium or Low)",
      "max_score": 10
    },
    {
      "name": "High: controller multi-step workflow",
      "description": "A controller action that performs multiple sequential domain operations is classified as High severity",
      "max_score": 10
    },
    {
      "name": "Boundary problems prioritised",
      "description": "The findings section leads with boundary violations (cross-layer coupling, misplaced logic) rather than style or naming issues",
      "max_score": 8
    },
    {
      "name": "Medium finding identified",
      "description": "At least one finding is correctly classified as Medium severity (e.g. duplicated workflow, heavy scope/class method, concern mixing responsibilities)",
      "max_score": 8
    },
    {
      "name": "Affected files per finding",
      "description": "Each finding names the specific file(s) where the problem exists",
      "max_score": 10
    },
    {
      "name": "Risk described per finding",
      "description": "Each finding includes a description of the risk — what goes wrong if the issue is not addressed",
      "max_score": 10
    },
    {
      "name": "Improvement per finding",
      "description": "Each finding includes a smallest credible improvement — a specific, actionable refactor suggestion rather than a generic principle",
      "max_score": 10
    },
    {
      "name": "Concerns and helpers checked",
      "description": "The review addresses at least one concern, helper, or module mixin — assessing whether it mixes responsibilities",
      "max_score": 10
    }
  ]
}