{
  "context": "Evaluates how well the solution leverages Luigi to dynamically expand manifest-driven hourly tasks and execute them in batched runs before aggregating outputs. Scoring looks only at correct use of Luigi primitives to express dynamic dependencies, batching, idempotent targets, and pipeline execution.",
  "type": "weighted_checklist",
  "checklist": [
    {
      "name": "Dynamic yield",
      "description": "Daily orchestration task reads the manifest and spawns per-hour work by yielding other luigi.Task instances (or using DynamicRequirements) from run() rather than hard-coding requires().",
      "max_score": 25
    },
    {
      "name": "Batch parameter",
      "description": "Hourly cleaning task defines its hour/date parameter with a batch_method on luigi.Parameter (or Date/IntParameter variants) so the scheduler can coalesce multiple hours into one run.",
      "max_score": 25
    },
    {
      "name": "Batch sizing",
      "description": "Hourly task exposes max_batch_size (or registers a task batcher) tied to the max_batch_size argument of build_and_run so grouped executions honor the configured limit.",
      "max_score": 15
    },
    {
      "name": "Target completeness",
      "description": "Outputs use luigi.Target implementations (e.g., LocalTarget) and rely on Target.exists/Task.complete to skip rewriting preexisting hourly files while still feeding downstream aggregation.",
      "max_score": 15
    },
    {
      "name": "Pipeline run",
      "description": "build_and_run drives the graph via luigi.build/run, wiring input()/output() across tasks so aggregation waits on dynamically discovered hour tasks and returns a boolean success indicator.",
      "max_score": 20
    }
  ]
}