{
  "context": "Evaluates how the solution uses pylibmc's cache invalidation APIs—single deletion, bulk deletion, TTL refresh, and full flush—to meet the spec requirements without reimplementing cache mechanics.",
  "type": "weighted_checklist",
  "checklist": [
    {
      "name": "Single delete",
      "description": "Per-key removal relies on `pylibmc.Client.delete` and returns its boolean result (handling `pylibmc.NotFound` as a miss) instead of mutating the cache through reads/writes.",
      "max_score": 25
    },
    {
      "name": "Bulk delete",
      "description": "Batch removal is implemented with `Client.delete_multi` (using `key_prefix` when applicable) rather than looping individual deletes, and the per-key status mapping reflects its success/failure outputs.",
      "max_score": 25
    },
    {
      "name": "TTL refresh",
      "description": "Expiration refresh uses `Client.touch` with the provided TTL to extend existing entries and returns its success indicator without overwriting values via set/get flows.",
      "max_score": 20
    },
    {
      "name": "Flush all",
      "description": "Whole-cache clearing calls `Client.flush_all()` directly and does not attempt to enumerate keys or simulate a flush through repeated deletes.",
      "max_score": 15
    },
    {
      "name": "Miss handling",
      "description": "Missing-key scenarios rely on pylibmc's miss signals (`NotFound` or False from delete/touch/delete_multi) rather than speculative reads to determine absence.",
      "max_score": 15
    }
  ]
}