{
  "context": "Tests whether the agent correctly classifies non-output-affecting warnings as cosmetic tier, applies the correct retry limit of 2 (not 3, not unlimited), and logs after retries are exhausted.",
  "type": "weighted_checklist",
  "checklist": [
    {
      "name": "Cosmetic tier label",
      "description": "Code or comments label this type of warning as 'cosmetic' (or equivalent lowest-severity tier name)",
      "max_score": 12
    },
    {
      "name": "Retry limit of 2",
      "description": "Retry logic allows at most 2 retry attempts (not 3, not unlimited) before giving up on a cosmetic failure",
      "max_score": 15
    },
    {
      "name": "Log after retries exhausted",
      "description": "After retry attempts are exhausted, code logs the warning (writes to log, prints warning, or calls a log function) and continues rather than escalating or raising",
      "max_score": 12
    },
    {
      "name": "No escalation for cosmetic",
      "description": "Code does NOT escalate (halt processing, raise exception, or trigger on-call alert) for cosmetic failures after retries are exhausted",
      "max_score": 10
    },
    {
      "name": "Tier-based branching",
      "description": "Code contains branching logic (if/elif/switch or dispatch) that handles different severity levels differently, not just a single uniform retry policy",
      "max_score": 10
    },
    {
      "name": "Report structure present",
      "description": "Module includes a report/result structure (dataclass, namedtuple, dict) capturing the failure signal and action taken",
      "max_score": 8
    },
    {
      "name": "Report includes tier",
      "description": "Report structure includes a field representing the tier or severity level assigned",
      "max_score": 8
    },
    {
      "name": "POLICY.md present",
      "description": "POLICY.md file exists and describes the retry/log strategy",
      "max_score": 10
    },
    {
      "name": "Condition: output still correct",
      "description": "Code or comments capture the condition for cosmetic classification — that output remains correct/complete despite the warning",
      "max_score": 8
    },
    {
      "name": "No suppression budget for cosmetic",
      "description": "Code does NOT implement a suppression budget (time window tracking, recurrence store) for cosmetic failures — suppression budget is reserved for operational tier",
      "max_score": 7
    }
  ]
}