{
  "id": "ops-default",
  "version": "0.1.0",
  "description": "Home/personal infrastructure operations: network diagnosis, device configuration, recurring-issue remediation. Command-based hard checks. Reads/writes OpenClaw memory for recurring state.",
  "domain": "ops",

  "assumedModel": {
    "minTier": "standard",
    "capabilities": ["tool-use", "code"]
  },

  "strategy": {
    "kind": "planner-worker-evaluator",
    "fixerEnabled": false
  },

  "evaluator": {
    "intensity": "deep",
    "fewShotCalibrationPath": "data/few-shot-rubrics/ops.md",
    "checklist": {
      "hard": [
        {
          "id": "H1",
          "title": "Diagnosis grounded",
          "passDescription": "every claim about device/network state is backed by a command hard check (ping, traceroute, ip, iw, log tail, …). No claims-without-evidence.",
          "failReworkDescription": "some claims lack command evidence; running the right diagnostic commands would produce it",
          "failEscalateDescription": "the device or network is genuinely unreachable or the issue requires physical access / hardware replacement"
        },
        {
          "id": "H2",
          "title": "Remediation reversible",
          "passDescription": "for every change command produced, a corresponding revert command is documented, or one-way changes are called out explicitly",
          "failReworkDescription": "revert commands are missing for some changes but can be derived from the runbook",
          "failEscalateDescription": "changes were applied destructively with no feasible revert path (e.g., overwritten config without backup)"
        },
        {
          "id": "H3",
          "title": "Memory write",
          "passDescription": "the terminal stage appends a structured entry to OpenClaw active memory with: symptom, diagnosis, remediation, devices touched",
          "failReworkDescription": "memory entry is partially written or missing some fields; can be completed in rework",
          "failEscalateDescription": "the issue could not be resolved, so a memory entry would be misleading or premature"
        }
      ],
      "soft": [
        { "id": "S1", "title": "Diagnostic completeness", "weight": 0.4, "scoringGuide": "5 = systematic elimination of hypotheses with evidence; 1 = guesswork without diagnostics" },
        { "id": "S2", "title": "Risk awareness", "weight": 0.3, "scoringGuide": "5 = every change evaluated for blast radius with rollback plan; 1 = changes applied blindly" },
        { "id": "S3", "title": "Memory entry quality", "weight": 0.2, "scoringGuide": "5 = future runs can reproduce diagnosis and remediation from entry alone; 1 = vague or missing" },
        { "id": "S4", "title": "Clarity", "weight": 0.1, "scoringGuide": "5 = clear runbook any admin can follow; 1 = disorganized or incomplete" }
      ]
    },
    "rubric": {
      "criteria": [
        {
          "id": "diagnostic-completeness",
          "weight": 0.4,
          "description": "Systematic elimination of hypotheses with command evidence.",
          "scoringGuide": "5 = every claim backed by command output; 1 = claims without evidence."
        },
        {
          "id": "risk-awareness",
          "weight": 0.3,
          "description": "Every change evaluated for blast radius with rollback plan.",
          "scoringGuide": "5 = all changes have revert documented; 1 = destructive without backup."
        },
        {
          "id": "memory-entry-quality",
          "weight": 0.2,
          "description": "Structured memory entry for future recurrence lookups.",
          "scoringGuide": "5 = future runs can reproduce fix from entry; 1 = vague or missing."
        },
        {
          "id": "clarity",
          "weight": 0.1,
          "description": "Clear runbook any admin can follow.",
          "scoringGuide": "5 = step-by-step with expected outputs; 1 = disorganized."
        }
      ]
    },
    "hardThresholds": [
      { "criterionId": "diagnostic-completeness", "min": 3 }
    ]
  },

  "recovery": {
    "resumeStrategy": "auto",
    "maxReworkPerStage": 3,
    "contextResetThresholdRatio": 0.7,
    "stuckDetection": {
      "noProgressReworkLimit": 2
    }
  },

  "budget": {
    "maxTokens": 800000,
    "maxWallClockMinutes": 240,
    "perStageTokenCap": 200000
  },

  "benchGoal": {
    "goal": "Diagnose why the home WiFi drops every evening around 9pm and produce a runbook with revert commands.",
    "expectedArtifacts": ["diagnosis.md", "runbook.md"],
    "maxTokens": 50000,
    "maxWallClockMs": 120000
  }
}