{
  "schema_version": "2026-05-24",
  "suite_id": "codegraph-claude",
  "source": "scripts/benchmark-repeat.sh aggregate.json",
  "reproducibility": {
    "manifest_sha256": "e291f95a5f13d752343d6e967fb45145c952e6bb4c82289556c4f2eb36683d8f",
    "manifest_schema_version": "2026-05-24",
    "harness": "claude",
    "required_repeats": 3,
    "fresh_context_contract": "Each repeat invokes one full baseline/optimized harness run. The harness creates fresh detached git worktrees and launches separate Claude/Codex task sessions for baseline and optimized runs.",
    "base_ref": "benchmark/owner-breakdown-v1-base",
    "quality_command": "go test ./..."
  },
  "required_repeats": 3,
  "completed_repeats": 3,
  "quality_passed": true,
  "failed_repeats": [],
  "delta_stats": {
    "avoidable_waste_high_points": {
      "n": 3,
      "mean": 1.0,
      "median": 1,
      "min": 0,
      "max": 2,
      "stdev": 1.0
    },
    "avoidable_waste_low_points": {
      "n": 3,
      "mean": 0.6666666666666666,
      "median": 1,
      "min": 0,
      "max": 1,
      "stdev": 0.5773502691896257
    },
    "claude_cache_creation_input_tokens": {
      "n": 3,
      "mean": 6608.666666666667,
      "median": 5805,
      "min": 5500,
      "max": 8521,
      "stdev": 1663.1356930008246
    },
    "claude_cache_read_input_tokens": {
      "n": 3,
      "mean": 164611.66666666666,
      "median": 173494,
      "min": 119105,
      "max": 201236,
      "stdev": 41779.745982154236
    },
    "claude_output_tokens": {
      "n": 3,
      "mean": 450.3333333333333,
      "median": 284,
      "min": 213,
      "max": 854,
      "stdev": 351.38345626015655
    },
    "claude_total_cost_usd": {
      "n": 3,
      "mean": 0.08125033333333334,
      "median": 0.07841195000000001,
      "min": 0.05987449999999997,
      "max": 0.10546455000000005,
      "stdev": 0.0229271777528512
    },
    "context_growth_events": {
      "n": 3,
      "mean": 0.0,
      "median": 0,
      "min": 0,
      "max": 0,
      "stdev": 0.0
    },
    "estimated_tokens": {
      "n": 3,
      "mean": 6093.666666666667,
      "median": 6435,
      "min": 4317,
      "max": 7529,
      "stdev": 1632.9780565988428
    },
    "failed_commands": {
      "n": 3,
      "mean": 2.6666666666666665,
      "median": 3,
      "min": 2,
      "max": 3,
      "stdev": 0.5773502691896257
    },
    "rereads": {
      "n": 3,
      "mean": 1.0,
      "median": 1,
      "min": 0,
      "max": 2,
      "stdev": 1.0
    },
    "tool_output_tokens": {
      "n": 3,
      "mean": 4046.3333333333335,
      "median": 3992,
      "min": 3111,
      "max": 5036,
      "stdev": 963.6494867602707
    }
  },
  "run_dirs": [
    "run-01",
    "run-02",
    "run-03"
  ],
  "promotion_policy": "candidate_until_reviewed",
  "candidate_source_url": "https://github.com/colbymchenry/codegraph",
  "candidate_issue": "https://github.com/Priivacy-ai/agent-log-analyzer/issues/173",
  "published_api_cost_estimate": {
    "harness": "claude",
    "complete_cost_surface": true,
    "multi_session_runs_not_in_stdout_usage": [],
    "rates": {
      "model_assumption": "Claude Sonnet 4.6",
      "input_tokens_per_mtok_usd": 3.0,
      "cache_creation_input_tokens_per_mtok_usd": 6.0,
      "cache_read_input_tokens_per_mtok_usd": 0.3,
      "output_tokens_per_mtok_usd": 15.0
    },
    "baseline_mean_usd": 0.17649499999999999,
    "optimized_mean_usd": 0.2723205,
    "delta_mean_usd": 0.09582550000000002,
    "savings_rate": -0.5429360605116293,
    "savings_percent": -54.29360605116293,
    "scaled_examples": {
      "savings_rate": -0.5429360605116293,
      "savings_percent": -54.29360605116293,
      "weekly_100_usd": -54.29360605116293,
      "weekly_500_usd": -271.46803025581465,
      "monthly_2000_usd": -1085.8721210232586,
      "monthly_5000_usd": -2714.6803025581466,
      "monthly_10000_usd": -5429.360605116293,
      "basis": "Scaled linearly from the suite's optimized-vs-baseline published API-rate delta: savings = comparable baseline spend * savings_percent."
    },
    "delta_min_usd": 0.07195350000000003,
    "delta_max_usd": 0.1243458,
    "runs": [
      {
        "run": "run-01",
        "baseline_usd": 0.178113,
        "optimized_usd": 0.26929020000000004,
        "delta_usd": 0.09117720000000004
      },
      {
        "run": "run-02",
        "baseline_usd": 0.178806,
        "optimized_usd": 0.2507595,
        "delta_usd": 0.07195350000000003
      },
      {
        "run": "run-03",
        "baseline_usd": 0.172566,
        "optimized_usd": 0.2969118,
        "delta_usd": 0.1243458
      }
    ],
    "note": "Uses exposed Claude Code aggregate usage fields repriced at published Claude Sonnet 4.6 API rates. Claude Code native billing can differ because raw runs also report internal model usage."
  },
  "privacy_boundary": "Raw Claude/Codex logs, raw prompts, local paths, and secrets are not included in public aggregate artifacts."
}