{
  "schema_version": "2026-05-24",
  "suite_id": "agent-analyzer-guided-v3",
  "source": "scripts/benchmark-repeat.sh aggregate.json",
  "reproducibility": {
    "manifest_sha256": "08d073a336c630aa6dc23fa3b9a2a71db131b9e8d1f8296b9b48cf4b6fd160e0",
    "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": "b96b8a7f5cc57c4335bc7bc85ec726c836ed0996",
    "quality_command": "go test ./..."
  },
  "required_repeats": 3,
  "completed_repeats": 3,
  "quality_passed": true,
  "failed_repeats": [],
  "delta_stats": {
    "avoidable_waste_high_points": {
      "n": 3,
      "mean": -2.3333333333333335,
      "median": -2,
      "min": -3,
      "max": -2,
      "stdev": 0.5773502691896257
    },
    "avoidable_waste_low_points": {
      "n": 3,
      "mean": -2.3333333333333335,
      "median": -2,
      "min": -3,
      "max": -2,
      "stdev": 0.5773502691896257
    },
    "claude_cache_creation_input_tokens": {
      "n": 3,
      "mean": -6560.666666666667,
      "median": -6264,
      "min": -10979,
      "max": -2439,
      "stdev": 4277.72233008798
    },
    "claude_cache_read_input_tokens": {
      "n": 3,
      "mean": -40954.333333333336,
      "median": -53478,
      "min": -85579,
      "max": 16194,
      "stdev": 52029.48733490782
    },
    "claude_output_tokens": {
      "n": 3,
      "mean": -503.6666666666667,
      "median": -618,
      "min": -682,
      "max": -211,
      "stdev": 255.4688500254646
    },
    "claude_total_cost_usd": {
      "n": 3,
      "mean": -0.04421946666666666,
      "median": -0.04631104999999999,
      "min": -0.048578399999999994,
      "max": -0.037768949999999996,
      "stdev": 0.005700183589660365
    },
    "context_growth_events": {
      "n": 3,
      "mean": -1.0,
      "median": -1,
      "min": -1,
      "max": -1,
      "stdev": 0.0
    },
    "estimated_tokens": {
      "n": 3,
      "mean": -12370.333333333334,
      "median": -10840,
      "min": -16581,
      "max": -9690,
      "stdev": 3691.599969299671
    },
    "failed_commands": {
      "n": 3,
      "mean": -0.3333333333333333,
      "median": 0,
      "min": -1,
      "max": 0,
      "stdev": 0.5773502691896257
    },
    "rereads": {
      "n": 3,
      "mean": -0.3333333333333333,
      "median": 0,
      "min": -1,
      "max": 0,
      "stdev": 0.5773502691896257
    },
    "tool_output_tokens": {
      "n": 3,
      "mean": -12697.666666666666,
      "median": -12064,
      "min": -15919,
      "max": -10110,
      "stdev": 2955.8874020052476
    }
  },
  "run_dirs": [
    "run-01",
    "run-02",
    "run-03"
  ],
  "promotion_policy": "recommendation_evidence",
  "candidate_source_url": null,
  "candidate_issue": null,
  "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": 0.2468368,
    "optimized_mean_usd": 0.18762949999999998,
    "delta_mean_usd": -0.05920730000000001,
    "savings_rate": 0.23986415315706577,
    "savings_percent": 23.986415315706576,
    "scaled_examples": {
      "savings_rate": 0.23986415315706577,
      "savings_percent": 23.986415315706576,
      "weekly_100_usd": 23.986415315706576,
      "weekly_500_usd": 119.93207657853289,
      "monthly_2000_usd": 479.72830631413154,
      "monthly_5000_usd": 1199.3207657853288,
      "monthly_10000_usd": 2398.6415315706577,
      "basis": "Scaled linearly from the suite's optimized-vs-baseline published API-rate delta: savings = comparable baseline spend * savings_percent."
    },
    "delta_min_usd": -0.07123680000000004,
    "delta_max_usd": -0.04348469999999999,
    "runs": [
      {
        "run": "run-01",
        "baseline_usd": 0.22676159999999998,
        "optimized_usd": 0.1832769,
        "delta_usd": -0.04348469999999999
      },
      {
        "run": "run-02",
        "baseline_usd": 0.2643324,
        "optimized_usd": 0.19309559999999998,
        "delta_usd": -0.07123680000000004
      },
      {
        "run": "run-03",
        "baseline_usd": 0.24941639999999998,
        "optimized_usd": 0.186516,
        "delta_usd": -0.0629004
      }
    ],
    "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."
}