{
  "schema_version": "2026-05-24",
  "suite_id": "gathon-claude",
  "source": "scripts/benchmark-repeat.sh aggregate.json",
  "reproducibility": {
    "manifest_sha256": "7350fd19ea2980c9fbbc5c31358c8228b3d1810f2f1a8c886bd5101482a40380",
    "manifest_schema_version": "2026-06-04",
    "harness": "claude",
    "required_repeats": 3,
    "completed_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.",
    "analyzer_commit": "e28e71a965aeef22fa1f68542af2d516e4f126aa",
    "analyzer_dirty": true,
    "source_commit": "1116b800a71cc532b37422f9b56b4391b3cf81f9",
    "source_dirty": true,
    "base_ref": "benchmark/owner-breakdown-v1-base",
    "quality_command": "go test ./...",
    "suite_file_sha256": "d52a79404542508e2ecc2b118277bad22e46df86d95069bede5e8f4f9dbd599b",
    "task_prompt_sha256": "abef759344047294d3b4a1b279a7a5fb72731764406c3d879efbfc6100d8ddc5",
    "guidance_sha256": "d36e664e923df7916ac78753a5d8f9e9178de76595c967797a2e9474a44f1431",
    "mcp_config_sha256": "902fd11d303d7c4b5bf8e6cd1fee743aff06a36772dda736c7621e1b0775ada6",
    "tool_versions": {
      "git": "git version 2.52.0",
      "go": "go version go1.25.4 darwin/arm64",
      "python3": "Python 3.14.5",
      "node": "v25.2.1",
      "npm": "11.6.2",
      "npx": "11.6.2",
      "claude": "2.1.165 (Claude Code)",
      "codex": "codex-cli 0.134.0",
      "uvx": "uv-tool-uvx 0.5.13 (c456bae5e 2024-12-27)",
      "rtk": "rtk 0.41.0",
      "grepai": "available"
    }
  },
  "required_repeats": 3,
  "completed_repeats": 3,
  "quality_passed": true,
  "failed_repeats": [],
  "delta_stats": {
    "avoidable_waste_high_points": {
      "n": 3,
      "mean": -0.6666666666666666,
      "median": -1,
      "min": -1,
      "max": 0,
      "stdev": 0.5773502691896257
    },
    "avoidable_waste_low_points": {
      "n": 3,
      "mean": -0.6666666666666666,
      "median": -1,
      "min": -1,
      "max": 0,
      "stdev": 0.5773502691896257
    },
    "claude_cache_creation_input_tokens": {
      "n": 3,
      "mean": 5171.333333333333,
      "median": 5235,
      "min": 4198,
      "max": 6081,
      "stdev": 943.1131073913316
    },
    "claude_cache_read_input_tokens": {
      "n": 3,
      "mean": 246904.0,
      "median": 219108,
      "min": 208047,
      "max": 313557,
      "stdev": 57987.526563908556
    },
    "claude_output_tokens": {
      "n": 3,
      "mean": 2071.6666666666665,
      "median": 2173,
      "min": 1807,
      "max": 2235,
      "stdev": 231.29490554989172
    },
    "claude_total_cost_usd": {
      "n": 3,
      "mean": 0.12492836666666664,
      "median": 0.11834864999999997,
      "min": 0.10565159999999998,
      "max": 0.15078484999999997,
      "stdev": 0.023274923575617452
    },
    "context_growth_events": {
      "n": 3,
      "mean": 0.0,
      "median": 0,
      "min": 0,
      "max": 0,
      "stdev": 0.0
    },
    "estimated_tokens": {
      "n": 3,
      "mean": 9471.333333333334,
      "median": 8531,
      "min": 7662,
      "max": 12221,
      "stdev": 2420.597102644993
    },
    "failed_commands": {
      "n": 3,
      "mean": 1.6666666666666667,
      "median": 2,
      "min": 1,
      "max": 2,
      "stdev": 0.5773502691896257
    },
    "rereads": {
      "n": 3,
      "mean": 0.6666666666666666,
      "median": 0,
      "min": 0,
      "max": 2,
      "stdev": 1.1547005383792515
    },
    "tool_output_tokens": {
      "n": 3,
      "mean": 5042.0,
      "median": 4103,
      "min": 3891,
      "max": 7132,
      "stdev": 1813.0943163553295
    }
  },
  "run_dirs": [
    "run-01",
    "run-02",
    "run-03"
  ],
  "promotion_policy": "candidate_until_reviewed",
  "candidate_source_url": "https://github.com/pauldx/gathon",
  "candidate_issue": "https://github.com/Priivacy-ai/agent-log-analyzer/issues/175",
  "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": 0.1686022,
    "optimized_mean_usd": 0.3048184,
    "delta_mean_usd": 0.13621619999999998,
    "savings_rate": -0.8079147247188944,
    "savings_percent": -80.79147247188943,
    "scaled_examples": {
      "savings_rate": -0.8079147247188944,
      "savings_percent": -80.79147247188943,
      "weekly_100_usd": -80.79147247188943,
      "weekly_500_usd": -403.9573623594472,
      "monthly_2000_usd": -1615.8294494377888,
      "monthly_5000_usd": -4039.573623594472,
      "monthly_10000_usd": -8079.147247188944,
      "basis": "Scaled linearly from the suite's optimized-vs-baseline published API-rate delta: savings = comparable baseline spend * savings_percent."
    },
    "delta_min_usd": 0.11474610000000002,
    "delta_max_usd": 0.16412609999999997,
    "runs": [
      {
        "run": "run-01",
        "baseline_usd": 0.18047760000000002,
        "optimized_usd": 0.310254,
        "delta_usd": 0.12977639999999996
      },
      {
        "run": "run-02",
        "baseline_usd": 0.1661184,
        "optimized_usd": 0.2808645,
        "delta_usd": 0.11474610000000002
      },
      {
        "run": "run-03",
        "baseline_usd": 0.1592106,
        "optimized_usd": 0.3233367,
        "delta_usd": 0.16412609999999997
      }
    ],
    "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."
}