{
  "schema": "bull-benchmark-v1",
  "source_commit": "cd461ae05a0925d6bd32381f8ef4da9e99bb2ccb",
  "run": {
    "timestamp_utc": "20260920T170536Z",
    "commit": "cd461ae05a0925d6bd32381f8ef4da9e99bb2ccb",
    "kernel": "Linux 7.2.6-1-cachyos x86_64 GNU/Linux",
    "python": "Python 3.14.7",
    "repeats": 3,
    "hot_iters": 500,
    "cpu_model": "AMD Ryzen 5 7430U with Radeon Graphics",
    "cpu_count": 12,
    "memory": "11Gi"
  },
  "selected_redteam": {
    "runs": 33,
    "passes": 33,
    "failures": 0,
    "attacks": [
      {
        "attack_id": "path_traversal",
        "attack": "Path traversal encodings",
        "surface": "filesystem",
        "runs": 3,
        "passed": 3,
        "failed": 0,
        "pass_rate_pct": 100,
        "mean_ms": 338.406,
        "median_ms": 328.454,
        "min_ms": 310.374,
        "max_ms": 376.391,
        "peak_rss_mb": 45.5
      },
      {
        "attack_id": "unbound_command",
        "attack": "Read authority → shell execution",
        "surface": "authorization",
        "runs": 3,
        "passed": 3,
        "failed": 0,
        "pass_rate_pct": 100,
        "mean_ms": 366.427,
        "median_ms": 360.8,
        "min_ms": 360.342,
        "max_ms": 378.14,
        "peak_rss_mb": 44.883
      },
      {
        "attack_id": "argv_drift",
        "attack": "Authorized argv → force-push drift",
        "surface": "authorization",
        "runs": 3,
        "passed": 3,
        "failed": 0,
        "pass_rate_pct": 100,
        "mean_ms": 318.111,
        "median_ms": 326.729,
        "min_ms": 298.313,
        "max_ms": 329.291,
        "peak_rss_mb": 44.926
      },
      {
        "attack_id": "credential_from_external",
        "attack": "External content → credential read",
        "surface": "secrets",
        "runs": 3,
        "passed": 3,
        "failed": 0,
        "pass_rate_pct": 100,
        "mean_ms": 301.671,
        "median_ms": 299.895,
        "min_ms": 295.504,
        "max_ms": 309.615,
        "peak_rss_mb": 44.855
      },
      {
        "attack_id": "sandbox_egress_bypass",
        "attack": "SANDBOX verdict → egress broker",
        "surface": "egress",
        "runs": 3,
        "passed": 3,
        "failed": 0,
        "pass_rate_pct": 100,
        "mean_ms": 292.603,
        "median_ms": 292.307,
        "min_ms": 291.569,
        "max_ms": 293.934,
        "peak_rss_mb": 44.734
      },
      {
        "attack_id": "sandbox_secret_bypass",
        "attack": "SANDBOX verdict → secret broker",
        "surface": "secrets",
        "runs": 3,
        "passed": 3,
        "failed": 0,
        "pass_rate_pct": 100,
        "mean_ms": 304.441,
        "median_ms": 308.532,
        "min_ms": 295.086,
        "max_ms": 309.706,
        "peak_rss_mb": 44.668
      },
      {
        "attack_id": "ssrf_private_addresses",
        "attack": "Private/metadata SSRF targets",
        "surface": "egress",
        "runs": 3,
        "passed": 3,
        "failed": 0,
        "pass_rate_pct": 100,
        "mean_ms": 355.926,
        "median_ms": 381.879,
        "min_ms": 301.842,
        "max_ms": 384.057,
        "peak_rss_mb": 44.797
      },
      {
        "attack_id": "wrong_peer_uid",
        "attack": "Wrong Unix peer UID",
        "surface": "secrets",
        "runs": 3,
        "passed": 3,
        "failed": 0,
        "pass_rate_pct": 100,
        "mean_ms": 360.088,
        "median_ms": 365.086,
        "min_ms": 349.124,
        "max_ms": 366.053,
        "peak_rss_mb": 44.809
      },
      {
        "attack_id": "prompt_injection",
        "attack": "Prompt injection",
        "surface": "multi-agent",
        "runs": 3,
        "passed": 3,
        "failed": 0,
        "pass_rate_pct": 100,
        "mean_ms": 351.591,
        "median_ms": 348.819,
        "min_ms": 345.773,
        "max_ms": 360.182,
        "peak_rss_mb": 46.086
      },
      {
        "attack_id": "forged_policy_verdict",
        "attack": "Forged policy verdict",
        "surface": "multi-agent",
        "runs": 3,
        "passed": 3,
        "failed": 0,
        "pass_rate_pct": 100,
        "mean_ms": 367.611,
        "median_ms": 384.904,
        "min_ms": 331.814,
        "max_ms": 386.116,
        "peak_rss_mb": 45.012
      },
      {
        "attack_id": "approval_replay",
        "attack": "Approval-token replay",
        "surface": "multi-agent",
        "runs": 3,
        "passed": 3,
        "failed": 0,
        "pass_rate_pct": 100,
        "mean_ms": 376.64,
        "median_ms": 378.265,
        "min_ms": 369.628,
        "max_ms": 382.027,
        "peak_rss_mb": 44.992
      }
    ]
  },
  "full_regression_suite": {
    "status": "PASS",
    "return_code": 0,
    "duration_seconds": 4.587
  },
  "hotpath": [
    {
      "benchmark": "Traversal rejection",
      "iterations": 500,
      "expected_matches": 500,
      "elapsed_seconds": 0.002449,
      "mean_us": 4.897,
      "ops_per_second": 204187.31
    },
    {
      "benchmark": "Prompt-injection denial",
      "iterations": 500,
      "expected_matches": 500,
      "elapsed_seconds": 33.279928,
      "mean_us": 66559.856,
      "ops_per_second": 15.02
    },
    {
      "benchmark": "Benign multi-agent allow",
      "iterations": 500,
      "expected_matches": 500,
      "elapsed_seconds": 219.279526,
      "mean_us": 438559.052,
      "ops_per_second": 2.28
    }
  ],
  "caveats": [
    "Selected red-team latency is end-to-end pytest latency, not pure policy-decision latency.",
    "Prompt-injection and benign multi-agent hot-path measurements reuse one MultiAgentSystem and include retained trace/honeytoken state growth.",
    "A PASS is evidence for the defined regression corpus, not independent security certification."
  ]
}
