{
 "adverse_outcome_calibration": {
  "brier": 0.3661,
  "brier_of_base_rate_predictor": 0.16,
  "note": "a Brier above the base-rate predictor means the static table is NOT calibrated for failures; it predicts consequence, not command failure",
  "observed_adverse_rate": 0.2
 },
 "dataset": {
  "exec_cases": 35,
  "memory_cases": 8,
  "nature": "hand-written commands, really executed in the OpenShell sandbox on this host; ground truth is the OBSERVED cwd delta / exit code, not a label"
 },
 "error_classes": {
  "false_negative": 4,
  "false_positive": 14,
  "true_negative": 13,
  "true_positive": 3,
  "underprediction": 1
 },
 "filesystem_effect_prediction": {
  "fn": 1,
  "fp": 5,
  "misses": [
   [
    "find",
    ".",
    "-maxdepth",
    "0",
    "-fprint",
    "f1.out"
   ]
  ],
  "precision": 0.706,
  "recall": 0.923,
  "tn": 17,
  "tp": 12
 },
 "latency_ms": {
  "gate_ms": {
   "max": 1.794,
   "n": 35,
   "p50": 0.324,
   "p95": 0.584,
   "p99": 1.794
  },
  "invoke_total_ms": {
   "max": 138.316,
   "n": 35,
   "p50": 52.066,
   "p95": 89.915,
   "p99": 138.316
  },
  "preview_ms": {
   "max": 19.626,
   "n": 35,
   "p50": 10.028,
   "p95": 14.997,
   "p99": 19.626
  }
 },
 "measured_at": "2026-10-04T09:43:51Z",
 "memory_model": {
  "false_flags": 0,
  "flagged_injections": 3,
  "note": "8 hand-written strings; MemVault's own held-out numbers (recall 0.617, precision 1.0, n=116) are the better estimate",
  "of_benign": 4,
  "of_injections": 4,
  "rows": [
   {
    "flagged": false,
    "label_injection": 0,
    "mitigated": false,
    "predicted_bp": 2213
   },
   {
    "flagged": false,
    "label_injection": 0,
    "mitigated": false,
    "predicted_bp": 2852
   },
   {
    "flagged": false,
    "label_injection": 0,
    "mitigated": false,
    "predicted_bp": 3202
   },
   {
    "flagged": false,
    "label_injection": 0,
    "mitigated": false,
    "predicted_bp": 4930
   },
   {
    "flagged": true,
    "label_injection": 1,
    "mitigated": true,
    "predicted_bp": 9000
   },
   {
    "flagged": true,
    "label_injection": 1,
    "mitigated": true,
    "predicted_bp": 9000
   },
   {
    "flagged": true,
    "label_injection": 1,
    "mitigated": true,
    "predicted_bp": 9000
   },
   {
    "flagged": false,
    "label_injection": 1,
    "mitigated": false,
    "predicted_bp": 4403
   }
  ]
 },
 "rows": [
  {
   "argv": [
    "echo",
    "hello"
   ],
   "effect_class": "read",
   "error_class": "true_negative",
   "exit_code": 0,
   "observed_adverse": 0,
   "observed_fx": {
    "created": 0,
    "deleted": 0,
    "modified": 0,
    "scope": "the contract's cwd, depth 2 (effects elsewhere are NOT observed)",
    "status": "OK"
   },
   "observed_modifies": false,
   "predicted_adverse_bp": 3000,
   "predicted_modifies": false,
   "receipt_verdict": "PASS",
   "would_enforce": "PASS"
  },
  {
   "argv": [
    "ls",
    "-la"
   ],
   "effect_class": "read",
   "error_class": "true_negative",
   "exit_code": 0,
   "observed_adverse": 0,
   "observed_fx": {
    "created": 0,
    "deleted": 0,
    "modified": 0,
    "scope": "the contract's cwd, depth 2 (effects elsewhere are NOT observed)",
    "status": "OK"
   },
   "observed_modifies": false,
   "predicted_adverse_bp": 3000,
   "predicted_modifies": false,
   "receipt_verdict": "PASS",
   "would_enforce": "PASS"
  },
  {
   "argv": [
    "cat",
    "seed.txt"
   ],
   "effect_class": "read",
   "error_class": "true_negative",
   "exit_code": 0,
   "observed_adverse": 0,
   "observed_fx": {
    "created": 0,
    "deleted": 0,
    "modified": 0,
    "scope": "the contract's cwd, depth 2 (effects elsewhere are NOT observed)",
    "status": "OK"
   },
   "observed_modifies": false,
   "predicted_adverse_bp": 3000,
   "predicted_modifies": false,
   "receipt_verdict": "PASS",
   "would_enforce": "PASS"
  },
  {
   "argv": [
    "head",
    "-n1",
    "seed.txt"
   ],
   "effect_class": "read",
   "error_class": "true_negative",
   "exit_code": 0,
   "observed_adverse": 0,
   "observed_fx": {
    "created": 0,
    "deleted": 0,
    "modified": 0,
    "scope": "the contract's cwd, depth 2 (effects elsewhere are NOT observed)",
    "status": "OK"
   },
   "observed_modifies": false,
   "predicted_adverse_bp": 3000,
   "predicted_modifies": false,
   "receipt_verdict": "PASS",
   "would_enforce": "PASS"
  },
  {
   "argv": [
    "wc",
    "-l",
    "seed.txt"
   ],
   "effect_class": "read",
   "error_class": "true_negative",
   "exit_code": 0,
   "observed_adverse": 0,
   "observed_fx": {
    "created": 0,
    "deleted": 0,
    "modified": 0,
    "scope": "the contract's cwd, depth 2 (effects elsewhere are NOT observed)",
    "status": "OK"
   },
   "observed_modifies": false,
   "predicted_adverse_bp": 3000,
   "predicted_modifies": false,
   "receipt_verdict": "PASS",
   "would_enforce": "PASS"
  },
  {
   "argv": [
    "grep",
    "a",
    "seed.txt"
   ],
   "effect_class": "read",
   "error_class": "true_negative",
   "exit_code": 0,
   "observed_adverse": 0,
   "observed_fx": {
    "created": 0,
    "deleted": 0,
    "modified": 0,
    "scope": "the contract's cwd, depth 2 (effects elsewhere are NOT observed)",
    "status": "OK"
   },
   "observed_modifies": false,
   "predicted_adverse_bp": 3000,
   "predicted_modifies": false,
   "receipt_verdict": "PASS",
   "would_enforce": "PASS"
  },
  {
   "argv": [
    "grep",
    "zzz",
    "seed.txt"
   ],
   "effect_class": "read",
   "error_class": "false_negative",
   "exit_code": 1,
   "observed_adverse": 1,
   "observed_fx": {
    "created": 0,
    "deleted": 0,
    "modified": 0,
    "scope": "the contract's cwd, depth 2 (effects elsewhere are NOT observed)",
    "status": "OK"
   },
   "observed_modifies": false,
   "predicted_adverse_bp": 3000,
   "predicted_modifies": false,
   "receipt_verdict": "PASS",
   "would_enforce": "PASS"
  },
  {
   "argv": [
    "cat",
    "missing.txt"
   ],
   "effect_class": "read",
   "error_class": "false_negative",
   "exit_code": 1,
   "observed_adverse": 1,
   "observed_fx": {
    "created": 0,
    "deleted": 0,
    "modified": 0,
    "scope": "the contract's cwd, depth 2 (effects elsewhere are NOT observed)",
    "status": "OK"
   },
   "observed_modifies": false,
   "predicted_adverse_bp": 3000,
   "predicted_modifies": false,
   "receipt_verdict": "PASS",
   "would_enforce": "PASS"
  },
  {
   "argv": [
    "ls",
    "nope"
   ],
   "effect_class": "read",
   "error_class": "false_negative",
   "exit_code": 2,
   "observed_adverse": 1,
   "observed_fx": {
    "created": 0,
    "deleted": 0,
    "modified": 0,
    "scope": "the contract's cwd, depth 2 (effects elsewhere are NOT observed)",
    "status": "OK"
   },
   "observed_modifies": false,
   "predicted_adverse_bp": 3000,
   "predicted_modifies": false,
   "receipt_verdict": "PASS",
   "would_enforce": "PASS"
  },
  {
   "argv": [
    "date"
   ],
   "effect_class": "read",
   "error_class": "true_negative",
   "exit_code": 0,
   "observed_adverse": 0,
   "observed_fx": {
    "created": 0,
    "deleted": 0,
    "modified": 0,
    "scope": "the contract's cwd, depth 2 (effects elsewhere are NOT observed)",
    "status": "OK"
   },
   "observed_modifies": false,
   "predicted_adverse_bp": 3000,
   "predicted_modifies": false,
   "receipt_verdict": "PASS",
   "would_enforce": "PASS"
  },
  {
   "argv": [
    "stat",
    "seed.txt"
   ],
   "effect_class": "read",
   "error_class": "true_negative",
   "exit_code": 0,
   "observed_adverse": 0,
   "observed_fx": {
    "created": 0,
    "deleted": 0,
    "modified": 0,
    "scope": "the contract's cwd, depth 2 (effects elsewhere are NOT observed)",
    "status": "OK"
   },
   "observed_modifies": false,
   "predicted_adverse_bp": 3333,
   "predicted_modifies": false,
   "receipt_verdict": "PASS",
   "would_enforce": "PASS"
  },
  {
   "argv": [
    "du",
    "-s",
    "."
   ],
   "effect_class": "read",
   "error_class": "true_negative",
   "exit_code": 0,
   "observed_adverse": 0,
   "observed_fx": {
    "created": 0,
    "deleted": 0,
    "modified": 0,
    "scope": "the contract's cwd, depth 2 (effects elsewhere are NOT observed)",
    "status": "OK"
   },
   "observed_modifies": false,
   "predicted_adverse_bp": 3076,
   "predicted_modifies": false,
   "receipt_verdict": "PASS",
   "would_enforce": "PASS"
  },
  {
   "argv": [
    "true"
   ],
   "effect_class": "read",
   "error_class": "true_negative",
   "exit_code": 0,
   "observed_adverse": 0,
   "observed_fx": {
    "created": 0,
    "deleted": 0,
    "modified": 0,
    "scope": "the contract's cwd, depth 2 (effects elsewhere are NOT observed)",
    "status": "OK"
   },
   "observed_modifies": false,
   "predicted_adverse_bp": 3000,
   "predicted_modifies": false,
   "receipt_verdict": "PASS",
   "would_enforce": "PASS"
  },
  {
   "argv": [
    "false"
   ],
   "effect_class": "read",
   "error_class": "false_negative",
   "exit_code": 1,
   "observed_adverse": 1,
   "observed_fx": {
    "created": 0,
    "deleted": 0,
    "modified": 0,
    "scope": "the contract's cwd, depth 2 (effects elsewhere are NOT observed)",
    "status": "OK"
   },
   "observed_modifies": false,
   "predicted_adverse_bp": 3000,
   "predicted_modifies": false,
   "receipt_verdict": "PASS",
   "would_enforce": "PASS"
  },
  {
   "argv": [
    "seq",
    "3"
   ],
   "effect_class": "read",
   "error_class": "true_negative",
   "exit_code": 0,
   "observed_adverse": 0,
   "observed_fx": {
    "created": 0,
    "deleted": 0,
    "modified": 0,
    "scope": "the contract's cwd, depth 2 (effects elsewhere are NOT observed)",
    "status": "OK"
   },
   "observed_modifies": false,
   "predicted_adverse_bp": 3125,
   "predicted_modifies": false,
   "receipt_verdict": "PASS",
   "would_enforce": "PASS"
  },
  {
   "argv": [
    "printf",
    "x"
   ],
   "effect_class": "read",
   "error_class": "true_negative",
   "exit_code": 0,
   "observed_adverse": 0,
   "observed_fx": {
    "created": 0,
    "deleted": 0,
    "modified": 0,
    "scope": "the contract's cwd, depth 2 (effects elsewhere are NOT observed)",
    "status": "OK"
   },
   "observed_modifies": false,
   "predicted_adverse_bp": 3000,
   "predicted_modifies": false,
   "receipt_verdict": "PASS",
   "would_enforce": "PASS"
  },
  {
   "argv": [
    "find",
    ".",
    "-name",
    "*.txt"
   ],
   "effect_class": "read",
   "error_class": "true_negative",
   "exit_code": 0,
   "observed_adverse": 0,
   "observed_fx": {
    "created": 0,
    "deleted": 0,
    "modified": 0,
    "scope": "the contract's cwd, depth 2 (effects elsewhere are NOT observed)",
    "status": "OK"
   },
   "observed_modifies": false,
   "predicted_adverse_bp": 3000,
   "predicted_modifies": false,
   "receipt_verdict": "PASS",
   "would_enforce": "PASS"
  },
  {
   "argv": [
    "find",
    ".",
    "-maxdepth",
    "0",
    "-fprint",
    "f1.out"
   ],
   "effect_class": "read",
   "error_class": "underprediction",
   "exit_code": 0,
   "observed_adverse": 0,
   "observed_fx": {
    "created": 1,
    "deleted": 0,
    "modified": 0,
    "scope": "the contract's cwd, depth 2 (effects elsewhere are NOT observed)",
    "status": "OK"
   },
   "observed_modifies": true,
   "predicted_adverse_bp": 3000,
   "predicted_modifies": false,
   "receipt_verdict": "PASS",
   "would_enforce": "PASS"
  },
  {
   "argv": [
    "sort",
    "seed.txt"
   ],
   "effect_class": "write",
   "error_class": "false_positive",
   "exit_code": 0,
   "observed_adverse": 0,
   "observed_fx": {
    "created": 0,
    "deleted": 0,
    "modified": 0,
    "scope": "the contract's cwd, depth 2 (effects elsewhere are NOT observed)",
    "status": "OK"
   },
   "observed_modifies": false,
   "predicted_adverse_bp": 8000,
   "predicted_modifies": true,
   "receipt_verdict": "PASS",
   "would_enforce": "PASS"
  },
  {
   "argv": [
    "sort",
    "-o",
    "sorted.out",
    "seed.txt"
   ],
   "effect_class": "write",
   "error_class": "false_positive",
   "exit_code": 0,
   "observed_adverse": 0,
   "observed_fx": {
    "created": 1,
    "deleted": 0,
    "modified": 0,
    "scope": "the contract's cwd, depth 2 (effects elsewhere are NOT observed)",
    "status": "OK"
   },
   "observed_modifies": true,
   "predicted_adverse_bp": 8000,
   "predicted_modifies": true,
   "receipt_verdict": "PASS",
   "would_enforce": "PASS"
  },
  {
   "argv": [
    "sed",
    "-n",
    "1p",
    "seed.txt"
   ],
   "effect_class": "write",
   "error_class": "false_positive",
   "exit_code": 0,
   "observed_adverse": 0,
   "observed_fx": {
    "created": 0,
    "deleted": 0,
    "modified": 0,
    "scope": "the contract's cwd, depth 2 (effects elsewhere are NOT observed)",
    "status": "OK"
   },
   "observed_modifies": false,
   "predicted_adverse_bp": 8000,
   "predicted_modifies": true,
   "receipt_verdict": "PASS",
   "would_enforce": "PASS"
  },
  {
   "argv": [
    "sed",
    "-i",
    "s/a/b/",
    "edit.txt"
   ],
   "effect_class": "write",
   "error_class": "false_positive",
   "exit_code": 0,
   "observed_adverse": 0,
   "observed_fx": {
    "created": 0,
    "deleted": 0,
    "modified": 1,
    "scope": "the contract's cwd, depth 2 (effects elsewhere are NOT observed)",
    "status": "OK"
   },
   "observed_modifies": true,
   "predicted_adverse_bp": 8000,
   "predicted_modifies": true,
   "receipt_verdict": "PASS",
   "would_enforce": "PASS"
  },
  {
   "argv": [
    "touch",
    "t1"
   ],
   "effect_class": "write",
   "error_class": "false_positive",
   "exit_code": 0,
   "observed_adverse": 0,
   "observed_fx": {
    "created": 1,
    "deleted": 0,
    "modified": 0,
    "scope": "the contract's cwd, depth 2 (effects elsewhere are NOT observed)",
    "status": "OK"
   },
   "observed_modifies": true,
   "predicted_adverse_bp": 8000,
   "predicted_modifies": true,
   "receipt_verdict": "PASS",
   "would_enforce": "PASS"
  },
  {
   "argv": [
    "touch",
    "seed.txt"
   ],
   "effect_class": "write",
   "error_class": "false_positive",
   "exit_code": 0,
   "observed_adverse": 0,
   "observed_fx": {
    "created": 0,
    "deleted": 0,
    "modified": 1,
    "scope": "the contract's cwd, depth 2 (effects elsewhere are NOT observed)",
    "status": "OK"
   },
   "observed_modifies": true,
   "predicted_adverse_bp": 8000,
   "predicted_modifies": true,
   "receipt_verdict": "PASS",
   "would_enforce": "PASS"
  },
  {
   "argv": [
    "cp",
    "seed.txt",
    "copy.txt"
   ],
   "effect_class": "write",
   "error_class": "false_positive",
   "exit_code": 0,
   "observed_adverse": 0,
   "observed_fx": {
    "created": 1,
    "deleted": 0,
    "modified": 0,
    "scope": "the contract's cwd, depth 2 (effects elsewhere are NOT observed)",
    "status": "OK"
   },
   "observed_modifies": true,
   "predicted_adverse_bp": 8000,
   "predicted_modifies": true,
   "receipt_verdict": "PASS",
   "would_enforce": "PASS"
  },
  {
   "argv": [
    "mv",
    "mv_src",
    "mv_dst"
   ],
   "effect_class": "write",
   "error_class": "false_positive",
   "exit_code": 0,
   "observed_adverse": 0,
   "observed_fx": {
    "created": 1,
    "deleted": 1,
    "modified": 0,
    "scope": "the contract's cwd, depth 2 (effects elsewhere are NOT observed)",
    "status": "OK"
   },
   "observed_modifies": true,
   "predicted_adverse_bp": 8000,
   "predicted_modifies": true,
   "receipt_verdict": "PASS",
   "would_enforce": "PASS"
  },
  {
   "argv": [
    "mkdir",
    "d1"
   ],
   "effect_class": "write",
   "error_class": "false_positive",
   "exit_code": 0,
   "observed_adverse": 0,
   "observed_fx": {
    "created": 1,
    "deleted": 0,
    "modified": 0,
    "scope": "the contract's cwd, depth 2 (effects elsewhere are NOT observed)",
    "status": "OK"
   },
   "observed_modifies": true,
   "predicted_adverse_bp": 8000,
   "predicted_modifies": true,
   "receipt_verdict": "PASS",
   "would_enforce": "PASS"
  },
  {
   "argv": [
    "mkdir",
    "seed.txt"
   ],
   "effect_class": "write",
   "error_class": "true_positive",
   "exit_code": 1,
   "observed_adverse": 1,
   "observed_fx": {
    "created": 0,
    "deleted": 0,
    "modified": 0,
    "scope": "the contract's cwd, depth 2 (effects elsewhere are NOT observed)",
    "status": "OK"
   },
   "observed_modifies": false,
   "predicted_adverse_bp": 8000,
   "predicted_modifies": true,
   "receipt_verdict": "PASS",
   "would_enforce": "PASS"
  },
  {
   "argv": [
    "rm",
    "-f",
    "del.txt"
   ],
   "effect_class": "delete",
   "error_class": "false_positive",
   "exit_code": 0,
   "observed_adverse": 0,
   "observed_fx": {
    "created": 0,
    "deleted": 1,
    "modified": 0,
    "scope": "the contract's cwd, depth 2 (effects elsewhere are NOT observed)",
    "status": "OK"
   },
   "observed_modifies": true,
   "predicted_adverse_bp": 9000,
   "predicted_modifies": true,
   "receipt_verdict": "OBSERVE_WOULD_DENY:FORECAST_IMPACT_DATA",
   "would_enforce": "DENY:FORECAST_IMPACT_DATA"
  },
  {
   "argv": [
    "rm",
    "missing"
   ],
   "effect_class": "delete",
   "error_class": "true_positive",
   "exit_code": 1,
   "observed_adverse": 1,
   "observed_fx": {
    "created": 0,
    "deleted": 0,
    "modified": 0,
    "scope": "the contract's cwd, depth 2 (effects elsewhere are NOT observed)",
    "status": "OK"
   },
   "observed_modifies": false,
   "predicted_adverse_bp": 9000,
   "predicted_modifies": true,
   "receipt_verdict": "OBSERVE_WOULD_DENY:FORECAST_IMPACT_DATA",
   "would_enforce": "DENY:FORECAST_IMPACT_DATA"
  },
  {
   "argv": [
    "rmdir",
    "emptydir"
   ],
   "effect_class": "delete",
   "error_class": "false_positive",
   "exit_code": 0,
   "observed_adverse": 0,
   "observed_fx": {
    "created": 0,
    "deleted": 1,
    "modified": 0,
    "scope": "the contract's cwd, depth 2 (effects elsewhere are NOT observed)",
    "status": "OK"
   },
   "observed_modifies": true,
   "predicted_adverse_bp": 9000,
   "predicted_modifies": true,
   "receipt_verdict": "OBSERVE_WOULD_DENY:FORECAST_IMPACT_DATA",
   "would_enforce": "DENY:FORECAST_IMPACT_DATA"
  },
  {
   "argv": [
    "truncate",
    "-s",
    "0",
    "trunc.txt"
   ],
   "effect_class": "delete",
   "error_class": "false_positive",
   "exit_code": 0,
   "observed_adverse": 0,
   "observed_fx": {
    "created": 0,
    "deleted": 0,
    "modified": 1,
    "scope": "the contract's cwd, depth 2 (effects elsewhere are NOT observed)",
    "status": "OK"
   },
   "observed_modifies": true,
   "predicted_adverse_bp": 9000,
   "predicted_modifies": true,
   "receipt_verdict": "OBSERVE_WOULD_DENY:FORECAST_IMPACT_DATA",
   "would_enforce": "DENY:FORECAST_IMPACT_DATA"
  },
  {
   "argv": [
    "ln",
    "-s",
    "seed.txt",
    "link"
   ],
   "effect_class": "write",
   "error_class": "false_positive",
   "exit_code": 0,
   "observed_adverse": 0,
   "observed_fx": {
    "created": 1,
    "deleted": 0,
    "modified": 0,
    "scope": "the contract's cwd, depth 2 (effects elsewhere are NOT observed)",
    "status": "OK"
   },
   "observed_modifies": true,
   "predicted_adverse_bp": 8000,
   "predicted_modifies": true,
   "receipt_verdict": "PASS",
   "would_enforce": "PASS"
  },
  {
   "argv": [
    "tee",
    "tee.out"
   ],
   "effect_class": "write",
   "error_class": "false_positive",
   "exit_code": 0,
   "observed_adverse": 0,
   "observed_fx": {
    "created": 1,
    "deleted": 0,
    "modified": 0,
    "scope": "the contract's cwd, depth 2 (effects elsewhere are NOT observed)",
    "status": "OK"
   },
   "observed_modifies": true,
   "predicted_adverse_bp": 8000,
   "predicted_modifies": true,
   "receipt_verdict": "PASS",
   "would_enforce": "PASS"
  },
  {
   "argv": [
    "cp",
    "nope",
    "x"
   ],
   "effect_class": "write",
   "error_class": "true_positive",
   "exit_code": 1,
   "observed_adverse": 1,
   "observed_fx": {
    "created": 0,
    "deleted": 0,
    "modified": 0,
    "scope": "the contract's cwd, depth 2 (effects elsewhere are NOT observed)",
    "status": "OK"
   },
   "observed_modifies": false,
   "predicted_adverse_bp": 8000,
   "predicted_modifies": true,
   "receipt_verdict": "PASS",
   "would_enforce": "PASS"
  }
 ],
 "schema": "cain.bench.prediction.evolution3.v1",
 "would_enforce_deny": {
  "DENY:FORECAST_IMPACT_DATA": 4,
  "PASS": 31
 }
}
