evaluation/pilot-eval.json

{
  "release": "v6",
  "phase": "F7/211-270",
  "benchmark_type": "simulated multi-tenant pilot engineering benchmark",
  "real_company_pilot": false,
  "passed": true,
  "case_counts": {
    "tenants": 3,
    "controlled_attacks": 12,
    "simulated_baseline_events": 18,
    "analyst_feedback": 6,
    "executive_summaries": 3,
    "tenant_isolation_checks": 3,
    "audit_records": 45
  },
  "metrics": {
    "controlled_attack_recall": 1.0,
    "simulated_baseline_specificity": 1.0,
    "tenant_isolation": 1.0,
    "audit_chain_integrity": 1.0,
    "audit_tamper_detection": 1.0,
    "feedback_does_not_mutate_weights": 1.0,
    "version_evidence_approval_complete": 1.0,
    "executive_summary_schema_complete": 1.0,
    "no_auto_execute": 1.0
  },
  "gates": {
    "controlled_attack_recall": true,
    "simulated_baseline_specificity": true,
    "tenant_isolation": true,
    "audit_chain_integrity": true,
    "audit_tamper_detection": true,
    "feedback_does_not_mutate_weights": true,
    "version_evidence_approval_complete": true,
    "executive_summary_schema_complete": true,
    "no_auto_execute": true
  },
  "limitations": [
    "No real company telemetry was used; this is not a production pilot claim.",
    "Risk scores are controlled fixtures used to verify tenant thresholds and audit behavior.",
    "Production false-positive and critical-miss rates remain unmeasured until a redacted 30-60 day tenant baseline is approved.",
    "The in-memory reference must be backed by tenant-scoped durable storage before production use."
  ]
}