데이터 게이트 검증기와 DAG 연결

docs/ROADMAP_WBS.md

@@ -36,7 +36,7 @@
 | **D2 공식 결정론** | 149개 공식 ID 전부 lifecycle 등록 | 269개 등록 (100%) ✅ |
 | **D3 리스크 제어** | Core/Satellite/Cash 버킷 밴드 위반 0건 | RISK_ON 밴드 내 유지 중 |
 | **D4 알파 피드백** | 예측→실현 수익 루프 30건 이상 누적 | 0건 (DATA_GATED ~2026-07-15) |
-| **D5 실행 자동화** | run_all 1회 실행으로 전체 파이프라인 완결 | 90단계 DAG 구축 완료 ✅ |
+| **D5 실행 자동화** | run_all 1회 실행으로 전체 파이프라인 완결 | 96단계 DAG 구축 완료 ✅ |
 
 ---
 
@@ -586,7 +586,7 @@ CI 게이트:
   honest_proof_score: 50.95 →  목표: ≥70 (T+20 30건 → 70.95 자동 달성 예상)
 
 자동화:
-  run_all 성공률:      90단계 DAG PASS →  목표: ≥95% ✅ (step_count=90, wave_0~9)
+  run_all 성공률:      96단계 DAG PASS →  목표: ≥95% ✅ (step_count=96, wave_0~9)
   CI/CD 커버리지:      100%   →  목표: 100% ✅ (Synology act_runner 온라인, 4게이트 PASS)
   수동 개입 횟수:      매일   →  목표: ≤1회/주 (setupDailyRunAllTrigger 설정 후)
 ```

package.json

@@ -16,8 +16,14 @@
     "ops:package": "python tools/refresh_trading_calendar.py && python tools/prepare_upload_zip.py --validation-mode release --profile",
     "prepare-upload-zip": "python tools/refresh_trading_calendar.py && python tools/prepare_upload_zip.py",
     "ops:audit": "python tools/harness_coverage_auditor.py",
+    "build-prediction-accuracy-harness": "python tools/build_prediction_accuracy_harness_v2.py",
+    "build-alpha-feedback-loop": "python tools/build_alpha_feedback_loop_v2.py",
+    "build-operational-alpha-calibration": "python tools/build_operational_alpha_calibration_v2.py",
     "build-realized-performance": "python tools/build_realized_performance_v1.py",
     "validate-completion-harness": "python tools/validate_completion_harness_instructions_v1.py",
+    "validate-prediction-accuracy-harness": "python tools/validate_prediction_accuracy_harness_v2.py",
+    "validate-alpha-feedback-loop": "python tools/validate_alpha_feedback_loop_v2.py",
+    "validate-operational-alpha-calibration": "python tools/validate_operational_alpha_calibration_v2.py",
     "validate-realized-performance": "python tools/validate_realized_performance_v1.py",
     "validate-gas-recovery": "python tools/validate_gas_orchestration_recovery_v1.py",
     "ops:clean": "python tools/clean_temp_artifacts_v1.py",

runtime/refactor_baseline_v1.yaml

@@ -1,9 +1,9 @@
 {
   "formula_id": "AUDIT_REPOSITORY_ENTROPY_V2",
   "gate": "PASS",
-  "total_file_count": 1884,
+  "total_file_count": 1890,
-  "package_script_count": 23,
+  "package_script_count": 29,
-  "temp_json_count": 188,
+  "temp_json_count": 191,
   "budget": {
     "schema_version": "repository_entropy_budget.v1",
     "max_total_files": 2200,
@@ -15,5 +15,5 @@
       "keep package scripts within release envelope"
     ]
   },
-  "source_zip_sha256": "469e09441818688b861efa6d6ee1bd6123806150c2940608ba964b44bcf50eb0"
+  "source_zip_sha256": "1b582cc260c09b41dfd2c361bdd40705ac003b88ad8a09b42bbba615189cbf2b"
 }

spec/41_release_dag.yaml

@@ -1,5 +1,5 @@
 schema_version: release_dag.v3
-step_count: 90
+step_count: 96
 goal: Linearize package.json scripts into a validated DAG execution graph.
 execution_order:
   # 토폴로지 정렬 기준 병렬 실행 wave (의존성 없는 노드들을 동시에 실행 가능)
@@ -85,13 +85,19 @@ execution_order:
   wave_6:
     - build_algorithm_guidance_proof
     - build_artifact_chain_hash
+    - build_alpha_feedback_loop
     - build_honest_proof_gap_analyzer
+    - build_operational_alpha_calibration
+    - build_prediction_accuracy_harness
     - validate_json_generator_outputs
+    - validate_alpha_feedback_loop
     - validate_llm_copy_only
     - validate_llm_determinism
     - validate_llm_regression
     - validate_low_capability
     - validate_provenance
+    - validate_prediction_accuracy_harness
+    - validate_operational_alpha_calibration
     - validate_render_diff
     - validate_report_numeric_consistency
     - validate_report_section_completeness
@@ -163,6 +169,42 @@ dag:
       artifact_policy: "keep"
       note: "WBS-4.1 realized performance replay summary — non-blocking diagnostic"
 
+    build_prediction_accuracy_harness:
+      id: build_prediction_accuracy_harness
+      command: ["python", "tools/build_prediction_accuracy_harness_v2.py"]
+      inputs: ["tools/build_prediction_accuracy_harness_v2.py", "Temp/proposal_evaluation_history.json"]
+      outputs: ["Temp/prediction_accuracy_harness_v2.json"]
+      depends_on: ["update_proposal_evaluation_history"]
+      timeout_sec: 30
+      cache_key: "build_prediction_accuracy_harness_v2"
+      strict: false
+      artifact_policy: "keep"
+      note: "WBS-4.2 prediction accuracy harness — non-blocking diagnostic"
+
+    build_alpha_feedback_loop:
+      id: build_alpha_feedback_loop
+      command: ["python", "tools/build_alpha_feedback_loop_v2.py"]
+      inputs: ["tools/build_alpha_feedback_loop_v2.py", "Temp/proposal_evaluation_history.json"]
+      outputs: ["Temp/alpha_feedback_loop_v2.json"]
+      depends_on: ["update_proposal_evaluation_history"]
+      timeout_sec: 30
+      cache_key: "build_alpha_feedback_loop_v2"
+      strict: false
+      artifact_policy: "keep"
+      note: "WBS-4.3 alpha feedback loop — non-blocking diagnostic"
+
+    build_operational_alpha_calibration:
+      id: build_operational_alpha_calibration
+      command: ["python", "tools/build_operational_alpha_calibration_v2.py"]
+      inputs: ["tools/build_operational_alpha_calibration_v2.py", "Temp/outcome_quality_score_v1.json", "Temp/prediction_accuracy_harness_v2.json", "Temp/trade_quality_from_t5_v1.json", "Temp/smart_cash_recovery_v5.json"]
+      outputs: ["Temp/operational_alpha_calibration_v2.json"]
+      depends_on: ["build_prediction_accuracy_harness", "build_alpha_feedback_loop", "build_realized_performance"]
+      timeout_sec: 30
+      cache_key: "build_operational_alpha_calibration_v2"
+      strict: false
+      artifact_policy: "keep"
+      note: "WBS-4.3 operational alpha calibration — non-blocking diagnostic"
+
     build_factor_shadow_eligibility:
       id: build_factor_shadow_eligibility
       command: ["python", "tools/build_factor_shadow_eligibility_v1.py"]
@@ -581,6 +623,39 @@ dag:
       strict: true
       artifact_policy: "keep"
 
+    validate_prediction_accuracy_harness:
+      id: validate_prediction_accuracy_harness
+      command: ["python", "tools/validate_prediction_accuracy_harness_v2.py"]
+      inputs: ["tools/validate_prediction_accuracy_harness_v2.py", "Temp/prediction_accuracy_harness_v2.json"]
+      outputs: ["Temp/validate_prediction_accuracy_harness_v2.json"]
+      depends_on: ["build_prediction_accuracy_harness"]
+      timeout_sec: 30
+      cache_key: "validate_prediction_accuracy_harness_v2"
+      strict: true
+      artifact_policy: "keep"
+
+    validate_alpha_feedback_loop:
+      id: validate_alpha_feedback_loop
+      command: ["python", "tools/validate_alpha_feedback_loop_v2.py"]
+      inputs: ["tools/validate_alpha_feedback_loop_v2.py", "Temp/alpha_feedback_loop_v2.json"]
+      outputs: ["Temp/validate_alpha_feedback_loop_v2.json"]
+      depends_on: ["build_alpha_feedback_loop"]
+      timeout_sec: 30
+      cache_key: "validate_alpha_feedback_loop_v2"
+      strict: true
+      artifact_policy: "keep"
+
+    validate_operational_alpha_calibration:
+      id: validate_operational_alpha_calibration
+      command: ["python", "tools/validate_operational_alpha_calibration_v2.py"]
+      inputs: ["tools/validate_operational_alpha_calibration_v2.py", "Temp/operational_alpha_calibration_v2.json"]
+      outputs: ["Temp/validate_operational_alpha_calibration_v2.json"]
+      depends_on: ["build_operational_alpha_calibration"]
+      timeout_sec: 30
+      cache_key: "validate_operational_alpha_calibration_v2"
+      strict: true
+      artifact_policy: "keep"
+
     validate_realized_performance:
       id: validate_realized_performance
       command: ["python", "tools/validate_realized_performance_v1.py"]
@@ -1214,7 +1289,7 @@ dag:
       command: ["python", "tools/prepare_upload_zip.py", "--skip-validate", "--skip-convert", "--validation-mode", "package-only"]
       inputs: ["tools/prepare_upload_zip.py"]
       outputs: []
-      depends_on: ["audit_entropy", "validate_specs", "validate_active_manifest", "validate_report_sync", "validate_report_numeric_consistency", "validate_field_dict", "validate_provenance", "validate_low_capability", "validate_golden_coverage", "validate_calibration", "validate_schema_model", "validate_gas_adapter", "validate_agents_shrink", "validate_no_replay_live_mix", "validate_realized_performance", "validate_runtime_source_whitelist", "validate_cash_ledger", "validate_factor_lifecycle", "validate_factor_lifecycle_completeness", "validate_metric_alias_collision", "validate_architecture_boundaries", "validate_module_io_coverage", "validate_artifact_chain_hash", "validate_artifact_sync", "validate_renderer_no_calc", "validate_packaged_refs", "validate_property_invariants", "validate_anti_late_entry", "validate_rule_lifecycle", "validate_change_requests", "validate_completion_harness_instructions", "validate_engine_health_card", "validate_llm_regression", "validate_llm_copy_only", "build_final_decision", "build_final_context", "build_provenance_ledger", "build_live_replay_separation", "build_late_chase_attribution", "build_profit_giveback_ratchet", "build_shadow_ledger", "build_operating_cadence_signal", "build_engine_health_card", "build_module_io_coverage", "build_artifact_chain_hash", "build_report", "build_bundle", "build_schema_models", "build_architecture_boundaries", "validate_decision_trace", "validate_factor_conflicts", "validate_no_lookahead", "validate_execution_sim", "validate_render_diff", "build_shadow_promotion", "validate_llm_determinism", "build_time_stop_forecast", "validate_live_activation", "build_rebalance_sheet"]
+      depends_on: ["audit_entropy", "validate_specs", "validate_active_manifest", "validate_report_sync", "validate_report_numeric_consistency", "validate_field_dict", "validate_provenance", "validate_low_capability", "validate_golden_coverage", "validate_calibration", "validate_schema_model", "validate_gas_adapter", "validate_agents_shrink", "validate_no_replay_live_mix", "validate_prediction_accuracy_harness", "validate_alpha_feedback_loop", "validate_operational_alpha_calibration", "validate_realized_performance", "validate_runtime_source_whitelist", "validate_cash_ledger", "validate_factor_lifecycle", "validate_factor_lifecycle_completeness", "validate_metric_alias_collision", "validate_architecture_boundaries", "validate_module_io_coverage", "validate_artifact_chain_hash", "validate_artifact_sync", "validate_renderer_no_calc", "validate_packaged_refs", "validate_property_invariants", "validate_anti_late_entry", "validate_rule_lifecycle", "validate_change_requests", "validate_completion_harness_instructions", "validate_engine_health_card", "validate_llm_regression", "validate_llm_copy_only", "build_final_decision", "build_final_context", "build_provenance_ledger", "build_live_replay_separation", "build_late_chase_attribution", "build_profit_giveback_ratchet", "build_shadow_ledger", "build_operating_cadence_signal", "build_engine_health_card", "build_module_io_coverage", "build_artifact_chain_hash", "build_report", "build_bundle", "build_schema_models", "build_architecture_boundaries", "validate_decision_trace", "validate_factor_conflicts", "validate_no_lookahead", "validate_execution_sim", "validate_render_diff", "build_shadow_promotion", "validate_llm_determinism", "build_time_stop_forecast", "validate_live_activation", "build_rebalance_sheet", "build_prediction_accuracy_harness", "build_alpha_feedback_loop", "build_operational_alpha_calibration"]
       timeout_sec: 60
       cache_key: "prepare_zip_v1"
       strict: true

src/quant_engine/prepare_upload_zip.py

@@ -89,6 +89,11 @@ TEMP_KEEP_FILES = {
     "canonical_artifact_resolver_v1.json",
     "final_execution_decision_v2.json",
     "prediction_accuracy_harness_v2.json",
+    "validate_prediction_accuracy_harness_v2.json",
+    "alpha_feedback_loop_v2.json",
+    "validate_alpha_feedback_loop_v2.json",
+    "operational_alpha_calibration_v2.json",
+    "validate_operational_alpha_calibration_v2.json",
     "realized_performance_v1.json",
     "validate_realized_performance_v1.json",
     "single_truth_ledger_v2.json",

tools/validate_alpha_feedback_loop_v2.py

@@ -0,0 +1,92 @@
+"""validate_alpha_feedback_loop_v2.py — ALPHA_FEEDBACK_LOOP_VALIDATE_V2
+
+Temp/alpha_feedback_loop_v2.json의 구조와 데이터 게이트 상태를 검증한다.
+"""
+from __future__ import annotations
+
+import argparse
+import json
+from pathlib import Path
+from typing import Any
+
+ROOT = Path(__file__).resolve().parents[1]
+DEFAULT_INPUT = ROOT / "Temp" / "alpha_feedback_loop_v2.json"
+DEFAULT_OUT = ROOT / "Temp" / "validate_alpha_feedback_loop_v2.json"
+FORMULA_ID = "ALPHA_FEEDBACK_LOOP_VALIDATE_V2"
+
+
+def _load(path: Path) -> Any:
+    if not path.exists():
+        return {}
+    try:
+        return json.loads(path.read_text(encoding="utf-8"))
+    except Exception:
+        return {}
+
+
+def _is_dict(value: Any) -> bool:
+    return isinstance(value, dict)
+
+
+def main() -> int:
+    ap = argparse.ArgumentParser()
+    ap.add_argument("--input", default=str(DEFAULT_INPUT))
+    ap.add_argument("--out", default=str(DEFAULT_OUT))
+    args = ap.parse_args()
+
+    input_path = Path(args.input)
+    input_path = input_path if input_path.is_absolute() else ROOT / input_path
+    out_path = Path(args.out)
+    out_path = out_path if out_path.is_absolute() else ROOT / out_path
+
+    payload = _load(input_path)
+    errors: list[str] = []
+
+    if not _is_dict(payload):
+        errors.append("payload must be object")
+    else:
+        if payload.get("formula_id") != "ALPHA_FEEDBACK_LOOP_V2":
+            errors.append("formula_id mismatch")
+
+        status = str(payload.get("status") or "")
+        if status not in {"DATA_INSUFFICIENT", "ANALYZED"}:
+            errors.append(f"status={status}")
+
+        if "cases_analyzed" not in payload or not isinstance(payload.get("cases_analyzed"), int):
+            errors.append("cases_analyzed must be int")
+        if "recommended_adjustments" not in payload or not isinstance(payload.get("recommended_adjustments"), list):
+            errors.append("recommended_adjustments must be list")
+
+        cases = payload.get("cases_analyzed")
+        recs = payload.get("recommended_adjustments")
+        if isinstance(cases, int):
+            if cases < 10 and status != "DATA_INSUFFICIENT":
+                errors.append("cases_analyzed < 10 requires DATA_INSUFFICIENT")
+            if cases >= 10 and status != "ANALYZED":
+                errors.append("cases_analyzed >= 10 requires ANALYZED")
+        if isinstance(recs, list) and status == "DATA_INSUFFICIENT" and recs:
+            errors.append("DATA_INSUFFICIENT must not carry recommendations")
+        if status == "ANALYZED":
+            for key in [
+                "active_signal_rate_pct", "active_signal_n",
+                "passive_signal_rate_pct", "passive_signal_n",
+                "combined_rate_pct", "sell_signal_rate_pct", "sell_signal_n",
+                "pa1_current_ratio", "pa1_thesis_sum", "pa1_antithesis_sum",
+                "component_analysis", "note",
+            ]:
+                if key not in payload:
+                    errors.append(f"missing field: {key}")
+
+    result = {
+        "formula_id": FORMULA_ID,
+        "gate": "PASS" if not errors else "FAIL",
+        "checked_file": str(Path(args.input).as_posix()),
+        "errors": errors,
+    }
+    out_path.write_text(json.dumps(result, ensure_ascii=False, indent=2) + "\n", encoding="utf-8")
+    print(json.dumps(result, ensure_ascii=False, indent=2))
+    return 0 if not errors else 1
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())

tools/validate_operational_alpha_calibration_v2.py

@@ -0,0 +1,105 @@
+"""validate_operational_alpha_calibration_v2.py — OPERATIONAL_ALPHA_CALIBRATION_VALIDATE_V2
+
+Temp/operational_alpha_calibration_v2.json의 최소 계약을 검증한다.
+"""
+from __future__ import annotations
+
+import argparse
+import json
+from pathlib import Path
+from typing import Any
+
+ROOT = Path(__file__).resolve().parents[1]
+DEFAULT_INPUT = ROOT / "Temp" / "operational_alpha_calibration_v2.json"
+DEFAULT_OUT = ROOT / "Temp" / "validate_operational_alpha_calibration_v2.json"
+FORMULA_ID = "OPERATIONAL_ALPHA_CALIBRATION_VALIDATE_V2"
+
+
+def _load(path: Path) -> Any:
+    if not path.exists():
+        return {}
+    try:
+        return json.loads(path.read_text(encoding="utf-8"))
+    except Exception:
+        return {}
+
+
+def _is_dict(value: Any) -> bool:
+    return isinstance(value, dict)
+
+
+def main() -> int:
+    ap = argparse.ArgumentParser()
+    ap.add_argument("--input", default=str(DEFAULT_INPUT))
+    ap.add_argument("--out", default=str(DEFAULT_OUT))
+    args = ap.parse_args()
+
+    input_path = Path(args.input)
+    input_path = input_path if input_path.is_absolute() else ROOT / input_path
+    out_path = Path(args.out)
+    out_path = out_path if out_path.is_absolute() else ROOT / out_path
+
+    payload = _load(input_path)
+    errors: list[str] = []
+
+    if not _is_dict(payload):
+        errors.append("payload must be object")
+    else:
+        if payload.get("formula_id") != "OPERATIONAL_ALPHA_CALIBRATION_V2":
+            errors.append("formula_id mismatch")
+        if payload.get("gate") not in {"PERFORMANCE_READY", "NOT_READY"}:
+            errors.append(f"gate={payload.get('gate')}")
+        if "performance_ready" not in payload or not isinstance(payload.get("performance_ready"), bool):
+            errors.append("performance_ready must be bool")
+        if "confidence_score" not in payload or not isinstance(payload.get("confidence_score"), (int, float)):
+            errors.append("confidence_score must be numeric")
+        for key in ["metrics", "targets", "readiness_reasons"]:
+            if key not in payload:
+                errors.append(f"missing field: {key}")
+        metrics = payload.get("metrics")
+        if isinstance(metrics, dict):
+            for key in [
+                "outcome_quality_score",
+                "t20_operational_sample",
+                "t20_operational_pass_rate",
+                "t5_operational_sample",
+                "t5_operational_pass_rate",
+                "trade_quality_t5_score",
+                "value_damage_pct_avg",
+            ]:
+                if key not in metrics:
+                    errors.append(f"missing field: metrics.{key}")
+        targets = payload.get("targets")
+        if isinstance(targets, dict):
+            for key in [
+                "outcome_quality_score_min",
+                "t20_operational_sample_min",
+                "t20_operational_pass_rate_min",
+                "t5_operational_sample_min",
+                "t5_operational_pass_rate_min",
+                "trade_quality_t5_score_min",
+                "value_damage_pct_avg_max",
+            ]:
+                if key not in targets:
+                    errors.append(f"missing field: targets.{key}")
+
+        reasons = payload.get("readiness_reasons")
+        if isinstance(reasons, list):
+            if payload.get("gate") == "PERFORMANCE_READY" and reasons:
+                errors.append("PERFORMANCE_READY must not have readiness reasons")
+            if payload.get("gate") == "NOT_READY" and not reasons:
+                errors.append("NOT_READY must have readiness reasons")
+
+    result = {
+        "formula_id": FORMULA_ID,
+        "gate": "PASS" if not errors else "FAIL",
+        "checked_file": str(Path(args.input).as_posix()),
+        "errors": errors,
+    }
+    out_path.write_text(json.dumps(result, ensure_ascii=False, indent=2) + "\n", encoding="utf-8")
+    print(json.dumps(result, ensure_ascii=False, indent=2))
+    return 0 if not errors else 1
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())

tools/validate_prediction_accuracy_harness_v2.py

@@ -0,0 +1,131 @@
+"""validate_prediction_accuracy_harness_v2.py — PREDICTION_ACCURACY_HARNESS_VALIDATE_V2
+
+Temp/prediction_accuracy_harness_v2.json의 기본 구조와 허용된 데이터 게이트 상태를 검증한다.
+현재는 운영 T+5/T+20 표본이 부족할 수 있으므로 INSUFFICIENT_SAMPLES는 허용한다.
+"""
+from __future__ import annotations
+
+import argparse
+import json
+from pathlib import Path
+from typing import Any
+
+ROOT = Path(__file__).resolve().parents[1]
+DEFAULT_INPUT = ROOT / "Temp" / "prediction_accuracy_harness_v2.json"
+DEFAULT_OUT = ROOT / "Temp" / "validate_prediction_accuracy_harness_v2.json"
+FORMULA_ID = "PREDICTION_ACCURACY_HARNESS_VALIDATE_V2"
+ALLOWED_CALIBRATION = {
+    "CALIBRATED",
+    "MONITOR",
+    "PAE_CALIBRATION_REQUIRED",
+    "BUY_PROPOSAL_FROZEN_RECOMMEND",
+    "INSUFFICIENT_SAMPLES",
+}
+
+
+def _load(path: Path) -> Any:
+    if not path.exists():
+        return {}
+    try:
+        return json.loads(path.read_text(encoding="utf-8"))
+    except Exception:
+        return {}
+
+
+def _is_dict(value: Any) -> bool:
+    return isinstance(value, dict)
+
+
+def _ensure_fields(payload: dict[str, Any], path: str, fields: list[str], errors: list[str]) -> None:
+    block = payload
+    if path:
+        for part in path.split("."):
+            block = block.get(part) if isinstance(block, dict) else None
+    if not isinstance(block, dict):
+        errors.append(f"{path or 'root'} must be object")
+        return
+    for field in fields:
+        if field not in block:
+            errors.append(f"missing field: {path + '.' if path else ''}{field}")
+
+
+def main() -> int:
+    ap = argparse.ArgumentParser()
+    ap.add_argument("--input", default=str(DEFAULT_INPUT))
+    ap.add_argument("--out", default=str(DEFAULT_OUT))
+    args = ap.parse_args()
+
+    input_path = Path(args.input)
+    input_path = input_path if input_path.is_absolute() else ROOT / input_path
+    out_path = Path(args.out)
+    out_path = out_path if out_path.is_absolute() else ROOT / out_path
+
+    payload = _load(input_path)
+    errors: list[str] = []
+
+    if not _is_dict(payload):
+        errors.append("payload must be object")
+    else:
+        if payload.get("formula_id") != "PREDICTION_ACCURACY_HARNESS_V2":
+            errors.append("formula_id mismatch")
+
+        calibration_state = str(payload.get("calibration_state") or "")
+        if calibration_state not in ALLOWED_CALIBRATION:
+            errors.append(f"calibration_state={calibration_state}")
+
+        for key in [
+            "as_of_date",
+            "data_origin_audit",
+            "windows",
+            "evaluation_methodology",
+        ]:
+            if key not in payload:
+                errors.append(f"missing field: {key}")
+
+        audit = payload.get("data_origin_audit")
+        if isinstance(audit, dict):
+            for key in [
+                "operational_sample_count",
+                "replay_sample_count",
+                "untagged_row_count",
+                "unrealized_outcome_row_count",
+                "replay_in_live_stats",
+                "operational_only_accuracy",
+            ]:
+                if key not in audit:
+                    errors.append(f"missing field: data_origin_audit.{key}")
+
+        for key in [
+            "t1_op_rate", "t1_sample", "t5_op_rate", "t5_sample",
+            "t20_op_rate", "t20_sample", "t20_replay_rate", "t20_replay_sample",
+            "t20_replay_avg_return_pct", "t20_replay_stdev_return_pct",
+            "window_90d_rate",
+        ]:
+            if key not in payload:
+                errors.append(f"missing field: {key}")
+
+        windows = payload.get("windows")
+        if isinstance(windows, dict):
+            _ensure_fields(windows, "t1", ["all", "30d", "7d"], errors)
+            _ensure_fields(windows, "t5", ["all", "active_passive", "30d", "90d"], errors)
+            _ensure_fields(windows, "t20", ["operational", "operational_30d", "replay", "replay_return_dist"], errors)
+        else:
+            errors.append("windows must be object")
+
+        t5_sample = payload.get("t5_sample")
+        if isinstance(t5_sample, int) and t5_sample < 30 and calibration_state != "INSUFFICIENT_SAMPLES":
+            errors.append("t5_sample < 30 requires INSUFFICIENT_SAMPLES")
+
+    result = {
+        "formula_id": FORMULA_ID,
+        "gate": "PASS" if not errors else "FAIL",
+        "checked_file": str(Path(args.input).as_posix()),
+        "errors": errors,
+    }
+    out_path.write_text(json.dumps(result, ensure_ascii=False, indent=2) + "\n", encoding="utf-8")
+    print(json.dumps(result, ensure_ascii=False, indent=2))
+    return 0 if not errors else 1
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())