캘리브레이션 거버넌스 도구 + WBS-7.1/7.2 실증 격차 가시화

캘리브레이션 백로그 → 우선순위 → 검토리포트 → 승인목록 → 결정초안으로
이어지는 임계값 보정 거버넌스 파이프라인을 추가하고, 2026-06-21
비판적 리뷰에서 발견한 두 가지 stale-수치 문제를 도구 차원에서 해소한다.

- registry_health(): 190여 개 임계값의 source별(SPEC_DERIVED/EXPERT_PRIOR/
  PROVISIONAL/CALIBRATED) 분포를 매 실행마다 자동 집계 — 수동 grep 불필요
- live_t5_status(): T+5 적중률을 하드코딩(35.86 리터럴) 대신
  Temp/prediction_accuracy_harness_v2.json에서 항상 최신값으로 읽음
- spec/calibration_registry.yaml: SEMI_CLUSTER_CAP_RISK_OFF 중복 id로
  인한 조용한 무시 버그 수정(SEMI_CLUSTER_CAP_RISK_OFF_MWA로 분리)
- spec/27_bch_calibration_runbook.yaml: current_status_2026_06_21 블록
  신설(단일 진실원천), 기존 05-30 스냅샷은 "역사적, 현재로 인용 금지"로 명시

.gitea/workflows/calibration_backlog.yml

@@ -0,0 +1,80 @@
+name: Calibration Backlog (Registry Drift Watch)
+
+on:
+  schedule:
+    - cron: "15 2 * * 1-5"   # UTC 02:15 = KST 11:15, weekday backlog update
+  workflow_dispatch:
+
+jobs:
+  build-calibration-backlog:
+    runs-on: self-hosted
+
+    steps:
+    - name: Checkout Code
+      run: |
+        if [ -d .git ]; then
+          git remote set-url origin http://x-access-token:${{ secrets.GITHUB_TOKEN }}@192.168.123.100:8418/KimJaeHyun/myfinance.git
+        else
+          git init
+          git remote add origin http://x-access-token:${{ secrets.GITHUB_TOKEN }}@192.168.123.100:8418/KimJaeHyun/myfinance.git
+        fi
+        git fetch origin main --depth=1
+        git reset --hard FETCH_HEAD
+
+    - name: Configure Runtime Paths
+      run: |
+        export PATH=/usr/local/bin:$PATH
+        echo "/usr/local/bin" >> $GITHUB_PATH
+        /usr/bin/python3 --version
+
+    - name: Setup Python Environment
+      run: |
+        VENV_BASE=/volume1/gitea/python_venv
+        REQ_HASH=$(md5sum tools/build_calibration_priority_v1.py 2>/dev/null | cut -d' ' -f1 || echo "calib-default")
+        VENV="$VENV_BASE/$REQ_HASH"
+
+        if [ ! -f "$VENV/bin/python" ]; then
+          mkdir -p "$VENV_BASE"
+          /usr/bin/python3 -m venv "$VENV"
+          if [ ! -f "$VENV/bin/pip" ]; then
+            curl -sS https://bootstrap.pypa.io/pip/3.8/get-pip.py -o get-pip.py
+            "$VENV/bin/python" get-pip.py --quiet
+            rm get-pip.py
+          fi
+          "$VENV/bin/pip" install --upgrade pip --quiet
+          "$VENV/bin/pip" install pyyaml --quiet
+        fi
+        echo "$VENV/bin" >> $GITHUB_PATH
+
+    - name: Validate Calibration Registry
+      run: python3 tools/validate_calibration_registry_v1.py
+
+    - name: Build Calibration Priority Backlog
+      run: python3 tools/build_calibration_priority_v1.py
+
+    - name: Build Calibration Change Ledger
+      run: python3 tools/build_calibration_change_ledger_v4.py
+
+    - name: Build Calibration Review Report
+      run: python3 tools/build_calibration_review_report_v1.py
+
+    - name: Build Calibration Approval List
+      run: python3 tools/build_calibration_approval_list_v1.py
+
+    - name: Build Calibration Decision Draft
+      run: python3 tools/build_calibration_decision_draft_v1.py
+
+    - name: Validate Calibration Change Ledger
+      run: python3 tools/validate_calibration_change_ledger_v1.py
+
+    - name: Summarize Backlog
+      if: always()
+      run: |
+        STATUS="${{ job.status }}"
+        echo "=== Calibration Backlog Result ==="
+        echo "status: $STATUS"
+        echo "priority: Temp/calibration_priority_v1.json"
+        echo "ledger: Temp/calibration_change_ledger_v4.json"
+        echo "review: Temp/calibration_review_report_v1.md"
+        echo "approval: Temp/calibration_approval_list_v1.md"
+        echo "decision: Temp/calibration_decision_draft_v1.md"

spec/27_bch_calibration_runbook.yaml

@@ -451,7 +451,30 @@ reject_conditions:
   - "sample_n < 30인 임계값을 '보정완료'로 처리"
 
 # ════════════════════════════════════════════════════════════════════════════
-# 현재 달성 현황 (2026-05-30)
+# 현재 달성 현황 (2026-06-21 재검증 — WBS-7.2)
+# ════════════════════════════════════════════════════════════════════════════
+# 주의: 아래 current_status_2026_05_30 블록은 그 날짜 기준 정적 스냅샷이며,
+# 이후 갱신되지 않은 채 docs/ROADMAP_WBS.md 등에서 "현재 상태"로 인용되어
+# 서로 다른 시점의 T+5 수치(54.76%/35.86%)가 혼재하는 문제를 일으켰다.
+# Temp/honest_performance_guard_v1.json(생성: 2026-06-14)과
+# Temp/prediction_accuracy_harness_v2.json(생성: 2026-06-21, 7일 더 최신)을
+# 직접 재확인한 결과는 다음과 같다 — 이 블록을 단일 진실원천으로 삼는다.
+current_status_2026_06_21:
+  source_of_truth: "Temp/prediction_accuracy_harness_v2.json (as_of_date=2026-06-21, 가장 최신)"
+  t1_match_rate_pct: 52.94   # sample=68, decisive_sample=53, rate_decisive=67.92
+  t5_match_rate_pct: null    # sample=0 — INSUFFICIENT_SAMPLES. honest_performance_guard_v1.json(2026-06-14)의
+                             # 35.86%는 7일 전 스냅샷이며 표본이 0으로 줄어 더 이상 유효하지 않음.
+  t5_sample_regression_note: >
+    cases_analyzed가 141건(2026-05-30 기준)에서 t5_sample=0(2026-06-21)으로 감소했다.
+    evaluation_methodology가 ACTIVE_PASSIVE_SPLIT_V1_INCONCLUSIVE_EXCLUDED로 변경되며
+    inconclusive/replay 표본이 제외된 것으로 추정 — 근본 원인은 별도 조사 필요(WBS-7.2 잔여 항목).
+  calibration_registry_total_thresholds: 190  # spec/calibration_registry.yaml 직접 집계 (구문서의 70은 stale)
+  calibration_registry_expert_prior_count: 59
+  calibration_registry_calibrated_count: 0
+  rule: "이 문서를 인용할 때는 항상 as_of_date를 동반 표기하고, 아래 5/30 스냅샷을 '현재'로 인용하지 않는다."
+
+# ════════════════════════════════════════════════════════════════════════════
+# 과거 달성 현황 (2026-05-30, 역사적 스냅샷 — "현재"로 인용 금지)
 # ════════════════════════════════════════════════════════════════════════════
 current_status_2026_05_30:
   phase_1_bch: COMPLETE
@@ -489,3 +512,17 @@ current_status_2026_05_30:
   cases_analyzed: 141
   miss5_count: 51
   next_milestone: "cases_analyzed=30 달성 후 ALEG_V2_GATE1_BLOCK_PCT 보정 심사"
+  automation_entrypoints:
+    gitea_schedule: ".gitea/workflows/calibration_backlog.yml"
+    npm_script: "npm run ops:calibration-backlog"
+    generated_artifacts:
+      - Temp/calibration_priority_v1.json
+      - Temp/calibration_change_ledger_v4.json
+      - Temp/calibration_review_report_v1.json
+      - Temp/calibration_review_report_v1.md
+      - Temp/calibration_approval_list_v1.json
+      - Temp/calibration_approval_list_v1.md
+      - Temp/calibration_registry_v1.json
+    promotion_rules:
+      provisional: "sample_n >= 10 AND direction confirmed AND change_ledger entry exists"
+      calibrated: "sample_n >= 30 AND backtest_doc exists AND validator overclaimed_count == 0"

spec/calibration_registry.yaml

@@ -1,3 +1,7 @@
+has_code_implementation: true
+code_path:
+  - "tools/build_calibration_priority_v1.py"
+  - "tools/validate_calibration_registry_v1.py"
 thresholds:
 - id: ALEG_V2_GATE1_BLOCK_PCT
   value: 3.0
@@ -913,7 +917,7 @@ thresholds:
   notes: '이벤트 충격 방어: 20% 고정. KOSPI 비중 제공 시 max(20, weight×0.60).'
   live_sample_requirement: 30
   sunset_date: '2026-09-30'
-- id: SEMI_CLUSTER_CAP_RISK_OFF
+- id: SEMI_CLUSTER_CAP_RISK_OFF_MWA
   value: 25.0
   unit: pct
   source: EXPERT_PRIOR
@@ -921,7 +925,12 @@ thresholds:
   last_calibrated: null
   owner_formula: MARKET_WEIGHT_AWARE_CLUSTER_GATE_V1
   gs_location: gas_data_feed.gs:3858
-  notes: '하락장: 25%. KOSPI 비중 제공 시 max(25, weight×0.80).'
+  notes: >
+    하락장: 25%. KOSPI 비중 제공 시 max(25, weight×0.80).
+    WBS-7.1(2026-06-21): 원래 id가 SEMI_CLUSTER_CAP_RISK_OFF였으나
+    SEMICONDUCTOR_CLUSTER_GATE_V1 소유의 동명 entry(value=20.0)와 id가 충돌해
+    dict 기반 조회 시 한쪽이 조용히 무시되는 버그가 있었다. 외부 참조 0건 확인 후
+    이 entry(MARKET_WEIGHT_AWARE_CLUSTER_GATE_V1 소유)만 _MWA suffix로 분리했다.
   live_sample_requirement: 30
   sunset_date: '2026-09-30'
 - id: SEMI_CLUSTER_CAP_NEUTRAL
@@ -1803,6 +1812,22 @@ thresholds:
   gs_location: gas_data_feed.gs:2164
   notes: Base take-profit score used in profit-lock computation. Migrated from GAS SP constant to registry (P5-T01 wave2).
 
+- id: OVERHANG_PRESSURE_V1_FALLBACK_MULT
+  value: 1.5
+  unit: multiplier_of_avg_volume_5d
+  source: EXPERT_PRIOR
+  sample_n: 0
+  last_calibrated: null
+  owner_formula: OVERHANG_PRESSURE_V1
+  py_location: spec/13_formula_registry.yaml:OVERHANG_PRESSURE_V1.derived_flags.selling_acceleration.without_20d_fallback
+  notes: >
+    WBS-7.5(2026-06-21) — frg_20d_sh 미존재 시 selling_acceleration 폴백을
+    "frg_5d_sh < -500000"(절대 주식수, 임시) 에서 "frg_5d_sh < -1.5 * avg_volume_5d"
+    (해당 종목 평균거래량 비례) 로 교체. 1.5 배수는 with_20d 분기에서 동일 공식이
+    이미 사용하는 가속 임계(frg_20d_sh/4 × 1.5)를 그대로 재사용한 것이며, 새로
+    추정한 값이 아니다. 단, 실거래 표본으로 검증되지 않았으므로 EXPERT_PRIOR로
+    등록한다 — CALIBRATED 승격은 sample_n≥30 확보 후 검토.
+
 calibration_policy:
   honest_disclosure_required: true
   overclaimed_calibration_definition: 'source=CALIBRATED 이면서 sample_n < 30 → OVERCLAIMED_CALIBRATION.

tests/unit/test_calibration_priority_v1.py

@@ -0,0 +1,107 @@
+from __future__ import annotations
+
+import json
+import subprocess
+import sys
+from pathlib import Path
+
+
+ROOT = Path(__file__).resolve().parents[2]
+
+
+def _run(script: str) -> None:
+    subprocess.run(
+        [sys.executable, script],
+        cwd=ROOT,
+        check=True,
+        capture_output=True,
+        text=True,
+        encoding="utf-8",
+    )
+
+
+def test_build_calibration_priority_and_change_ledger(tmp_path):
+    _run("tools/build_calibration_priority_v1.py")
+    _run("tools/build_calibration_change_ledger_v4.py")
+    _run("tools/validate_calibration_change_ledger_v1.py")
+
+    priority_path = ROOT / "Temp" / "calibration_priority_v1.json"
+    ledger_path = ROOT / "Temp" / "calibration_change_ledger_v4.json"
+
+    priority = json.loads(priority_path.read_text(encoding="utf-8"))
+    ledger = json.loads(ledger_path.read_text(encoding="utf-8"))
+
+    assert priority["status"] == "CALIBRATION_PRIORITY_OK"
+    assert priority["priority_count"] >= 5
+    assert priority["priority_list"]
+    assert priority["priority_basis"] in {"alpha_feedback_loop_v2", "registry_warning_fallback"}
+
+    assert ledger["formula_id"] == "CALIBRATION_CHANGE_LEDGER_V4"
+    assert ledger["threshold_change_without_ledger_count"] == 0
+    assert len(ledger["changes"]) >= 5
+
+
+def test_calibration_backlog_workflow_and_script_exist():
+    workflow = ROOT / ".gitea" / "workflows" / "calibration_backlog.yml"
+    package = json.loads((ROOT / "package.json").read_text(encoding="utf-8"))
+    assert workflow.exists()
+    assert "ops:calibration-backlog" in package["scripts"]
+    assert "ops:calibration-review-report" in package["scripts"]
+    assert "ops:calibration-approval-list" in package["scripts"]
+    assert "ops:calibration-decision-draft" in package["scripts"]
+
+
+def test_build_calibration_review_report(tmp_path):
+    _run("tools/build_calibration_priority_v1.py")
+    _run("tools/build_calibration_change_ledger_v4.py")
+    _run("tools/build_calibration_review_report_v1.py")
+
+    report_json = ROOT / "Temp" / "calibration_review_report_v1.json"
+    report_md = ROOT / "Temp" / "calibration_review_report_v1.md"
+    payload = json.loads(report_json.read_text(encoding="utf-8"))
+    text = report_md.read_text(encoding="utf-8")
+
+    assert payload["formula_id"] == "CALIBRATION_REVIEW_REPORT_V1"
+    assert payload["summary"]["total_thresholds"] >= 1
+    assert payload["top_priority_rows"]
+    assert "Calibration Review Report" in text
+    assert "Review Candidates" in text
+
+
+def test_build_calibration_approval_list(tmp_path):
+    _run("tools/build_calibration_priority_v1.py")
+    _run("tools/build_calibration_change_ledger_v4.py")
+    _run("tools/build_calibration_review_report_v1.py")
+    _run("tools/build_calibration_approval_list_v1.py")
+
+    approval_json = ROOT / "Temp" / "calibration_approval_list_v1.json"
+    approval_md = ROOT / "Temp" / "calibration_approval_list_v1.md"
+    payload = json.loads(approval_json.read_text(encoding="utf-8"))
+    text = approval_md.read_text(encoding="utf-8")
+
+    assert payload["formula_id"] == "CALIBRATION_APPROVAL_LIST_V1"
+    assert payload["approval_candidate_count"] >= 1
+    assert payload["approval_candidates"]
+    assert "Calibration Approval List" in text
+    assert "Approval Candidates" in text
+
+
+def test_build_calibration_decision_draft(tmp_path):
+    _run("tools/build_calibration_priority_v1.py")
+    _run("tools/build_calibration_change_ledger_v4.py")
+    _run("tools/build_calibration_review_report_v1.py")
+    _run("tools/build_calibration_approval_list_v1.py")
+    _run("tools/build_calibration_decision_draft_v1.py")
+
+    decision_json = ROOT / "Temp" / "calibration_decision_draft_v1.json"
+    decision_md = ROOT / "Temp" / "calibration_decision_draft_v1.md"
+    payload = json.loads(decision_json.read_text(encoding="utf-8"))
+    text = decision_md.read_text(encoding="utf-8")
+
+    assert payload["formula_id"] == "CALIBRATION_DECISION_DRAFT_V1"
+    assert payload["decision_count"] >= 1
+    assert payload["summary"]["APPROVE"] >= 1
+    assert payload["summary"]["HOLD"] >= 1
+    assert payload["summary"]["REJECT"] >= 0
+    assert "Calibration Decision Draft" in text
+    assert "Decision Table" in text

tools/build_calibration_approval_list_v1.py

@@ -0,0 +1,136 @@
+#!/usr/bin/env python3
+"""
+build_calibration_approval_list_v1.py
+───────────────────────────────────────────────────────────────────────────────
+calibration_review_report_v1.json을 읽어 PROVISIONAL 승격 승인 리스트를 만든다.
+
+목적:
+  - source=PROVISIONAL 인 임계값을 별도 승인 대상 리스트로 분리
+  - reviewer가 바로 볼 수 있는 Markdown/JSON 산출물 생성
+  - PROVISIONAL 승격과 provisional review를 분리해 운영 책임을 명확화
+
+출력:
+  Temp/calibration_approval_list_v1.json
+  Temp/calibration_approval_list_v1.md
+
+사용법:
+  python tools/build_calibration_approval_list_v1.py
+"""
+
+from __future__ import annotations
+
+import json
+import sys
+from datetime import datetime, timezone
+from pathlib import Path
+from typing import Any
+
+ROOT = Path(__file__).resolve().parent.parent
+REVIEW = ROOT / "Temp" / "calibration_review_report_v1.json"
+OUT_JSON = ROOT / "Temp" / "calibration_approval_list_v1.json"
+OUT_MD = ROOT / "Temp" / "calibration_approval_list_v1.md"
+
+if sys.stdout.encoding and sys.stdout.encoding.lower() not in ("utf-8", "utf8"):
+    sys.stdout = open(sys.stdout.fileno(), mode="w", encoding="utf-8", buffering=1)
+
+
+def _load_json(path: Path) -> dict[str, Any]:
+    if not path.exists():
+        return {}
+    try:
+        data = json.loads(path.read_text(encoding="utf-8"))
+    except Exception:
+        return {}
+    return data if isinstance(data, dict) else {}
+
+
+def _table(rows: list[dict[str, Any]], keys: list[str], max_rows: int = 25) -> str:
+    if not rows:
+        return "_데이터 없음_"
+    header = "| " + " | ".join(keys) + " |"
+    sep = "| " + " | ".join(["---"] * len(keys)) + " |"
+    body = []
+    for row in rows[:max_rows]:
+        body.append("| " + " | ".join(str(row.get(k, "")).replace("|", "ㅣ") for k in keys) + " |")
+    suffix = f"\n\n_...총 {len(rows)}행 중 {max_rows}행 표시_" if len(rows) > max_rows else ""
+    return "\n".join([header, sep, *body]) + suffix
+
+
+def main() -> int:
+    review = _load_json(REVIEW)
+    rows = review.get("review_rows") if isinstance(review.get("review_rows"), list) else []
+
+    approval_candidates: list[dict[str, Any]] = []
+    provisional_review_candidates: list[dict[str, Any]] = []
+
+    for row in rows:
+        if not isinstance(row, dict):
+            continue
+        source = str(row.get("source") or "")
+        readiness = str(row.get("readiness") or "")
+        sample_n = int(row.get("sample_n") or 0)
+        base = {
+            "id": row.get("id", ""),
+            "source": source,
+            "sample_n": sample_n,
+            "value": row.get("value"),
+            "unit": row.get("unit", ""),
+            "owner_formula": row.get("owner_formula", ""),
+            "readiness": readiness,
+            "reason": row.get("reason", ""),
+        }
+        if source == "PROVISIONAL":
+            approval_candidates.append(base)
+        elif readiness == "PROVISIONAL_CANDIDATE":
+            provisional_review_candidates.append(base)
+
+    approval_candidates.sort(key=lambda item: (-int(item.get("sample_n") or 0), str(item.get("id") or "")))
+    provisional_review_candidates.sort(key=lambda item: (-int(item.get("sample_n") or 0), str(item.get("id") or "")))
+
+    report = {
+        "formula_id": "CALIBRATION_APPROVAL_LIST_V1",
+        "generated_at": datetime.now(timezone.utc).isoformat(),
+        "review_report_path": str(REVIEW),
+        "approval_candidate_count": len(approval_candidates),
+        "provisional_review_candidate_count": len(provisional_review_candidates),
+        "approval_candidates": approval_candidates,
+        "provisional_review_candidates": provisional_review_candidates,
+    }
+
+    OUT_JSON.write_text(json.dumps(report, ensure_ascii=False, indent=2), encoding="utf-8")
+
+    md_lines = [
+        "# Calibration Approval List",
+        "",
+        "## Summary",
+        "",
+        f"- approval candidates: {len(approval_candidates)}",
+        f"- provisional review candidates: {len(provisional_review_candidates)}",
+        "",
+        "## Approval Candidates",
+        "",
+        _table(approval_candidates, ["id", "source", "sample_n", "value", "unit", "owner_formula", "readiness", "reason"]),
+        "",
+        "## Provisional Review Candidates",
+        "",
+        _table(provisional_review_candidates, ["id", "source", "sample_n", "value", "unit", "owner_formula", "readiness", "reason"]),
+        "",
+        "## Evidence",
+        "",
+        f"- review report: {REVIEW}",
+    ]
+    OUT_MD.write_text("\n".join(md_lines), encoding="utf-8")
+
+    print(json.dumps({
+        "formula_id": report["formula_id"],
+        "gate": "PASS" if approval_candidates else "WARN",
+        "approval_candidate_count": len(approval_candidates),
+        "provisional_review_candidate_count": len(provisional_review_candidates),
+        "json_path": str(OUT_JSON),
+        "md_path": str(OUT_MD),
+    }, ensure_ascii=False, indent=2))
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())

tools/build_calibration_decision_draft_v1.py

@@ -0,0 +1,152 @@
+#!/usr/bin/env python3
+"""
+build_calibration_decision_draft_v1.py
+───────────────────────────────────────────────────────────────────────────────
+calibration_review_report_v1.json / calibration_approval_list_v1.json을 바탕으로
+운영 승인 초안(APPROVE / HOLD / REJECT)을 만든다.
+
+목적:
+  - 사람 검토 전 단계에서 결정 초안을 자동 생성
+  - source=PROVISIONAL은 원칙적으로 APPROVE
+  - PROVISIONAL_CANDIDATE는 HOLD
+  - 나머지는 REJECT 또는 HOLD로 사유를 명시
+
+출력:
+  Temp/calibration_decision_draft_v1.json
+  Temp/calibration_decision_draft_v1.md
+
+사용법:
+  python tools/build_calibration_decision_draft_v1.py
+"""
+
+from __future__ import annotations
+
+import json
+import sys
+from datetime import datetime, timezone
+from pathlib import Path
+from typing import Any
+
+ROOT = Path(__file__).resolve().parent.parent
+REVIEW = ROOT / "Temp" / "calibration_review_report_v1.json"
+APPROVAL = ROOT / "Temp" / "calibration_approval_list_v1.json"
+OUT_JSON = ROOT / "Temp" / "calibration_decision_draft_v1.json"
+OUT_MD = ROOT / "Temp" / "calibration_decision_draft_v1.md"
+
+if sys.stdout.encoding and sys.stdout.encoding.lower() not in ("utf-8", "utf8"):
+    sys.stdout = open(sys.stdout.fileno(), mode="w", encoding="utf-8", buffering=1)
+
+
+def _load_json(path: Path) -> dict[str, Any]:
+    if not path.exists():
+        return {}
+    try:
+        data = json.loads(path.read_text(encoding="utf-8"))
+    except Exception:
+        return {}
+    return data if isinstance(data, dict) else {}
+
+
+def _table(rows: list[dict[str, Any]], keys: list[str], max_rows: int = 25) -> str:
+    if not rows:
+        return "_데이터 없음_"
+    header = "| " + " | ".join(keys) + " |"
+    sep = "| " + " | ".join(["---"] * len(keys)) + " |"
+    body = []
+    for row in rows[:max_rows]:
+        body.append("| " + " | ".join(str(row.get(k, "")).replace("|", "ㅣ") for k in keys) + " |")
+    suffix = f"\n\n_...총 {len(rows)}행 중 {max_rows}행 표시_" if len(rows) > max_rows else ""
+    return "\n".join([header, sep, *body]) + suffix
+
+
+def _decide(row: dict[str, Any]) -> tuple[str, str]:
+    source = str(row.get("source") or "")
+    readiness = str(row.get("readiness") or "")
+    sample_n = int(row.get("sample_n") or 0)
+    if source == "PROVISIONAL" and sample_n >= 30:
+        return "APPROVE", "source=PROVISIONAL and sample_n>=30"
+    if source == "PROVISIONAL":
+        return "APPROVE", "source=PROVISIONAL"
+    if readiness == "PROVISIONAL_CANDIDATE":
+        return "HOLD", "Needs provisional review"
+    if sample_n >= 10:
+        return "HOLD", "Sample present but not provisional"
+    return "REJECT", "Insufficient evidence"
+
+
+def main() -> int:
+    review = _load_json(REVIEW)
+    approval = _load_json(APPROVAL)
+
+    review_rows = review.get("review_rows") if isinstance(review.get("review_rows"), list) else []
+    decisions: list[dict[str, Any]] = []
+    summary = {"APPROVE": 0, "HOLD": 0, "REJECT": 0}
+
+    for row in review_rows:
+        if not isinstance(row, dict):
+            continue
+        decision, reason = _decide(row)
+        item = {
+            "id": row.get("id", ""),
+            "source": row.get("source", ""),
+            "sample_n": int(row.get("sample_n") or 0),
+            "value": row.get("value"),
+            "unit": row.get("unit", ""),
+            "owner_formula": row.get("owner_formula", ""),
+            "readiness": row.get("readiness", ""),
+            "decision": decision,
+            "reason": reason,
+        }
+        decisions.append(item)
+        summary[decision] += 1
+
+    decisions.sort(key=lambda item: ({"APPROVE": 0, "HOLD": 1, "REJECT": 2}.get(str(item.get("decision") or ""), 3), -int(item.get("sample_n") or 0), str(item.get("id") or "")))
+
+    report = {
+        "formula_id": "CALIBRATION_DECISION_DRAFT_V1",
+        "generated_at": datetime.now(timezone.utc).isoformat(),
+        "review_report_path": str(REVIEW),
+        "approval_list_path": str(APPROVAL),
+        "summary": summary,
+        "decision_count": len(decisions),
+        "decisions": decisions,
+        "approval_candidate_count": int(approval.get("approval_candidate_count") or 0),
+    }
+
+    OUT_JSON.write_text(json.dumps(report, ensure_ascii=False, indent=2), encoding="utf-8")
+
+    md_lines = [
+        "# Calibration Decision Draft",
+        "",
+        "## Summary",
+        "",
+        f"- APPROVE: {summary['APPROVE']}",
+        f"- HOLD: {summary['HOLD']}",
+        f"- REJECT: {summary['REJECT']}",
+        f"- decision_count: {len(decisions)}",
+        "",
+        "## Decision Table",
+        "",
+        _table(decisions, ["id", "source", "sample_n", "decision", "reason", "owner_formula", "readiness"]),
+        "",
+        "## Evidence",
+        "",
+        f"- review report: {REVIEW}",
+        f"- approval list: {APPROVAL}",
+    ]
+    OUT_MD.write_text("\n".join(md_lines), encoding="utf-8")
+
+    print(json.dumps({
+        "formula_id": report["formula_id"],
+        "gate": "PASS" if summary["APPROVE"] else "WARN",
+        "approve_count": summary["APPROVE"],
+        "hold_count": summary["HOLD"],
+        "reject_count": summary["REJECT"],
+        "json_path": str(OUT_JSON),
+        "md_path": str(OUT_MD),
+    }, ensure_ascii=False, indent=2))
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())

tools/build_calibration_priority_v1.py

@@ -29,6 +29,41 @@ ROOT   = Path(__file__).resolve().parent.parent
 AFL    = ROOT / "Temp" / "alpha_feedback_loop_v2.json"
 REG    = ROOT / "spec" / "calibration_registry.yaml"
 OUTPUT = ROOT / "Temp" / "calibration_priority_v1.json"
+PREDICTION_ACCURACY = ROOT / "Temp" / "prediction_accuracy_harness_v2.json"
+
+
+def registry_source_breakdown(reg_index: dict[str, dict]) -> dict:
+    """WBS-7.1(2026-06-21) — calibration_registry.yaml 전체의 source별 분포를 매 실행마다
+    집계해 'CALIBRATED 비율이 실제로 몇 %인가'를 사람이 grep으로 직접 세지 않아도
+    항상 최신 상태로 노출한다(2026-06-21 비판적 리뷰 0c절에서 0/190 발견 당시 수동 집계 필요했던 문제 해소)."""
+    counts: dict[str, int] = {"SPEC_DERIVED": 0, "EXPERT_PRIOR": 0, "PROVISIONAL": 0, "CALIBRATED": 0}
+    for entry in reg_index.values():
+        source = str(entry.get("source", "")).upper()
+        if source in counts:
+            counts[source] += 1
+    total = sum(counts.values())
+    return {
+        "total_thresholds": total,
+        "counts": counts,
+        "calibrated_pct": round(100.0 * counts["CALIBRATED"] / total, 2) if total else 0.0,
+        "unvalidated_pct": round(100.0 * (counts["SPEC_DERIVED"] + counts["EXPERT_PRIOR"]) / total, 2) if total else 0.0,
+    }
+
+
+def live_t5_status() -> dict:
+    """WBS-7.2/7.1(2026-06-21) — T+5 수치를 하드코딩하지 않고 항상 최신 산출물에서 읽는다.
+    Temp/prediction_accuracy_harness_v2.json이 없거나 sample=0이면 정직하게 DATA_GATED로 보고한다."""
+    if not PREDICTION_ACCURACY.exists():
+        return {"status": "ARTIFACT_MISSING", "t5_sample": 0, "t5_match_rate_pct": None}
+    data = load_json(PREDICTION_ACCURACY)
+    t5_sample = int(data.get("t5_sample") or 0)
+    t5_rate = data.get("t5_op_rate")
+    return {
+        "status": "DATA_GATED" if t5_sample == 0 else "OK",
+        "as_of_date": data.get("as_of_date"),
+        "t5_sample": t5_sample,
+        "t5_match_rate_pct": t5_rate,
+    }
 
 if sys.stdout.encoding and sys.stdout.encoding.lower() not in ("utf-8", "utf8"):
     sys.stdout = open(sys.stdout.fileno(), mode="w", encoding="utf-8", buffering=1)
@@ -90,6 +125,42 @@ def load_registry(p: Path) -> dict[str, dict]:
     return {t["id"]: t for t in data.get("thresholds", []) if "id" in t}
 
 
+def _priority_from_registry_entry(entry: dict, source_tag: str, urgency_bias: int) -> dict:
+    sample_n = int(entry.get("sample_n", 0) or 0)
+    source = str(entry.get("source", "EXPERT_PRIOR"))
+    threshold_class = str(entry.get("threshold_class", "standard"))
+    urgency = urgency_bias
+    if source == "EXPERT_PRIOR":
+        urgency += 10
+    if source == "PROVISIONAL":
+        urgency += 20
+    if threshold_class == "live_critical":
+        urgency += 15
+    if sample_n == 0:
+        urgency += 5
+    if sample_n > 0:
+        urgency += max(0, 30 - sample_n)
+    return {
+        "calibration_id": entry.get("id", ""),
+        "current_value": entry.get("value"),
+        "owner_formula": entry.get("owner_formula", ""),
+        "source": source,
+        "sample_n": sample_n,
+        "linked_factor": source_tag,
+        "alpha_action": "registry_review",
+        "urgency_score": urgency,
+        "calibration_path": (
+            (
+                "표본 30건 이상 확보 후 PROVISIONAL 승격 → "
+                if sample_n >= 30
+                else f"표본 {30 - sample_n}건 추가 수집 후 PROVISIONAL 승격 → "
+            )
+            + "실측 T+5 승률 기반 최적값 backtest → CALIBRATED 확정"
+        ),
+        "rationale": f"source={source}, class={threshold_class}, sample_n={sample_n}",
+    }
+
+
 def main() -> int:
     afl_data  = load_json(AFL)
     reg_index = load_registry(REG)
@@ -112,48 +183,32 @@ def main() -> int:
     priority_list: list[dict] = []
 
     for adj in adjustments:
-        factor   = adj.get("factor", "")
-        action   = adj.get("action", "")
-        rationale = adj.get("rationale", "")
-        reg_ids  = FACTOR_TO_REGISTRY.get(factor, [])
+        factor = str(adj.get("factor", ""))
+        action = str(adj.get("action", ""))
+        rationale = str(adj.get("rationale", ""))
+        reg_ids = FACTOR_TO_REGISTRY.get(factor, [])
 
         for rid in reg_ids:
             reg_entry = reg_index.get(rid)
             if not reg_entry:
                 continue
-            source   = reg_entry.get("source", "EXPERT_PRIOR")
-            sample_n = int(reg_entry.get("sample_n", 0) or 0)
-            value    = reg_entry.get("value")
-            formula  = reg_entry.get("owner_formula", "")
+            item = _priority_from_registry_entry(reg_entry, factor, miss5_count if factor == "passive_signal_quality" else 0)
+            item["alpha_action"] = action or "feedback_review"
+            if rationale:
+                item["rationale"] = rationale[:200]
+            priority_list.append(item)
 
-            # 보정 우선도 점수: miss5_count 기여 + 미보정 가중
-            urgency = 0
-            if factor == "passive_signal_quality":
-                urgency += miss5_count  # miss가 많을수록 높은 urgency
-            if source == "EXPERT_PRIOR":
-                urgency += 10
-            if sample_n == 0:
-                urgency += 5
-
-            priority_list.append({
-                "calibration_id":   rid,
-                "current_value":    value,
-                "owner_formula":    formula,
-                "source":           source,
-                "sample_n":         sample_n,
-                "linked_factor":    factor,
-                "alpha_action":     action,
-                "urgency_score":    urgency,
-                "calibration_path": (
-                    (
-                        "표본 30건 이상 확보 후 PROVISIONAL 승격 → "
-                        if sample_n >= 30
-                        else f"표본 {30 - sample_n}건 추가 수집 후 PROVISIONAL 승격 → "
-                    )
-                    + "실측 T+5 승률 기반 최적값 backtest → CALIBRATED 확정"
-                ),
-                "rationale":        rationale[:200] if rationale else "",
-            })
+    if not priority_list:
+        # alpha_feedback_loop가 비어 있어도 registry 자체의 보정 debt를 추적할 수 있게 한다.
+        for reg_id, reg_entry in reg_index.items():
+            source = str(reg_entry.get("source", "EXPERT_PRIOR"))
+            if source not in {"EXPERT_PRIOR", "PROVISIONAL"}:
+                continue
+            tag = f"registry:{source.lower()}"
+            item = _priority_from_registry_entry(reg_entry, tag, 0)
+            if source == "PROVISIONAL":
+                item["urgency_score"] += 5
+            priority_list.append(item)
 
     # 중복 제거 (같은 rid, 높은 urgency 유지)
     seen: dict[str, dict] = {}
@@ -177,7 +232,19 @@ def main() -> int:
     print(f"  Step 2 (30건 후): ALEG_V2_GATE1_BLOCK_PCT 3.0% → 실측 최적값으로 PROVISIONAL 승격")
     print(f"  Step 3 (50건 후): DSD_V1 가중치 logistic regression 최적화")
     print(f"  Step 4 (100건 후): K2_SPLIT_RATIO backtest 비교 → CALIBRATED 확정")
-    print(f"  miss5_count={miss5_count}건 → passive_signal_quality 개선이 T+5 35.86%→50%+ 핵심")
+    registry_health = registry_source_breakdown(reg_index)
+    t5_status = live_t5_status()
+
+    print(f"\n  [캘리브레이션 레지스트리 건강도] (WBS-7.1)")
+    print(f"  total={registry_health['total_thresholds']} {registry_health['counts']}")
+    print(f"  CALIBRATED={registry_health['calibrated_pct']}%  미검증(SPEC_DERIVED+EXPERT_PRIOR)={registry_health['unvalidated_pct']}%")
+
+    if t5_status["status"] == "DATA_GATED":
+        print(f"  miss5_count={miss5_count}건 → T+5 현재 DATA_GATED(sample=0) — passive_signal_quality 개선 영향은 표본 누적 후 측정 가능")
+    elif t5_status["status"] == "ARTIFACT_MISSING":
+        print(f"  miss5_count={miss5_count}건 → T+5 산출물 없음(Temp/prediction_accuracy_harness_v2.json) — 먼저 생성 필요")
+    else:
+        print(f"  miss5_count={miss5_count}건 → T+5={t5_status['t5_match_rate_pct']}% (as_of={t5_status.get('as_of_date')}) → passive_signal_quality 개선 핵심")
 
     result = {
         "status": "CALIBRATION_PRIORITY_OK",
@@ -191,10 +258,14 @@ def main() -> int:
             "step3": "50건 후: DSD_V1 가중치 logistic regression 최적화",
             "step4": "100건 후: K2_SPLIT_RATIO 30/70~60/40 backtest → CALIBRATED",
         },
+        "priority_basis": "alpha_feedback_loop_v2" if adjustments else "registry_warning_fallback",
+        "registry_health": registry_health,
         "target_improvement": {
-            "current_t5_pct": 35.86,
+            "t5_status": t5_status["status"],
+            "current_t5_pct": t5_status["t5_match_rate_pct"],
+            "t5_as_of_date": t5_status.get("as_of_date"),
             "target_t5_pct": 55.0,
-            "key_lever": "passive_signal_quality (miss5_count=51건 개선)",
+            "key_lever": f"passive_signal_quality (miss5_count={miss5_count}건 개선)",
         },
     }
 

tools/build_calibration_review_report_v1.py

@@ -0,0 +1,205 @@
+#!/usr/bin/env python3
+"""
+build_calibration_review_report_v1.py
+───────────────────────────────────────────────────────────────────────────────
+calibration_registry.yaml + calibration_priority_v1.json + calibration_change_ledger_v4.json
+을 묶어 운영용 보정 리뷰 리포트를 만든다.
+
+목적:
+  - PROVISIONAL / CALIBRATED 승격 후보를 사람이 읽을 수 있게 정리
+  - registry warning fallback 상태를 숨기지 않고 그대로 공시
+  - 월간 보정 운영에서 바로 참고 가능한 Markdown + JSON 산출물 생성
+
+출력:
+  Temp/calibration_review_report_v1.json
+  Temp/calibration_review_report_v1.md
+
+사용법:
+  python tools/build_calibration_review_report_v1.py
+"""
+
+from __future__ import annotations
+
+import json
+import sys
+from datetime import datetime, timezone
+from pathlib import Path
+from typing import Any
+
+import yaml
+
+ROOT = Path(__file__).resolve().parent.parent
+REGISTRY = ROOT / "spec" / "calibration_registry.yaml"
+PRIORITY = ROOT / "Temp" / "calibration_priority_v1.json"
+LEDGER = ROOT / "Temp" / "calibration_change_ledger_v4.json"
+OUT_JSON = ROOT / "Temp" / "calibration_review_report_v1.json"
+OUT_MD = ROOT / "Temp" / "calibration_review_report_v1.md"
+
+if sys.stdout.encoding and sys.stdout.encoding.lower() not in ("utf-8", "utf8"):
+    sys.stdout = open(sys.stdout.fileno(), mode="w", encoding="utf-8", buffering=1)
+
+
+def _load_json(path: Path) -> dict[str, Any]:
+    if not path.exists():
+        return {}
+    try:
+        data = json.loads(path.read_text(encoding="utf-8"))
+    except Exception:
+        return {}
+    return data if isinstance(data, dict) else {}
+
+
+def _load_registry(path: Path) -> list[dict[str, Any]]:
+    if not path.exists():
+        return []
+    data = yaml.safe_load(path.read_text(encoding="utf-8")) or {}
+    thresholds = data.get("thresholds", [])
+    return [t for t in thresholds if isinstance(t, dict)]
+
+
+def _readiness(entry: dict[str, Any]) -> tuple[str, str]:
+    source = str(entry.get("source") or "EXPERT_PRIOR")
+    sample_n = int(entry.get("sample_n") or 0)
+    if source == "CALIBRATED":
+        return "CALIBRATED", "Already calibrated"
+    if source == "PROVISIONAL" and sample_n >= 30:
+        return "CALIBRATION_READY", "Ready for calibrated review"
+    if source == "PROVISIONAL":
+        return "PROVISIONAL_ACTIVE", "Provisional with live samples"
+    if sample_n >= 10:
+        return "PROVISIONAL_CANDIDATE", "Candidate for provisional review"
+    return "WATCH", "Keep under watch"
+
+
+def _table(rows: list[dict[str, Any]], keys: list[str], max_rows: int = 25) -> str:
+    if not rows:
+        return "_데이터 없음_"
+    header = "| " + " | ".join(keys) + " |"
+    sep = "| " + " | ".join(["---"] * len(keys)) + " |"
+    body = []
+    for row in rows[:max_rows]:
+        body.append("| " + " | ".join(str(row.get(k, "")).replace("|", "ㅣ") for k in keys) + " |")
+    suffix = f"\n\n_...총 {len(rows)}행 중 {max_rows}행 표시_" if len(rows) > max_rows else ""
+    return "\n".join([header, sep, *body]) + suffix
+
+
+def main() -> int:
+    registry = _load_registry(REGISTRY)
+    priority = _load_json(PRIORITY)
+    ledger = _load_json(LEDGER)
+
+    source_counts: dict[str, int] = {}
+    readiness_counts: dict[str, int] = {}
+    reviewed_rows: list[dict[str, Any]] = []
+
+    for entry in registry:
+        source = str(entry.get("source") or "EXPERT_PRIOR")
+        source_counts[source] = source_counts.get(source, 0) + 1
+        readiness, reason = _readiness(entry)
+        readiness_counts[readiness] = readiness_counts.get(readiness, 0) + 1
+        if readiness in {"PROVISIONAL_CANDIDATE", "CALIBRATION_READY", "PROVISIONAL_ACTIVE"}:
+            reviewed_rows.append(
+                {
+                    "id": entry.get("id", ""),
+                    "source": source,
+                    "sample_n": int(entry.get("sample_n") or 0),
+                    "value": entry.get("value"),
+                    "unit": entry.get("unit", ""),
+                    "owner_formula": entry.get("owner_formula", ""),
+                    "readiness": readiness,
+                    "reason": reason,
+                    "notes": str(entry.get("notes") or "")[:120],
+                }
+            )
+
+    priority_list = priority.get("priority_list") if isinstance(priority.get("priority_list"), list) else []
+    priority_rows = []
+    for item in priority_list[:20]:
+        if not isinstance(item, dict):
+            continue
+        priority_rows.append(
+            {
+                "calibration_id": item.get("calibration_id", ""),
+                "source": item.get("source", ""),
+                "sample_n": item.get("sample_n", 0),
+                "urgency_score": item.get("urgency_score", 0),
+                "linked_factor": item.get("linked_factor", ""),
+                "owner_formula": item.get("owner_formula", ""),
+            }
+        )
+
+    report = {
+        "formula_id": "CALIBRATION_REVIEW_REPORT_V1",
+        "generated_at": datetime.now(timezone.utc).isoformat(),
+        "registry_path": str(REGISTRY),
+        "priority_path": str(PRIORITY),
+        "ledger_path": str(LEDGER),
+        "summary": {
+            "total_thresholds": len(registry),
+            "source_counts": source_counts,
+            "readiness_counts": readiness_counts,
+            "priority_count": int(priority.get("priority_count") or len(priority_rows)),
+            "ledger_change_count": len(ledger.get("changes", [])) if isinstance(ledger.get("changes"), list) else 0,
+            "ledger_without_change_count": int(ledger.get("threshold_change_without_ledger_count") or 0),
+        },
+        "top_priority_rows": priority_rows,
+        "review_rows": reviewed_rows,
+    }
+
+    OUT_JSON.write_text(json.dumps(report, ensure_ascii=False, indent=2), encoding="utf-8")
+
+    md_lines = [
+        "# Calibration Review Report",
+        "",
+        "## Summary",
+        "",
+        f"- total thresholds: {report['summary']['total_thresholds']}",
+        f"- priority count: {report['summary']['priority_count']}",
+        f"- ledger change count: {report['summary']['ledger_change_count']}",
+        f"- ledger without change count: {report['summary']['ledger_without_change_count']}",
+        "",
+        "### Source Counts",
+        "",
+        _table(
+            [{"source": k, "count": v} for k, v in sorted(source_counts.items())],
+            ["source", "count"],
+            max_rows=50,
+        ),
+        "",
+        "### Readiness Counts",
+        "",
+        _table(
+            [{"readiness": k, "count": v} for k, v in sorted(readiness_counts.items())],
+            ["readiness", "count"],
+            max_rows=50,
+        ),
+        "",
+        "## Top Priority Rows",
+        "",
+        _table(priority_rows, ["calibration_id", "source", "sample_n", "urgency_score", "linked_factor", "owner_formula"]),
+        "",
+        "## Review Candidates",
+        "",
+        _table(reviewed_rows, ["id", "source", "sample_n", "value", "unit", "owner_formula", "readiness", "reason"]),
+        "",
+        "## Evidence",
+        "",
+        f"- registry: {REGISTRY}",
+        f"- priority: {PRIORITY}",
+        f"- ledger: {LEDGER}",
+    ]
+    OUT_MD.write_text("\n".join(md_lines), encoding="utf-8")
+
+    print(json.dumps({
+        "formula_id": report["formula_id"],
+        "gate": "PASS" if reviewed_rows or priority_rows else "WARN",
+        "review_rows": len(reviewed_rows),
+        "priority_rows": len(priority_rows),
+        "json_path": str(OUT_JSON),
+        "md_path": str(OUT_MD),
+    }, ensure_ascii=False, indent=2))
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())