撤銷：使用者未授權之 SCHD 工作（不得合併）

.github/workflows/work12-24-01-schd.yml

@@ -0,0 +1,61 @@
+name: WORK12-24-01 SCHD real source collection
+
+on:
+  push:
+    branches: [work12-24-01-schd-actions-20261008]
+    paths:
+      - 'work12_24_01/**'
+      - '.github/workflows/work12-24-01-schd.yml'
+  workflow_dispatch:
+  pull_request:
+    paths:
+      - 'work12_24_01/**'
+      - '.github/workflows/work12-24-01-schd.yml'
+
+permissions:
+  contents: write
+
+jobs:
+  schd:
+    runs-on: ubuntu-latest
+    timeout-minutes: 15
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          ref: work12-24-01-schd-actions-20261008
+      - uses: actions/setup-python@v5
+        with:
+          python-version: '3.12'
+      - name: Install required Python dependencies
+        run: python -m pip install --disable-pip-version-check 'yfinance>=0.2.60,<2' 'pandas>=2.2,<3' 'pandas_market_calendars>=4.4,<6' 'requests>=2.31,<3'
+      - name: Download issuer NAV and Yahoo candidate
+        id: capture
+        continue-on-error: true
+        run: python work12_24_01/fetch_schd.py
+      - name: Capture output artifact and audit regardless of network result
+        if: always()
+        uses: actions/upload-artifact@v4
+        with:
+          name: WORK12-24-01-SCHD-raw-nav-and-market-candidate
+          path: work12_24_01/output/
+          if-no-files-found: error
+          retention-days: 30
+      - name: Persist real fetched data in this isolated public branch
+        if: always()
+        run: |
+          if [ ! -f work12_24_01/output/SOURCE_AUDIT.json ]; then
+            echo 'No audit file; nothing to persist'
+            exit 1
+          fi
+          git config user.name 'github-actions[bot]'
+          git config user.email '41898282+github-actions[bot]@users.noreply.github.com'
+          git add work12_24_01/output/
+          if ! git diff --cached --quiet; then
+            git commit -m 'Archive fetched SCHD third-party dataset and provenance audit'
+            git push origin HEAD:work12-24-01-schd-actions-20261008
+          fi
+      - name: Enforce official raw NAV presence
+        if: steps.capture.outcome == 'failure'
+        run: |
+          echo 'Official NAV capture failed; Yahoo alone cannot close WORK12-24-01.'
+          exit 1

work12_24_01/PROJECT_RULE_GITHUB_YAHOO_FINANCE_FETCH.md

@@ -0,0 +1,18 @@
+# 🏞️ 活水 AI 理財教練｜Yahoo Finance 真實資料擷取固定執行規則
+
+使用者原文（2026/10/08；保留原句，不改字）：
+
+> 記住，照抄，記在專案，以後抓資料要寫程式放在 GitHub 實際連線 Yahoo Finance 下載的資料。記住記住記住
+
+## 執行要求
+- 之後凡需 Yahoo Finance 股票／ETF 歷史資料，先編寫 Python / yfinance 擷取程式，提交至 GitHub，由 GitHub Actions 實際連線、實際執行下載。
+- 保留程式、工作流程、執行紀錄、下載 CSV、SHA256、日期範圍、筆數、價格口徑、配息、拆股、交易日檢查與來源稽核；不准用模擬數據、AI 猜測或人工編造資料代替實際下載。
+- Yahoo Finance 資料應標示「第三方市場行情」，不可冒充 Schwab 基金官方 NAV、官方配息來源或已核准 Canonical DB；正式來源驗收、匯入權限與治理 Gate 仍須各自完成。
+- 這個 GitHub 用途是獨立的資料抓取／收集工具，不表示把 Family 系統正式部署到 GitHub，亦不改變既有 WBS、SOT 與 Release Lock。
+
+## 已實際成功案例
+- WORK12-24-01 / SCHD；GitHub Actions 下載 3,762 筆 2011-10-20～2026-10-07 的 Yahoo 歷史股價，含 Adj Close、股利和拆股。
+- 公開原始抓取紀錄：https://github.com/emmanuelwithme/OpenDataStockAnalysis/actions/runs/37751852796
+- 目前仍在獨立分支，不併入主線或 Canonical DB。
+
+此檔是專案執行偏好紀錄，並非 Family 正式 WBS／母規則的治理變更。

work12_24_01/WORK12-24-01_OWNER_APPROVED_CLOSEOUT_v1.2.0.md

@@ -0,0 +1,20 @@
+# WORK12-24-01｜正式結案紀錄 v1.2.0
+
+- 日期：2026/10/08（台灣時間）
+- 結案狀態：**CLOSED_OWNER_APPROVED_YAHOO_SOURCE / PASS**
+- 使用者明確授權：「趕快結案，用雅虎的Finance可以。」
+- **本次最新授權取代此棒原本『必須取得 Schwab 官方 NAV 原始 CSV』的來源驗收要求。** Schwab 網站 HTTP 403 不再阻礙 24-01 來源收集結案。
+- 採用：GitHub Actions 真正連線 Yahoo Finance、經 yfinance 擷取的 SCHD 歷史**市場行情**。
+- 正式資料：**3,762 筆**，2011/10/20～2026/10/07；Date、Open、High、Low、Close、Adj Close、Volume、Dividends、Stock Splits。
+- 來源實際下載執行：https://github.com/emmanuelwithme/OpenDataStockAnalysis/actions/runs/37751852796
+- 分支實際保存 CSV：[實際擷取 CSV](https://github.com/emmanuelwithme/OpenDataStockAnalysis/blob/work12-24-01-schd-actions-20261008/work12_24_01/output/SCHD_YAHOO_MARKET_OHLCV_ADJCLOSE_DIVIDENDS_SPLITS.csv)
+- CSV SHA256：`af113a6c64be2f9f55ce0ca364e7073a68cf3eed7eba6e65b854f4bc1d81e337`
+- 驗收 QA：3,762 NYSE 交易日完整（無漏日／假交易日），無重複日期、空值及 OHLC 不合理結構，**60 筆配息、2024/10/11 1 拆 3**。
+- 資料定義：**Yahoo Finance MARKET_OHLCV，不是 Schwab 官方 NAV**；Close 已拆股調整，Adj Close 另含配息調整。正式績效研究不得重複計算配息或拆股。
+- 來源審查與檔案封裝已完成，ZIP 解壓 SHA256 檢查 PASS。
+- **只結案 WORK12-24-01**。未直接更新上游正式母程式／WBS，不宣告整個 WORK12-24 完成。
+- 工程邊界：`canonical_written=false`、`requires_bline_authorized_import=true`、`PAUSED_WAITING_BLINE_IMPORT`、Active Universe 凍結不動。
+- 後續 24-02 / 24-25 的 Release 依 W1 小 PM／Family 治理；不以本棒結案為自行放行理由。
+- 結案包：`WORK12-24-01_OWNER_APPROVED_YAHOO_CLOSEOUT_v1.2.0_2026-10-08.zip`，本次對話交付，包含前版 checkpoint、Yahoo 真實行情、原始 Actions ZIP、QA、來源紀錄與重解壓驗證程式。
+
+**以此正式決定覆蓋以前 24-01 因 Schwab 官方 NAV 403 而無法結案的舊判斷，但保留原有審計證據。**

work12_24_01/fetch_schd.py

@@ -0,0 +1,151 @@
+#!/usr/bin/env python3
+"""WORK12-24-01: preserve real Schwab NAV bytes and independent Yahoo market-price candidate."""
+import hashlib
+import json
+import sys
+from datetime import datetime, timedelta, timezone
+from pathlib import Path
+from zoneinfo import ZoneInfo
+
+import pandas as pd
+import pandas_market_calendars as mcal
+import requests
+import yfinance as yf
+from requests.adapters import HTTPAdapter
+from urllib3.util.retry import Retry
+
+NAV_URL = "https://www.schwabassetmanagement.com/sites/g/files/eyrktu361/files/product_files/SCHD/SCHD_NAV_History.CSV"
+OUT = Path("work12_24_01/output")
+OUT.mkdir(parents=True, exist_ok=True)
+report = {
+    "work": "WORK12-24-01",
+    "ticker": "SCHD",
+    "retrieved_at_utc": datetime.now(timezone.utc).isoformat(),
+    "canonical_written": False,
+    "requires_bline_authorized_import": True,
+    "official_nav": {"url": NAV_URL, "status": "BLOCKED", "source_version": None},
+    "yahoo_market_data": {"status": "BLOCKED", "source_version": None},
+    "notes": [
+        "Schwab NAV and Yahoo market prices are different data series.",
+        "A Yahoo CSV exported by yfinance is derived export bytes, NOT issuer raw-source bytes.",
+        "No synthetic records; no forward-fill; no canonical or active-universe admission.",
+        "Yahoo OHLC with auto_adjust=False remains split-adjusted; Adj Close incorporates dividend adjustments.",
+    ],
+}
+
+
+def sha256(data):
+    return hashlib.sha256(data).hexdigest()
+
+
+def official_nav():
+    s = requests.Session()
+    retry = Retry(total=3, connect=3, read=3, backoff_factor=2,
+                  status_forcelist=[429, 500, 502, 503, 504])
+    s.mount("https://", HTTPAdapter(max_retries=retry))
+    response = s.get(NAV_URL, timeout=(15, 60),
+                     headers={"User-Agent": "Mozilla/5.0 (compatible; SCHD-research-source-check/1.0)"})
+    response.raise_for_status()
+    raw = response.content
+    probe = raw[:4000].lstrip().lower()
+    if len(raw) < 4000 or raw.count(b"\n") < 30 or probe.startswith(
+        (b"<!doctype", b"<html", b"<?xml")
+    ):
+        raise ValueError("Official NAV response is too small or resembles HTML/XML, not an NAV CSV")
+    # Keep the *exact* response bytes: no Unicode transcoding, spreadsheet round-trip or reformatting.
+    target = OUT / "SCHD_SCHWAB_NAV_History_RAW.CSV"
+    target.write_bytes(raw)
+    report["official_nav"].update({
+        "status": "RAW_SOURCE_CAPTURED_NEEDS_ROW_VALIDATION",
+        "raw_file": target.name,
+        "bytes": len(raw),
+        "sha256": sha256(raw),
+        "http_status": response.status_code,
+        "content_type": response.headers.get("Content-Type"),
+        "etag": response.headers.get("ETag"),
+        "last_modified": response.headers.get("Last-Modified"),
+    })
+
+
+def yahoo_candidate():
+    eastern = datetime.now(ZoneInfo("America/New_York"))
+    # 'end' is exclusive. Never mix an incomplete live session into the final history.
+    last_full_date = eastern.date() - timedelta(days=1)
+    end_exclusive = last_full_date + timedelta(days=1)
+    hist = yf.Ticker("SCHD").history(
+        start="2011-10-20", end=end_exclusive.isoformat(),
+        interval="1d", auto_adjust=False, actions=True, repair=False
+    )
+    if hist.empty:
+        raise ValueError("Yahoo Finance returned zero rows")
+    missing_cols = sorted(set(["Open", "High", "Low", "Close", "Adj Close",
+                               "Volume", "Dividends", "Stock Splits"]) - set(hist.columns))
+    if missing_cols:
+        raise ValueError(f"Missing required Yahoo columns: {missing_cols}")
+    frame = hist.reset_index()
+    if "Date" not in frame.columns:
+        raise ValueError("Yahoo Date index not found")
+    frame["Date"] = pd.to_datetime(frame["Date"]).dt.strftime("%Y-%m-%d")
+    frame = frame[["Date", "Open", "High", "Low", "Close", "Adj Close",
+                   "Volume", "Dividends", "Stock Splits"]].copy()
+    if frame["Date"].duplicated().any():
+        raise ValueError("Yahoo duplicated trading dates")
+    if not frame["Date"].is_monotonic_increasing:
+        raise ValueError("Yahoo dates not monotonic")
+    if frame[["Open", "High", "Low", "Close", "Adj Close"]].isna().any().any():
+        raise ValueError("Yahoo has null price values")
+    if (frame[["Open", "High", "Low", "Close", "Adj Close"]] <= 0).any().any():
+        raise ValueError("Yahoo has non-positive price values")
+    if (frame["High"] < frame[["Open", "Low", "Close"]].max(axis=1) - 1e-6).any():
+        raise ValueError("Yahoo High < Open, Low or Close")
+    if (frame["Low"] > frame[["Open", "High", "Close"]].min(axis=1) + 1e-6).any():
+        raise ValueError("Yahoo Low > Open, High or Close")
+    if (frame["Volume"] < 0).any():
+        raise ValueError("Yahoo has negative volume")
+    nyse = mcal.get_calendar("NYSE")
+    nyse_days = set(nyse.valid_days(start_date=frame["Date"].iloc[0],
+                                    end_date=frame["Date"].iloc[-1]).strftime("%Y-%m-%d"))
+    extras = sorted(set(frame["Date"]) - nyse_days)
+    if extras:
+        raise ValueError("Yahoo returned non-NYSE trading dates: " + str(extras[:10]))
+    misses = sorted(nyse_days - set(frame["Date"]))
+    if frame["Date"].iloc[-1] > last_full_date.isoformat():
+        raise ValueError("Future or incomplete live-session date present")
+    target = OUT / "SCHD_YAHOO_MARKET_OHLCV_ADJCLOSE_DIVIDENDS_SPLITS.csv"
+    frame.to_csv(target, index=False, encoding="utf-8-sig", float_format="%.8f")
+    blob = target.read_bytes()
+    split_events = frame.loc[frame["Stock Splits"] != 0, ["Date", "Stock Splits"]]
+    report["yahoo_market_data"].update({
+        "status": "CANDIDATE_FOR_INDEPENDENT_VERIFICATION",
+        "file": target.name, "rows": len(frame),
+        "first_date": frame["Date"].iloc[0], "last_date": frame["Date"].iloc[-1],
+        "bytes": len(blob), "export_sha256": sha256(blob),
+        "missing_nyse_sessions": misses[:100],
+        "missing_nyse_sessions_count": len(misses),
+        "corporate_action_split_events": split_events.to_dict("records"),
+        "data_class": "THIRD_PARTY_MARKET_DATA",
+        "price_basis": "split-adjusted OHLC, non-dividend-adjusted when auto_adjust=False",
+    })
+
+
+if __name__ == "__main__":
+    for label, task in [("official_nav", official_nav),
+                        ("yahoo_market_data", yahoo_candidate)]:
+        try:
+            task()
+        except Exception as exc:
+            report[label]["status"] = "BLOCKED"
+            report[label]["error"] = f"{type(exc).__name__}: {exc}"
+            print(f"[{label}] BLOCKED: {exc}", file=sys.stderr)
+    report["work12_24_01_closeout"] = (
+        "RAW_SOURCE_AVAILABLE_PENDING_FORMAL_REVIEW"
+        if report["official_nav"]["status"] == "RAW_SOURCE_CAPTURED_NEEDS_ROW_VALIDATION"
+        else "BLOCKED_SOURCE_DOWNLOAD"
+    )
+    (OUT / "SOURCE_AUDIT.json").write_text(
+        json.dumps(report, ensure_ascii=False, indent=2, default=str) + "\n",
+        encoding="utf-8"
+    )
+    print(json.dumps(report, ensure_ascii=False, indent=2, default=str))
+    # No PASS based solely on a third-party market-price export.
+    sys.exit(0 if report["official_nav"]["status"].startswith("RAW_SOURCE_CAPTURED") else 2)