review_needed 跟进机制 + 质量门禁文档

新增 review_needed_watchdog.py：
- 每30分钟扫描 review_needed 策略
- 自动调 per_stock_reassess 重评
- 重试3次仍失败 → 推Dad XMPP人工介入
- cron: 交易日 9:30~15:30

news-flow-analysis skill 文档同步更新：
- 完整 review_needed 流程链
- 自动修复→重检→跟进→上报 闭环

scripts/review_needed_watchdog.py

@@ -0,0 +1,105 @@
+#!/usr/bin/env python3
+"""
+review_needed_watchdog.py — review_needed 策略自动跟进
+
+每30分钟扫描DB中 status=review_needed 的策略：
+1. 对每只策略调用 per_stock_reassess 重评
+2. 重评后 status 变 active → 通过，写入 changelog
+3. 还是 review_needed → retry_count+=1
+4. retry_count>=3 → 推 Dad 人工介入
+"""
+
+import sys, json, os, datetime
+
+sys.path.insert(0, "/home/hmo/web-dashboard")
+os.chdir("/home/hmo/MoFin")
+
+DEC_PATH = "/home/hmo/web-dashboard/data/decisions.json"
+RETRY_FILE = "/home/hmo/web-dashboard/data/review_needed_retry.json"
+XMPP_USER = "hmo@yoin.fun"
+XMPP_BRIDGE = "http://192.168.1.246:5805/xmpp/send"
+
+def load_retry():
+    try:
+        return json.load(open(RETRY_FILE))
+    except:
+        return {}
+
+def save_retry(data):
+    json.dump(data, open(RETRY_FILE, "w"), indent=2)
+
+def push_xmpp(text):
+    try:
+        from urllib.request import Request, urlopen
+        payload = json.dumps({"to": XMPP_USER, "body": text.strip()}).encode()
+        req = Request(XMPP_BRIDGE, data=payload, headers={"Content-Type": "application/json"})
+        urlopen(req, timeout=5)
+        print(f"  [XMPP] 已推送")
+    except Exception as e:
+        print(f"  [XMPP推送失败] {e}")
+
+def main():
+    dec = json.load(open(DEC_PATH))
+    review_list = [d for d in dec.get("decisions", []) if d.get("status") == "review_needed"]
+    retry_data = load_retry()
+    today = datetime.date.today().isoformat()
+
+    if not review_list:
+        print("[SILENT] 无待处理策略")
+        return
+
+    print(f"发现 {len(review_list)} 只 review_needed 策略")
+    changes = False
+    for d in review_list:
+        code = d["code"]
+        name = d.get("name", code)
+        retries = retry_data.get(code, {}).get("count", 0) + 1
+        retry_data[code] = {"count": retries, "last_attempt": today}
+
+        if retries >= 3:
+            print(f"  ⛔ {name}({code}) 已重试{retries}次，跳过")
+            continue
+
+        print(f"  🔄 {name}({code}) 第{retries}次重试...")
+        import subprocess
+        r = subprocess.run(
+            [sys.executable, "/home/hmo/MoFin/scripts/per_stock_reassess.py", code],
+            capture_output=True, text=True, timeout=30
+        )
+        out = (r.stdout or "") + (r.stderr or "")
+        print(f"    {out[:200]}")
+
+        # 重读决策
+        dec2 = json.load(open(DEC_PATH))
+        for d2 in dec2.get("decisions", []):
+            if d2["code"] == code:
+                if d2.get("status") == "active":
+                    print(f"  ✅ {name}({code}) 重评通过！")
+                    retry_data[code] = {"count": 0, "last_attempt": today}
+                    changes = True
+                elif d2.get("status") == "review_needed":
+                    issues = d2.get("quality_issues", {}).get("critical", [])
+                    print(f"  ❌ {name}({code}) 仍 review_needed ({issues})")
+                break
+
+    # 3次以上失败 → 推 Dad
+    dead = [code for code, v in retry_data.items() if v.get("count", 0) >= 3]
+    if dead:
+        names = []
+        for code in dead:
+            for d in dec.get("decisions", []):
+                if d["code"] == code:
+                    names.append(f"{d.get('name', code)}({code})")
+                    break
+        msg = f"【知微】策略质量审核 {today}\n以下策略3次自动重评均失败，需人工介入：\n"
+        for n in names:
+            msg += f"  - {n}\n"
+        msg += "\n原因可能是：缺少技术面数据 / 行业信息不完整 / 利润保护目标无法确定。"
+        push_xmpp(msg)
+
+    save_retry(retry_data)
+    if not changes:
+        print("[SILENT] 状态无变化")
+
+if __name__ == "__main__":
+    main()