数据新鲜度校验: 所有模块加上日期/时效检查

1. search_cache.json → 添加 _metadata.updated_at 时间戳
   web_search.search_from_cache() 跳过超过36h的旧缓存
   防止某查询失败时残留旧数据

2. metrics_feedback.json → collector 检查mtime,超过24h不采用

3. trends.json → 已有 date==TODAY 校验(load_trends)

4. collector_ai_advice → DB每日覆盖,时序安全
   creator→topic状态位避免重复生成
   optimizer→文章状态位避免重复审查

scripts/collector.py

@@ -398,13 +398,18 @@ class SustainabilityCollector:
         metrics_file = DATA_DIR / "metrics_feedback.json"
         if metrics_file.exists():
             try:
-                feedback = json.loads(metrics_file.read_text(encoding='utf-8'))
-                top_domains = feedback.get("top_domains", [])
-                if top_domains:
-                    lines = ["## 历史表现反馈（高互动领域优先", ""]
-                    for d, s in top_domains[:3]:
-                        lines.append(f"- {d}：平均分 {s}")
-                    parts.append("\n".join(lines))
+                mtime = datetime.datetime.fromtimestamp(metrics_file.stat().st_mtime)
+                age = (datetime.datetime.now() - mtime).total_seconds()
+                if age > 86400:  # 超过24h的数据不采用
+                    logger.debug("metrics_feedback 过时(%.0fh)，跳过", age / 3600)
+                else:
+                    feedback = json.loads(metrics_file.read_text(encoding='utf-8'))
+                    top_domains = feedback.get("top_domains", [])
+                    if top_domains:
+                        lines = ["## 历史表现反馈（高互动领域优先", ""]
+                        for d, s in top_domains[:3]:
+                            lines.append(f"- {d}：平均分 {s}")
+                        parts.append("\n".join(lines))
             except Exception:
                 pass