diff --git a/app/services/agent/prompts.py b/app/services/agent/prompts.py
index ecd3bd0..57620ea 100644
--- a/app/services/agent/prompts.py
+++ b/app/services/agent/prompts.py
@@ -25,6 +25,7 @@ SYSTEM_PROMPT = """你是VSP物业AI助手，通过企业微信协助物业人
 ## 交互规则
 - 用户发图片时，如果有待处理工单，询问是否作为处理结果上传
 - 用户说"处理完了"并附带描述，自动提交结单
+- 如果用户已经在当前会话里发送过处理结果图片，绝对不要再次要求用户提供图片URL或COS链接，直接使用系统暂存图片
 - 回复简洁，适合手机阅读
 - 重要信息用【】标注
 - 禁止使用markdown语法（如![](url)、**加粗**、# 标题），企微聊天不支持
diff --git a/app/services/agent_dispatcher.py b/app/services/agent_dispatcher.py
index 2aef6a0..a5773b9 100644
--- a/app/services/agent_dispatcher.py
+++ b/app/services/agent_dispatcher.py
@@ -59,6 +59,7 @@ class AgentDispatcher:
         normalized_content = content.strip()
 
         try:
+            self._maybe_mark_close_photo_intent(session, normalized_content)
             reply = self._maybe_start_manual_order_from_text(session, normalized_content)
             if reply is None:
                 reply = await self._handle_manual_order_message(user_id, session, normalized_content)
@@ -106,6 +107,14 @@ class AgentDispatcher:
             del presign_url
             return await self._start_manual_order_flow(session, permanent_url)
 
+        if session.state == "waiting_close_photo":
+            del presign_url
+            session.pending_images.append(permanent_url)
+            return (
+                f"已收到图片，当前已暂存 {len(session.pending_images)} 张处理结果图片。\n"
+                "请继续告诉我工单编号或序号，并补充处理描述；无需再提供图片URL。"
+            )
+
         # 3. 检查用户是否有待处理工单
         handling_alarm_id = self._find_handling_alarm(user_id)
         if handling_alarm_id:
@@ -143,6 +152,19 @@ class AgentDispatcher:
 
     async def _langgraph_chat(self, user_id: str, content: str) -> str:
         """LangGraph 图调用"""
+        session = get_session_manager().get(user_id)
+        graph_content = content
+        hints: List[str] = []
+        if session.state == "waiting_close_photo":
+            hints.append("系统提示：用户当前正在补充工单处理结果图片。")
+        if session.pending_images:
+            hints.append(
+                f"系统提示：用户已上传 {len(session.pending_images)} 张图片，后台已有永久URL。"
+                "不要再要求用户提供图片URL或COS链接。若要提交处理结果，直接调用 submit_order_result，image_urls 可留空。"
+            )
+        if hints:
+            graph_content = f"{content}\n\n" + "\n".join(hints)
+
         config = {
             "configurable": {
                 "thread_id": f"wechat-{user_id}",
@@ -152,7 +174,7 @@ class AgentDispatcher:
 
         result = await self._graph.ainvoke(
             {
-                "messages": [{"role": "user", "content": content}],
+                "messages": [{"role": "user", "content": graph_content}],
                 "user_id": user_id,
                 "pending_images": [],
                 "user_uploaded_images": [],
@@ -223,6 +245,29 @@ class AgentDispatcher:
         session.state = "waiting_manual_order_image"
         return "请先上传现场图片，我会在收到图片后引导您选择区域并补充备注。"
 
+    @staticmethod
+    def _maybe_mark_close_photo_intent(session, content: str) -> None:
+        """用户明确表示要上传处理结果图片时，优先保留该意图。"""
+        trigger_phrases = (
+            "上传完结图片",
+            "上传完结的图片",
+            "上传完结照片",
+            "上传处理结果",
+            "提交处理结果",
+            "补传图片",
+            "补传照片",
+            "结单图片",
+            "完结图片",
+        )
+        if any(phrase in content for phrase in trigger_phrases):
+            if session.state not in {
+                "waiting_manual_order_image",
+                "waiting_manual_order_area",
+                "waiting_manual_order_remark",
+                "waiting_manual_order_confirm",
+            }:
+                session.state = "waiting_close_photo"
+
     async def _handle_manual_order_message(self, user_id: str, session, content: str) -> Optional[str]:
         """处理手动工单创建状态机。"""
         if session.state == "waiting_manual_order_image":