diff --git a/agent_system/triager/agent.py b/agent_system/triager/agent.py
index 7f3f2cc..7be0144 100644
--- a/agent_system/triager/agent.py
+++ b/agent_system/triager/agent.py
@@ -141,10 +141,10 @@ class TriageAgent(BaseAgent):
             TriageResult: 包含默认分诊建议的结果
         """
         return TriageResult(
+            triage_reasoning="由于分诊分析过程中出现异常，系统推荐全科就诊。建议患者先到全科进行初步评估，医生会根据具体情况进一步转诊到合适的专科。",
             primary_department="全科",
             secondary_department="全科（二级）",
-            confidence_score=0.3,
-            triage_reasoning="由于分诊分析过程中出现异常，系统推荐全科就诊。建议患者先到全科进行初步评估，医生会根据具体情况进一步转诊到合适的专科。"
+            confidence_score=0.3
         )
     
     def triage_by_chief_complaint(self, chief_complaint: str) -> TriageResult:
diff --git a/agent_system/triager/prompt.py b/agent_system/triager/prompt.py
index 266989f..f785930 100644
--- a/agent_system/triager/prompt.py
+++ b/agent_system/triager/prompt.py
@@ -19,11 +19,13 @@ class TriagerPrompt(BasePrompt):
     
     # 执行指令和注意事项
     instructions = [
-        "## 核心分诊任务",
-        "1. **症状系统分析**: 根据患者主诉和现病史，准确识别疾病涉及的主要器官系统",
-        "2. **科室匹配判断**: 结合症状特点，选择最适合的一级科室和对应的二级科室",
-        "3. **专业度评估**: 基于病情复杂程度和专业要求，确定最合适的专科科室",
-        "4. **信心度评价**: 对分诊决策的确信程度进行量化评估",
+        "## 分诊分析步骤",
+        "请按照以下步骤进行分诊分析：",
+        "",
+        "**第一步：症状分析与推理** - 分析患者症状特点，识别涉及的器官系统，提供详细的医学推理过程",
+        "**第二步：一级科室选择** - 根据症状系统归属和年龄特异性，选择最合适的一级科室",
+        "**第三步：二级科室匹配** - 基于病情复杂程度和专业要求，确定对应的二级科室",
+        "**第四步：信心度评估** - 对分诊决策的确信程度进行量化评估（0-1分）",
         "",
         "## 科室结构体系",
         "### 一级科室列表：",
@@ -59,10 +61,10 @@ class TriagerPrompt(BasePrompt):
         "",
         "## 示例输出格式（JSON）",
         "{",
+        "  \"triage_reasoning\": \"患者主诉胸闷、胸痛，伴有呼吸困难，症状提示心血管系统疾病。结合既往高血压病史，考虑冠心病可能性较大，建议心血管内科就诊进行进一步检查和治疗。\",",
         "  \"primary_department\": \"内科\",",
         "  \"secondary_department\": \"心血管内科\",",
-        "  \"confidence_score\": 0.85,",
-        "  \"triage_reasoning\": \"患者主诉胸闷、胸痛，伴有呼吸困难，症状提示心血管系统疾病。结合既往高血压病史，考虑冠心病可能性较大，建议心血管内科就诊进行进一步检查和治疗。\"",
+        "  \"confidence_score\": 0.85",
         "}"
     ]
     
@@ -75,8 +77,8 @@ class TriagerPrompt(BasePrompt):
             str: JSON 格式的示例输出
         """
         return """{
+  "triage_reasoning": "详细的分诊推理过程，包括症状分析、科室选择依据和建议理由",
   "primary_department": "推荐的一级科室（必须从科室列表中选择）",
   "secondary_department": "推荐的二级科室（必须是一级科室的下属科室）",
-  "confidence_score": 0.85,
-  "triage_reasoning": "详细的分诊推理过程，包括症状分析、科室选择依据和建议理由"
+  "confidence_score": 0.85
 }"""
\ No newline at end of file
diff --git a/agent_system/triager/response_model.py b/agent_system/triager/response_model.py
index 12e2f47..3fbde81 100644
--- a/agent_system/triager/response_model.py
+++ b/agent_system/triager/response_model.py
@@ -7,6 +7,11 @@ class TriageResult(BaseResponseModel):
     """
     科室分诊结果模型
     """
+    triage_reasoning: str = Field(
+        ...,
+        description="分诊推理过程，解释为什么推荐该科室"
+    )
+    
     primary_department: Literal[
         "内科", "外科", "儿科", "妇产科", "皮肤性病科", 
         "口腔科", "眼科", "肿瘤科", "耳鼻咽喉科", "康复科", 
@@ -26,9 +31,4 @@ class TriageResult(BaseResponseModel):
         ge=0.0,
         le=1.0,
         description="分诊信心度评分（0-1之间）"
-    )
-    
-    triage_reasoning: str = Field(
-        ...,
-        description="分诊推理过程，解释为什么推荐该科室"
     )
\ No newline at end of file