feat: refine risk level detection logic in code review results

Improved the accuracy of risk level assessment from LLM responses by implementing a prioritized detection strategy:
1. Matches leading "✅" for skipped or passed reviews.
2. Specifically searches for an emoji (✅, ⚠️, or 🚨) within a "Risk Level" line.
3. Implements a fallback that checks for actual list items within the "High Risk" section to avoid false positives triggered by section headers.

ai_review/main.py

@@ -2,6 +2,7 @@
 """主流程编排。"""
 
 import os
+import re
 import time
 
 from ai_review.config import (
@@ -137,9 +138,21 @@ def main():
     print(review_result)
     print("=" * 60)
 
-    has_high_risk = "🔴" in review_result or "高风险" in review_result
-    is_pass = "✅" in review_result and not has_high_risk
-    risk_emoji = "✅" if is_pass else ("🚨" if has_high_risk else "⚠️")
+    # 风险等级判定，按可靠度由高到低：
+    # 1. skip 路径返回的简短结论以 ✅ 开头
+    # 2. LLM 在 "风险等级" 行里明确给出 ✅/⚠️/🚨
+    # 3. fallback：看 "## 🔴 高风险" 节区里是否含实际条目（- [...] 格式），
+    #    避免被节区标题本身（"🔴"/"高风险"字样）误判为高风险
+    if re.match(r"^\s*✅", review_result):
+        risk_emoji = "✅"
+    else:
+        m = re.search(r"风险等级[：:]\s*(✅|⚠️|🚨)", review_result)
+        if m:
+            risk_emoji = m.group(1)
+        else:
+            sec = re.search(r"##\s*🔴[^\n]*\n([\s\S]*?)(?=\n##|\Z)", review_result)
+            has_high_risk = bool(sec and re.search(r"^\s*-\s*\[", sec.group(1), re.MULTILINE))
+            risk_emoji = "🚨" if has_high_risk else "⚠️"
 
     # 构造 diff 链接
     project_url = os.environ.get("CI_PROJECT_URL", "")