diff --git a/src/ai/client.py b/src/ai/client.py
index 5e4be00..44d2cc0 100644
--- a/src/ai/client.py
+++ b/src/ai/client.py
@@ -493,6 +493,11 @@ class AIClient:
             return None
         return min(limit, 5000)
 
+    @staticmethod
+    def _compact_identifier(text: str) -> str:
+        """Compact identifier for fuzzy matching (e.g. humanizer_zh -> humanizerzh)."""
+        return re.sub(r"[^a-z0-9]+", "", (text or "").lower())
+
     @staticmethod
     def _extract_forced_tool_name(
         user_message: str, available_tool_names: List[str]
@@ -500,7 +505,16 @@ class AIClient:
         if not user_message or not available_tool_names:
             return None
 
-        triggers = ["调用工具", "使用工具", "只调用", "务必调用", "必须调用", "tool"]
+        triggers = [
+            "调用工具",
+            "使用工具",
+            "只调用",
+            "务必调用",
+            "必须调用",
+            "调用",
+            "使用",
+            "tool",
+        ]
         if not any(trigger in user_message for trigger in triggers):
             return None
 
@@ -537,6 +551,38 @@ class AIClient:
             if len(prefix_tools) == 1:
                 return prefix_tools[0]
 
+        # 模糊匹配：支持省略下划线/点号的写法（如 humanizerzh）。
+        compact_message = AIClient._compact_identifier(user_message)
+        if compact_message:
+            compact_full_matches = []
+            for tool_name in available_tool_names:
+                compact_tool_name = AIClient._compact_identifier(tool_name)
+                if compact_tool_name and compact_tool_name in compact_message:
+                    compact_full_matches.append(tool_name)
+
+            if len(compact_full_matches) == 1:
+                return compact_full_matches[0]
+            if len(compact_full_matches) > 1:
+                return None
+
+            compact_prefix_map: Dict[str, List[str]] = {}
+            for tool_name in available_tool_names:
+                prefix = tool_name.split(".", 1)[0]
+                compact_prefix = AIClient._compact_identifier(prefix)
+                if not compact_prefix:
+                    continue
+                compact_prefix_map.setdefault(compact_prefix, []).append(tool_name)
+
+            compact_prefix_matches = [
+                compact_prefix
+                for compact_prefix in compact_prefix_map
+                if compact_prefix in compact_message
+            ]
+            if len(compact_prefix_matches) == 1:
+                matched_tools = compact_prefix_map[compact_prefix_matches[0]]
+                if len(matched_tools) == 1:
+                    return matched_tools[0]
+
         return None
     
     def set_personality(self, personality_name: str) -> bool:
diff --git a/tests/test_ai_client_forced_tool.py b/tests/test_ai_client_forced_tool.py
index 07ad623..cb2ac85 100644
--- a/tests/test_ai_client_forced_tool.py
+++ b/tests/test_ai_client_forced_tool.py
@@ -27,6 +27,18 @@ def test_extract_forced_tool_name_unique_prefix():
     assert forced == "humanizer_zh.read_skill_doc"
 
 
+def test_extract_forced_tool_name_compact_prefix_without_underscore():
+    tools = [
+        "humanizer_zh.read_skill_doc",
+        "skills_creator.create_skill",
+    ]
+    message = "调用humanizerzh人性化处理以下文本"
+
+    forced = AIClient._extract_forced_tool_name(message, tools)
+
+    assert forced == "humanizer_zh.read_skill_doc"
+
+
 def test_extract_forced_tool_name_ambiguous_prefix_returns_none():
     tools = [
         "skills_creator.create_skill",