diff --git a/plugins/ai_auto_response/config.toml b/plugins/ai_auto_response/config.toml
index 255eef2..e902b7e 100644
--- a/plugins/ai_auto_response/config.toml
+++ b/plugins/ai_auto_response/config.toml
@@ -35,31 +35,31 @@ casual_topic = 0.35
 
 [flow]
 enable_flow_state = true
-flow_decay_per_minute = 12
-idle_threshold = 30
-warming_threshold = 48
-engaged_threshold = 78
+flow_decay_per_minute = 16
+idle_threshold = 36
+warming_threshold = 60
+engaged_threshold = 96
 at_bot_boost = 40
 question_boost = 30
-followup_boost = 20
-topic_boost = 10
-returning_member_boost = 6
-response_accepted_boost = 10
-ignored_reply_penalty = 20
-over_reply_penalty = 22
+followup_boost = 12
+topic_boost = 4
+returning_member_boost = 3
+response_accepted_boost = 6
+ignored_reply_penalty = 26
+over_reply_penalty = 32
 night_penalty = 30
-max_bot_reply_streak = 2
+max_bot_reply_streak = 1
 
 [cooldown]
-group_reply_cooldown_sec = 90
-same_user_followup_cooldown_sec = 18
+group_reply_cooldown_sec = 150
+same_user_followup_cooldown_sec = 28
 at_mention_min_interval_sec = 5
 at_mention_burst_window_sec = 90
-at_mention_burst_limit = 5
+at_mention_burst_limit = 4
 at_mention_silent_sec = 180
 directed_burst_window_sec = 240
-directed_burst_limit = 4
-directed_burst_silent_sec = 480
+directed_burst_limit = 3
+directed_burst_silent_sec = 600
 night_silent_hours = ["01:00-07:30"]
 
 [memory]
@@ -102,6 +102,11 @@ ignore_prefixes = ["/", "#"]
 ignore_exact = ["收到", "好的", "嗯", "哦", "6", "1", "？", "?"]
 min_text_length = 1
 
+[spam_guard]
+repeat_window_sec = 10
+repeat_threshold = 2
+repeat_min_length = 4
+
 [logging]
 debug = true
 
diff --git a/plugins/ai_auto_response/main.py b/plugins/ai_auto_response/main.py
index a6d908c..2f2a0d3 100644
--- a/plugins/ai_auto_response/main.py
+++ b/plugins/ai_auto_response/main.py
@@ -130,6 +130,7 @@ class AIAutoResponsePlugin(MessagePluginInterface):
         self.cooldown_config = self._config.get("cooldown", {}) or {}
         self.cooldown = CooldownManager(self.cooldown_config)
         self.image_config = self._config.get("image", {}) or {}
+        self.spam_config = self._config.get("spam_guard", {}) or {}
         self._synced_member_context_versions: Dict[str, str] = {}
         self.log_debug = bool((self._config.get("logging", {}) or {}).get("debug", True))
         self.LOG.debug(f"[{self.name}] 初始化完成")
@@ -197,6 +198,23 @@ class AIAutoResponsePlugin(MessagePluginInterface):
                     reply_mode="defense",
                 )
                 return False, "ignored_prompt_attack"
+            if self.dedup.should_skip_repeated_room_content(
+                room_id=room_id,
+                content=content,
+                window_sec=int(self.spam_config.get("repeat_window_sec", 45) or 45),
+                repeat_threshold=int(self.spam_config.get("repeat_threshold", 3) or 3),
+                min_length=int(self.spam_config.get("repeat_min_length", 4) or 4),
+            ):
+                self._log_event(
+                    "skip",
+                    room_id=room_id,
+                    sender=sender,
+                    reason="repeated_room_content",
+                    trigger_type="spam_guard",
+                    reply_mode="guard",
+                    topic="-",
+                )
+                return False, "repeated_room_content"
             coding_work_request = is_coding_work_request(content)
             if coding_work_request and not is_at:
                 return False, "skip_coding_work"
diff --git a/plugins/ai_auto_response/safety/dedup.py b/plugins/ai_auto_response/safety/dedup.py
index 3c89783..6ae57d2 100644
--- a/plugins/ai_auto_response/safety/dedup.py
+++ b/plugins/ai_auto_response/safety/dedup.py
@@ -1,7 +1,8 @@
 from __future__ import annotations
 
+import re
 import time
-from typing import Dict, Set
+from typing import Dict, List, Set, Tuple
 
 
 class DedupManager:
@@ -9,6 +10,7 @@ class DedupManager:
         self.inflight_message_keys: Set[str] = set()
         self.recent_message_keys: Dict[str, float] = {}
         self.recent_reply_signatures: Dict[str, float] = {}
+        self.recent_room_content_hits: Dict[str, List[Tuple[float, str]]] = {}
 
     def begin_message_processing(self, message_key: str, expiry_sec: int) -> bool:
         if not message_key:
@@ -51,3 +53,31 @@ class DedupManager:
             return True
         self.recent_reply_signatures[signature] = now
         return False
+
+    def should_skip_repeated_room_content(
+        self,
+        *,
+        room_id: str,
+        content: str,
+        window_sec: int,
+        repeat_threshold: int,
+        min_length: int = 4,
+    ) -> bool:
+        text = self._normalize_room_content(content)
+        if not room_id or not text or len(text) < max(int(min_length or 4), 1):
+            return False
+        now = time.time()
+        window_sec = max(int(window_sec or 0), 1)
+        repeat_threshold = max(int(repeat_threshold or 0), 2)
+        room_items = self.recent_room_content_hits.get(room_id, [])
+        room_items = [(ts, item_text) for ts, item_text in room_items if now - ts <= window_sec]
+        same_count = sum(1 for _, item_text in room_items if item_text == text)
+        room_items.append((now, text))
+        self.recent_room_content_hits[room_id] = room_items[-80:]
+        return same_count + 1 >= repeat_threshold
+
+    @staticmethod
+    def _normalize_room_content(content: str) -> str:
+        text = str(content or "").strip().lower()
+        text = re.sub(r"\s+", "", text)
+        return text