diff --git a/.gitea/ai-review/exclusions.json b/.gitea/ai-review/exclusions.json
index 4ffa10c..6a36245 100644
--- a/.gitea/ai-review/exclusions.json
+++ b/.gitea/ai-review/exclusions.json
@@ -488,5 +488,35 @@
         "role": "Assassin",
         "original_finding": "extractUsage 對不預期 payload 僅返回 null，過度信任 API 回應結構，可能導致計費或配額相關監控被繞過；建議增加 Schema Validation、異常明確記錄。",
         "reason": "誤判／過度設計。extractUsage 僅用於「使用量顯示統計」，非計費或配額強制；回傳 null 是「此回應無可辨識 usage 資訊」的正確訊號，呼叫端以 0 計入並降級顯示，不影響任何金流或門檻判斷。對 best-effort 顯示統計加 schema validation 與錯誤記錄屬過度設計。"
+    },
+    {
+        "location": "app/comments.test.js:250",
+        "role": "Maya",
+        "original_finding": "新增的 postFindingsReview 使用統計功能，但在測試中完全未驗證輸出內容；應斷言 body 含 usageSection 與統計數據。",
+        "reason": "誤判，測試已存在。`app/comments.test.js` 的 'appends usageSection verbatim after the stats block' 斷言 body.endsWith(usageSection) 與結構，'counts both new and old findings in the summary' 斷言四欄新舊統計列；body 內容已被多個案例驗證。"
+    },
+    {
+        "location": "app/findings.test.js:154",
+        "role": "Maya",
+        "original_finding": "filterFalsePositivesWithAI 測試不足，缺乏對內部函數 judgeFindingIsFalsePositive 的獨立單元測試。",
+        "reason": "誤判／不適用。judgeFindingIsFalsePositive 是 findings.js 的私有函式（未匯出），其 verdict 處理（false_positive／confirmed／異常值／拋錯）已透過公開呼叫端 filterFalsePositivesWithAI 的多個案例完整覆蓋；為測試實作細節而匯出私有函式不符測試原則。"
+    },
+    {
+        "location": "app/findings.test.js:145",
+        "role": "Maya",
+        "original_finding": "filterFalsePositivesWithAI 未測試平行裁決部分成功、部分失敗時的結果一致性（失敗者保守保留）。",
+        "reason": "誤判，測試已存在。'keeps failed and confirmed, drops only confirmed false positives (mixed parallel)' 正是模擬一個拋錯、一個誤報、一個成立，驗證只剔除確認誤報、保留失敗與成立者。"
+    },
+    {
+        "location": "app/findings.test.js:189",
+        "role": "Maya",
+        "original_finding": "未測試 resolveMissingLineNumbers 當 chatFn 回傳無效行號時的處理（fallback）。",
+        "reason": "誤判，測試已存在。'resolveMissingLineNumbers keeps the filename after exhausting retries' 以 chatFn 持續回 {line:0}（無效行號）驗證進入 fallback、保留檔名；另有 'swallows chatFn exceptions' 覆蓋拋錯情境。"
+    },
+    {
+        "location": "app/resolve.test.js:21",
+        "role": "Maya",
+        "original_finding": "parseBotReviewComment 的測試沒有驗證解析失敗時回傳 null 的行為。",
+        "reason": "誤判，測試已存在。'returns null for free-form human comments' 已斷言自由格式留言、空字串、null 皆回傳 null。"
     }
 ]
diff --git a/.gitea/ai-review/findings.json b/.gitea/ai-review/findings.json
index 7669a87..fe51488 100644
--- a/.gitea/ai-review/findings.json
+++ b/.gitea/ai-review/findings.json
@@ -1,45 +1 @@
-[
-  {
-    "level": "critical",
-    "role": "Maya",
-    "location": "app/comments.test.js:250",
-    "problem": "新增的 `postFindingsReview` 使用統計功能，但在測試中完全未驗證輸出內容。",
-    "suggestion": "應斷言 `reviewCalls[0].body` 確實包含了預期的 `usageSection` 資訊與統計數據。"
-  },
-  {
-    "level": "critical",
-    "role": "Maya",
-    "location": "app/findings.test.js:154",
-    "problem": "`filterFalsePositivesWithAI` 測試不足，缺乏對 `judgeFindingIsFalsePositive` 內部的獨立單元測試。",
-    "suggestion": "為內部函數 `judgeFindingIsFalsePositive` 補寫測試，單獨驗證其對不同 Verdict 值（false_positive, confirmed, 異常值）的處理邏輯。"
-  },
-  {
-    "level": "warning",
-    "role": "Maya",
-    "problem": "在 `filterFalsePositivesWithAI` 的測試中，雖然模擬了平行處理，但並未測試當多個並行裁決（Promise.all）中，部分成功、部分失敗時的結果一致性（即確保失敗者保守保留）。",
-    "suggestion": "增加測試案例：模擬其中一個 sub-agent 拋出錯誤、另一個判為誤報、第三個判為成立，驗證最終結果是否正確地保留了「失敗者」與「成立者」，且只剔除「確認誤報者」。",
-    "location": "app/findings.test.js:145",
-    "is_new": false
-  },
-  {
-    "level": "warning",
-    "role": "Maya",
-    "location": "app/findings.test.js:189",
-    "problem": "未測試 `resolveMissingLineNumbers` 當 `chatFn` 回傳無效行號時的處理。",
-    "suggestion": "補上測試案例：模擬 `chatFn` 回傳無效行號，確保其進入 fallback 邏輯。"
-  },
-  {
-    "level": "warning",
-    "role": "Maya",
-    "location": "app/resolve.test.js:21",
-    "problem": "在 `parseBotReviewComment` 的測試中，沒有驗證解析失敗時的行為。",
-    "suggestion": "補上邊界測試：輸入不完整的內容，驗證函數是否正確回傳 `null`。"
-  },
-  {
-    "level": "info",
-    "role": "Maya",
-    "location": "app/resolve.test.js:133",
-    "problem": "`judgeConversations` 的測試中，未對「AI 回傳空陣列」或「所有 Verdict 皆為空」的情境進行邊界驗證。",
-    "suggestion": "補上邊界測試，驗證該情境下是否將所有對話歸類為 `open`（保守保留）。"
-  }
-]
+[]