Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →透過程式碼庫和文件驗證資訊準確性。「這個資訊對嗎？」「事實檢查」「驗證這個」等觸發。
.claude/skills/wasabeef-d8b66e/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | -9% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 9% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 44% | 0% |
| case-16 | ✗→✓ | ▲ Improved | 2% | 0% |
| case-17 | ✗→✓ | ▲ Improved | 8% | 0% |
參考項目內的代碼庫、文檔 (docs/、README.md 等)，確認所給資訊的正確性。
bash# 基本用法 /check-fact "Flutter 應用使用了 Riverpod" # 一次性確認多個資訊 /check-fact "這個項目使用 GraphQL，並通過 auto_route 管理路由" # 確認特定技術規範 /check-fact "使用 JWT 進行身份驗證，未使用 Firebase Auth"
✅ 正確 - 資訊與代碼庫完全一致❌ 錯誤 - 資訊明顯錯誤⚠️ 部分正確 - 部分準確但不完整❓ 無法判斷 - 缺少必要的確認資訊text## 事實檢查結果 ### 檢驗對象 「[用戶提供的資訊]」 ### 結論 [✅/❌/⚠️/❓] [判定結果] ### 依據 - **文件**: `path/to/file.dart:123` - **內容**: [相關代碼/文本] - **補充**: [额外說明] ### 詳细說明 [如果錯誤，提供正確資訊] [如果部分正確，指出不準確的部分] [如果無法判斷，說明缺少的資訊]
bash# 項目技術棧確認 /check-fact "這個應用是 Flutter + Riverpod + GraphQL 的架構" # 實現狀况確認 /check-fact "已實現暗黑模式功能，可從用戶設置切換" # 架構確認 /check-fact "狀態管理全部使用 Riverpod，未使用 BLoC" # 安全實現確認 /check-fact "認證令牌已加密存儲在 secure storage 中"
bash# 分析整個代碼庫後進行確認 ls -la && find . -name "pubspec.yaml" -exec cat {} \; /check-fact "這個項目使用的主要依賴有..." # 確認特定功能的實現狀况 grep -r "authentication" . --include="*.dart" /check-fact "認證功能為自定義實現，未使用第三方認證" # 確認與文檔的一致性 cat README.md /check-fact "README 中記載的功能都已實現"
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 22,469 | 5,555 | -75% | 1 | 1 | 0% | 3,988 | 1,118 | -72% | 0 | 0 | — |
case-02 | fail→fail | 16,294 | 15,517 | -5% | 1 | 1 | 0% | 2,579 | 2,683 | +4% | 0 | 0 | — |
case-03 | fail→pass | 16,992 | 11,105 | -35% | 1 | 1 | 0% | 2,906 | 2,637 | -9% | 0 | 0 | — |
case-04 | pass→pass | 12,896 | 6,307 | -51% | 1 | 1 | 0% | 2,135 | 1,872 | -12% | 0 | 0 | — |
case-05 | pass→pass | 9,947 | 4,739 | -52% | 1 | 1 | 0% | 1,829 | 1,759 | -4% | 0 | 0 | — |
case-06 | pass→pass | 10,452 | 4,180 | -60% | 1 | 1 | 0% | 1,706 | 1,568 | -8% | 0 | 0 | — |
case-07 | pass→pass | 10,992 | 5,393 | -51% | 1 | 1 | 0% | 1,984 | 1,773 | -11% | 0 | 0 | — |
case-08 | pass→pass | 9,934 | 5,012 | -50% | 1 | 1 | 0% | 1,664 | 1,610 | -3% | 0 | 0 | — |
case-09 | fail→pass | 7,174 | 3,428 | -52% | 1 | 1 | 0% | 1,308 | 1,429 | +9% | 0 | 0 | — |
case-10 | pass→fail | 11,927 | 9,838 | -18% | 1 | 1 | 0% | 1,979 | 2,553 | +29% | 0 | 0 | — |
case-11 | pass→pass | 15,551 | 21,888 | +41% | 1 | 1 | 0% | 3,192 | 4,359 | +37% | 0 | 0 | — |
case-12 | pass→pass | 9,595 | 5,091 | -47% | 1 | 1 | 0% | 1,768 | 1,682 | -5% | 0 | 0 | — |
case-13 | pass→fail | 12,336 | 2,769 | -78% | 1 | 1 | 0% | 2,146 | 1,024 | -52% | 0 | 0 | — |
case-14 | fail→pass | 6,055 | 6,408 | +6% | 1 | 1 | 0% | 1,025 | 1,474 | +44% | 0 | 0 | — |
case-15 | pass→pass | 10,181 | 4,852 | -52% | 1 | 1 | 0% | 1,982 | 1,828 | -8% | 0 | 0 | — |
case-16 | fail→pass | 12,014 | 7,108 | -41% | 1 | 1 | 0% | 2,132 | 2,167 | +2% | 0 | 0 | — |
case-17 | fail→pass | 9,732 | 5,596 | -42% | 1 | 1 | 0% | 1,641 | 1,774 | +8% | 0 | 0 | — |
case-18 | pass→pass | 16,458 | 6,066 | -63% | 1 | 1 | 0% | 2,769 | 1,817 | -34% | 0 | 0 | — |
case-19 | pass→pass | 10,963 | 4,631 | -58% | 1 | 1 | 0% | 1,846 | 1,573 | -15% | 0 | 0 | — |
case-20 | pass→pass | 8,212 | 4,683 | -43% | 1 | 1 | 0% | 1,386 | 1,652 | +19% | 0 | 0 | — |
case-21 | fail→pass | 7,806 | 4,377 | -44% | 1 | 1 | 0% | 1,484 | 1,634 | +10% | 0 | 0 | — |
case-22 | pass→pass | 13,155 | 6,889 | -48% | 1 | 1 | 0% | 2,128 | 1,862 | -13% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 21 counted toward the lift figure. The other 1 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +18 percentage points is the difference between those two pass rates over the 21 comparable cases. 2 cases got worse with the skill loaded, and they are included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.