Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when developing, operating, or deploying Google Apps Script projects with CI/CD automation.
.claude/skills/bilal140202-gas-ops/SKILL.md| Model | Eval pass | Runs |
|---|---|---|
| gemini-3.6-flash | 100% | 32 |
| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -22% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -4% | 0% |
| case-12 | ✗→✓ | ▲ Improved | -46% | 0% |
| case-13 | ✗→✓ | ▲ Improved | -56% | 0% |
| case-02 | ✓→✓ | ＝ Same ✓ | 7% | 0% |
此技能規範 Google Apps Script (GAS) 開發的標準流程，結合 clasp 工具與 GitHub Actions 實現自動化部屬與版本控制。
npm install -g @google/clasp)在本地進行開發後，通過 Git 提交代碼。若需從 GAS 網頁工作區同步回本地：
powershellclasp pull
專案應配置 .github/workflows/deploy.yml，在推送到特定分支（如 master 或 main）時自動執行部屬。
clasp 並利用 CLASPRC_JSON_BASE64 秘密金鑰進行授權。clasp push --force僅在 CI/CD 失效且需緊急修復時使用：
powershellclasp push
clasp login 獲取本地 ~/.clasprc.json。powershell [Convert]::ToBase64String([IO.File]::ReadAllBytes("$HOME\.clasprc.json"))
CLASPRC_JSON_BASE64。clasp pull 回本地。git push 觸發 Action 進行部屬，確保部署紀錄與 Git Commit 關聯。clasp push --force 以避免文件版本衝突。| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 13,864 | 7,473 | -46% | 1 | 1 | 0% | 2,811 | 2,185 | -22% | 0 | 0 | — |
case-02 | pass→pass | 4,646 | 2,234 | -52% | 1 | 1 | 0% | 902 | 967 | +7% | 0 | 0 | — |
case-03 | pass→pass | 5,474 | 1,729 | -68% | 1 | 1 | 0% | 1,130 | 832 | -26% | 0 | 0 | — |
case-04 | pass→pass | 5,372 | 3,208 | -40% | 1 | 1 | 0% | 1,025 | 762 | -26% | 0 | 0 | — |
case-05 | pass→pass | 1,487 | 1,343 | -10% | 1 | 1 | 0% | 261 | 760 | +191% | 0 | 0 | — |
case-06 | pass→pass | 7,760 | 1,907 | -75% | 1 | 1 | 0% | 1,551 | 873 | -44% | 0 | 0 | — |
case-07 | pass→pass | 8,819 | 4,074 | -54% | 1 | 1 | 0% | 1,693 | 1,369 | -19% | 0 | 0 | — |
case-08 | pass→pass | 9,224 | 5,710 | -38% | 1 | 1 | 0% | 1,607 | 1,543 | -4% | 0 | 0 | — |
case-09 | pass→pass | 3,059 | 3,151 | +3% | 1 | 1 | 0% | 473 | 730 | +54% | 0 | 0 | — |
case-10 | fail→pass | 6,953 | 3,080 | -56% | 1 | 1 | 0% | 1,188 | 1,142 | -4% | 0 | 0 | — |
case-11 | pass→pass | 2,712 | 2,321 | -14% | 1 | 1 | 0% | 454 | 901 | +98% | 0 | 0 | — |
case-12 | fail→pass | 7,386 | 2,752 | -63% | 1 | 1 | 0% | 1,392 | 748 | -46% | 0 | 0 | — |
case-13 | fail→pass | 14,194 | 3,487 | -75% | 1 | 1 | 0% | 2,521 | 1,112 | -56% | 0 | 0 | — |
case-14 | pass→pass | 4,363 | 1,970 | -55% | 1 | 1 | 0% | 909 | 844 | -7% | 0 | 0 | — |
case-15 | pass→pass | 3,962 | 1,529 | -61% | 1 | 1 | 0% | 729 | 782 | +7% | 0 | 0 | — |
case-16 | pass→pass | 4,910 | 1,376 | -72% | 1 | 1 | 0% | 951 | 738 | -22% | 0 | 0 | — |
case-17 | pass→pass | 9,467 | 3,593 | -62% | 1 | 1 | 0% | 1,575 | 1,071 | -32% | 0 | 0 | — |
case-18 | pass→pass | 4,278 | 1,481 | -65% | 1 | 1 | 0% | 773 | 800 | +3% | 0 | 0 | — |
case-19 | pass→pass | 9,260 | 10,744 | +16% | 1 | 1 | 0% | 1,679 | 2,370 | +41% | 0 | 0 | — |
case-20 | pass→pass | 6,289 | 4,663 | -26% | 1 | 1 | 0% | 1,363 | 1,446 | +6% | 0 | 0 | — |
case-21 | pass→pass | 2,777 | 2,472 | -11% | 1 | 1 | 0% | 547 | 1,003 | +83% | 0 | 0 | — |
case-22 | pass→pass | 2,930 | 2,913 | -1% | 1 | 1 | 0% | 378 | 822 | +117% | 0 | 0 | — |
case-23 | pass→pass | 6,758 | 2,259 | -67% | 1 | 1 | 0% | 1,167 | 924 | -21% | 0 | 0 | — |
case-24 | pass→pass | 9,915 | 1,779 | -82% | 1 | 1 | 0% | 1,804 | 866 | -52% | 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. 24 cases were attempted. The headline lift of +17 percentage points is the difference between those two pass rates over the 24 comparable cases.
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.