Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Claude Code 懶人包全集 — 環境建置、MCP 串接、技能安裝。說「Claude Code 懶人包」「安裝懶人包」時載入。
.claude/skills/mathruffian-dot-claude-code-lazy-packs/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-11 | ✗→✓ | ▲ Improved | -75% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -4% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -54% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -71% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -55% | 0% |
當使用者給你這個 repo 網址並說要安裝時：
| 編號 | Skill 名稱 | 說明 | |------|-----------|------| | 00 | 00-env-setup | 安裝 uv 等基礎工具 | | 01 | 01-notebooklm | 連接 NotebookLM MCP | | 02 | 02-github | 連接 GitHub CLI + push 驗證 | | 03 | 03-obsidian | 連接 Obsidian MCPVault | | 04 | 04-second-brain | 第二大腦三層結構 | | 05 | 04-supabase | 連接 Supabase 資料庫 | | 06 | 05-firebase | 連接 Firebase 資料庫 | | 07 | 06-ollama | 安裝本地 AI Ollama | | 08 | 07-gemini | 設定 Gemini 免費 API | | 09 | 08-workspace | 老師建專案工作模式 | | 10 | 09-draw | 安裝 gpt-image-2 生圖 skill | | 11 | 00-install-all | 一次安裝全部 |
問：「你要安裝哪些？輸入全部或編號組合（例如 00, 01, 03）。」
bashnpx skills add mathruffian-dot/claude-code-lazy-packs --skill <名稱> -g -y
若無法使用 npx skills add，改手動讀取 skills/<名稱>/SKILL.md 執行。
每項回報 ✅/⚠️/❌，最後列總表。
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 15,544 | 6,116 | -61% | 1 | 1 | 0% | 2,441 | 1,707 | -30% | 0 | 0 | — |
case-11 | fail→pass | 13,117 | 1,371 | -90% | 1 | 1 | 0% | 2,374 | 602 | -75% | 0 | 0 | — |
case-02 | fail→fail | 9,990 | 4,572 | -54% | 1 | 1 | 0% | 606 | 1,348 | +122% | 0 | 0 | — |
case-03 | fail→fail | 9,763 | 3,196 | -67% | 1 | 1 | 0% | 594 | 1,081 | +82% | 0 | 0 | — |
case-04 | pass→pass | 16,379 | 19,655 | +20% | 1 | 1 | 0% | 3,342 | 4,565 | +37% | 0 | 0 | — |
case-05 | pass→pass | 10,465 | 9,796 | -6% | 1 | 1 | 0% | 1,891 | 2,203 | +16% | 0 | 0 | — |
case-06 | pass→pass | 9,559 | 9,080 | -5% | 1 | 1 | 0% | 1,653 | 1,992 | +21% | 0 | 0 | — |
case-07 | fail→pass | 7,515 | 3,691 | -51% | 1 | 1 | 0% | 1,162 | 1,112 | -4% | 0 | 0 | — |
case-08 | fail→pass | 11,650 | 3,033 | -74% | 1 | 1 | 0% | 2,322 | 1,069 | -54% | 0 | 0 | — |
case-09 | fail→pass | 18,800 | 2,185 | -88% | 1 | 1 | 0% | 2,915 | 850 | -71% | 0 | 0 | — |
case-10 | fail→pass | 8,883 | 1,796 | -80% | 1 | 1 | 0% | 1,582 | 711 | -55% | 0 | 0 | — |
case-12 | fail→pass | 10,214 | 1,861 | -82% | 1 | 1 | 0% | 1,779 | 681 | -62% | 0 | 0 | — |
case-13 | fail→pass | 13,818 | 2,230 | -84% | 1 | 1 | 0% | 2,491 | 852 | -66% | 0 | 0 | — |
case-14 | fail→pass | 13,261 | 2,523 | -81% | 1 | 1 | 0% | 2,159 | 866 | -60% | 0 | 0 | — |
case-15 | fail→pass | 11,830 | 2,617 | -78% | 1 | 1 | 0% | 2,032 | 731 | -64% | 0 | 0 | — |
case-16 | fail→pass | 13,438 | 2,285 | -83% | 1 | 1 | 0% | 2,390 | 842 | -65% | 0 | 0 | — |
case-17 | fail→pass | 9,744 | 1,964 | -80% | 1 | 1 | 0% | 1,727 | 753 | -56% | 0 | 0 | — |
case-18 | fail→pass | 15,128 | 2,412 | -84% | 1 | 1 | 0% | 2,598 | 897 | -65% | 0 | 0 | — |
case-19 | fail→pass | 11,585 | 2,357 | -80% | 1 | 1 | 0% | 1,947 | 810 | -58% | 0 | 0 | — |
case-20 | fail→pass | 21,147 | 1,404 | -93% | 1 | 1 | 0% | 3,306 | 629 | -81% | 0 | 0 | — |
case-21 | pass→pass | 4,225 | 2,174 | -49% | 1 | 1 | 0% | 712 | 777 | +9% | 0 | 0 | — |
case-22 | pass→pass | 11,063 | 3,268 | -70% | 1 | 1 | 0% | 1,751 | 1,024 | -42% | 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. The headline lift of +64 percentage points is the difference between those two pass rates over the 22 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.