Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →ターゲット リポジトリの既存統合パターンに正確に一致する新しい API コネクターまたはプロバイダーを構築します。2 番目のアーキテクチャを発明せずに、1 つ以上の統合を追加するときに使用します。
.claude/skills/affaan-m-api-connector-builder/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-09 | ✗→✓ | ▲ Improved | -37% | 0% |
| case-01 | ✗→✓ | ▲ Improved | -79% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 3% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 6% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -15% | 0% |
当任务需要添加仓库原生的集成接口，而非仅通用 HTTP 客户端时使用此工具。
关键在于匹配宿主仓库的模式：
检查至少 2 个现有连接器/提供商，并映射：
仅定义仓库实际需要的接口：
典型分层：
新连接器应在代码库中显得自然，而非从不同生态导入。
textproviders/ existing_provider/ __init__.py provider.py config.py
textintegrations/ existing/ client.py models.py connector.py
textsrc/integrations/ existing/ index.ts client.ts types.ts test.ts
backend-patternsmcp-server-patternsgithub-ops| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-09 | fail→pass | 14,314 | 6,191 | -57% | 1 | 1 | 0% | 2,564 | 1,604 | -37% | 0 | 0 | — |
case-07 | pass→pass | 12,616 | 5,816 | -54% | 1 | 1 | 0% | 1,856 | 1,571 | -15% | 0 | 0 | — |
case-08 | pass→pass | 11,698 | 10,868 | -7% | 1 | 1 | 0% | 2,068 | 2,519 | +22% | 0 | 0 | — |
case-01 | fail→pass | 20,305 | 2,608 | -87% | 1 | 1 | 0% | 4,607 | 971 | -79% | 0 | 0 | — |
case-02 | fail→fail | 22,384 | 21,581 | -4% | 1 | 1 | 0% | 5,004 | 5,416 | +8% | 0 | 0 | — |
case-03 | fail→fail | 24,421 | 24,694 | +1% | 1 | 1 | 0% | 5,415 | 6,321 | +17% | 0 | 0 | — |
case-04 | fail→pass | 13,190 | 11,612 | -12% | 1 | 1 | 0% | 2,610 | 2,683 | +3% | 0 | 0 | — |
case-05 | fail→pass | 10,941 | 8,267 | -24% | 1 | 1 | 0% | 2,032 | 2,148 | +6% | 0 | 0 | — |
case-06 | pass→pass | 10,119 | 8,946 | -12% | 1 | 1 | 0% | 1,838 | 2,074 | +13% | 0 | 0 | — |
case-10 | pass→pass | 11,222 | 7,627 | -32% | 1 | 1 | 0% | 1,854 | 1,872 | +1% | 0 | 0 | — |
case-11 | fail→pass | 10,583 | 6,137 | -42% | 1 | 1 | 0% | 1,934 | 1,646 | -15% | 0 | 0 | — |
case-12 | fail→pass | 11,517 | 7,480 | -35% | 1 | 1 | 0% | 2,164 | 2,078 | -4% | 0 | 0 | — |
case-13 | fail→pass | 12,519 | 6,130 | -51% | 1 | 1 | 0% | 2,363 | 1,803 | -24% | 0 | 0 | — |
case-14 | fail→pass | 14,108 | 12,377 | -12% | 1 | 1 | 0% | 2,363 | 2,856 | +21% | 0 | 0 | — |
case-15 | pass→pass | 11,738 | 7,361 | -37% | 1 | 1 | 0% | 2,048 | 1,883 | -8% | 0 | 0 | — |
case-16 | fail→pass | 15,892 | 10,586 | -33% | 1 | 1 | 0% | 2,822 | 2,365 | -16% | 0 | 0 | — |
case-17 | pass→pass | 10,301 | 8,542 | -17% | 1 | 1 | 0% | 2,239 | 2,503 | +12% | 0 | 0 | — |
case-18 | pass→pass | 10,520 | 15,694 | +49% | 1 | 1 | 0% | 2,383 | 4,346 | +82% | 0 | 0 | — |
case-19 | pass→pass | 10,141 | 6,863 | -32% | 1 | 1 | 0% | 2,004 | 1,806 | -10% | 0 | 0 | — |
case-20 | pass→pass | 7,371 | 9,112 | +24% | 1 | 1 | 0% | 1,774 | 2,807 | +58% | 0 | 0 | — |
case-21 | pass→pass | 13,865 | 12,855 | -7% | 1 | 1 | 0% | 3,153 | 3,405 | +8% | 0 | 0 | — |
case-22 | pass→pass | 4,274 | 4,347 | +2% | 1 | 1 | 0% | 929 | 1,468 | +58% | 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 +41 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.