Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →跑批测试标准与双邮箱回归门禁。当任务涉及"跑批"、"batch test"、"真值审计"、"回归验证"、"regression"时触发。确保每次跑批都遵守 P0/P1/P2 标准，且触碰保护区域的改动必须通过双邮箱验证。
.claude/skills/ethanyoq-email-batch-test/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | -28% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -48% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -72% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -55% | 0% |
| case-12 | ✗→✓ | ▲ Improved | -33% | 0% |
当任务涉及以下关键词时自动触发：
CLAUDE.md — 了解 FROZEN/GUARDED/SAFE 区域定义references/test-standards.md — P0/P1/P2 量化标准references/regression-gate.md — 双邮箱回归门禁流程strict_truth_audit.py 对比真值test_dataset/qq_20260201_20260311_final/truth_manifest.json.claude/worktrees/test-163-batch/test_dataset/163_20260301_20260319/truth_manifest.json| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 13,564 | 3,093 | -77% | 1 | 1 | 0% | 2,145 | 671 | -69% | 0 | 0 | — |
case-02 | fail→fail | 12,287 | 2,918 | -76% | 1 | 1 | 0% | 1,906 | 712 | -63% | 0 | 0 | — |
case-03 | fail→fail | 15,572 | 2,935 | -81% | 1 | 1 | 0% | 2,760 | 825 | -70% | 0 | 0 | — |
case-04 | pass→pass | 9,158 | 7,280 | -21% | 1 | 1 | 0% | 1,763 | 1,794 | +2% | 0 | 0 | — |
case-05 | pass→pass | 10,534 | 8,172 | -22% | 1 | 1 | 0% | 2,022 | 2,023 | +0% | 0 | 0 | — |
case-06 | pass→pass | 16,153 | 10,702 | -34% | 1 | 1 | 0% | 3,051 | 2,377 | -22% | 0 | 0 | — |
case-07 | fail→pass | 14,540 | 7,247 | -50% | 1 | 1 | 0% | 2,304 | 1,663 | -28% | 0 | 0 | — |
case-08 | fail→pass | 8,709 | 2,777 | -68% | 1 | 1 | 0% | 1,508 | 790 | -48% | 0 | 0 | — |
case-09 | pass→pass | 11,281 | 10,032 | -11% | 1 | 1 | 0% | 1,759 | 2,064 | +17% | 0 | 0 | — |
case-10 | fail→pass | 12,402 | 2,327 | -81% | 1 | 1 | 0% | 2,037 | 566 | -72% | 0 | 0 | — |
case-11 | fail→pass | 9,609 | 1,981 | -79% | 1 | 1 | 0% | 1,603 | 714 | -55% | 0 | 0 | — |
case-12 | fail→pass | 11,528 | 4,761 | -59% | 1 | 1 | 0% | 1,715 | 1,144 | -33% | 0 | 0 | — |
case-13 | fail→pass | 18,646 | 6,822 | -63% | 1 | 1 | 0% | 2,821 | 1,451 | -49% | 0 | 0 | — |
case-14 | fail→pass | 12,705 | 2,355 | -81% | 1 | 1 | 0% | 2,048 | 636 | -69% | 0 | 0 | — |
case-15 | fail→pass | 10,998 | 1,790 | -84% | 1 | 1 | 0% | 1,795 | 649 | -64% | 0 | 0 | — |
case-16 | fail→pass | 10,597 | 1,796 | -83% | 1 | 1 | 0% | 1,813 | 643 | -65% | 0 | 0 | — |
case-17 | pass→pass | 11,260 | 3,159 | -72% | 1 | 1 | 0% | 1,828 | 891 | -51% | 0 | 0 | — |
case-18 | pass→pass | 12,779 | 3,747 | -71% | 1 | 1 | 0% | 1,863 | 953 | -49% | 0 | 0 | — |
case-19 | fail→pass | 9,980 | 4,476 | -55% | 1 | 1 | 0% | 1,720 | 1,110 | -35% | 0 | 0 | — |
case-20 | fail→pass | 18,249 | 7,785 | -57% | 1 | 1 | 0% | 2,784 | 1,746 | -37% | 0 | 0 | — |
case-21 | fail→pass | 11,090 | 4,185 | -62% | 1 | 1 | 0% | 1,635 | 1,059 | -35% | 0 | 0 | — |
case-22 | fail→pass | 14,175 | 6,959 | -51% | 1 | 1 | 0% | 2,282 | 1,551 | -32% | 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 +59 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.