Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →163 邮箱真值集构建、审计与跑批测试。当任务涉及 163 邮箱、网易邮箱、163 真值集、163 跑批时触发。包含 163 IMAP 特殊处理（ID 命令）、已知分类陷阱、已验证案例。
.claude/skills/ethanyoq-163-email-ground-truth/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | -24% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 86% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -49% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -49% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -65% | 0% |
当任务涉及以下关键词时自动触发：
references/workflow-163.md — 163 专属工作流程references/pitfalls-163.md — 163 踩坑记录references/case-study-163.md — 本次 41 张已验证案例163 和 QQ 共用同一套核心处理逻辑，本 Skill 只记录 163 的差异点。 通用规则请参考 qq-email-ground-truth Skill。
.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-06 | pass→pass | 11,950 | 7,249 | -39% | 1 | 1 | 0% | 2,061 | 1,510 | -27% | 0 | 0 | — |
case-01 | fail→fail | 20,894 | 2,816 | -87% | 1 | 1 | 0% | 3,765 | 829 | -78% | 0 | 0 | — |
case-02 | fail→pass | 22,113 | 15,436 | -30% | 1 | 1 | 0% | 4,538 | 3,468 | -24% | 0 | 0 | — |
case-03 | fail→fail | 15,457 | 6,875 | -56% | 1 | 1 | 0% | 2,425 | 962 | -60% | 0 | 0 | — |
case-04 | pass→pass | 9,922 | 5,293 | -47% | 1 | 1 | 0% | 1,934 | 1,465 | -24% | 0 | 0 | — |
case-05 | pass→pass | 9,958 | 7,448 | -25% | 1 | 1 | 0% | 1,670 | 1,695 | +1% | 0 | 0 | — |
case-07 | fail→pass | 7,191 | 2,149 | -70% | 1 | 1 | 0% | 478 | 889 | +86% | 0 | 0 | — |
case-08 | fail→pass | 7,699 | 1,628 | -79% | 1 | 1 | 0% | 1,331 | 678 | -49% | 0 | 0 | — |
case-09 | fail→pass | 9,515 | 2,469 | -74% | 1 | 1 | 0% | 1,598 | 821 | -49% | 0 | 0 | — |
case-10 | fail→pass | 10,837 | 1,496 | -86% | 1 | 1 | 0% | 1,873 | 650 | -65% | 0 | 0 | — |
case-11 | fail→pass | 15,428 | 1,884 | -88% | 1 | 1 | 0% | 2,468 | 698 | -72% | 0 | 0 | — |
case-12 | fail→pass | 10,655 | 1,385 | -87% | 1 | 1 | 0% | 1,821 | 655 | -64% | 0 | 0 | — |
case-13 | pass→pass | 13,673 | 8,838 | -35% | 1 | 1 | 0% | 2,537 | 2,053 | -19% | 0 | 0 | — |
case-14 | pass→pass | 11,762 | 5,530 | -53% | 1 | 1 | 0% | 1,955 | 1,376 | -30% | 0 | 0 | — |
case-15 | fail→pass | 17,788 | 12,577 | -29% | 1 | 1 | 0% | 2,741 | 2,514 | -8% | 0 | 0 | — |
case-16 | fail→pass | 3,467 | 2,071 | -40% | 1 | 1 | 0% | 496 | 702 | +42% | 0 | 0 | — |
case-17 | fail→pass | 8,720 | 7,540 | -14% | 1 | 1 | 0% | 1,488 | 1,684 | +13% | 0 | 0 | — |
case-18 | fail→fail | 12,775 | 7,368 | -42% | 1 | 1 | 0% | 2,307 | 1,610 | -30% | 0 | 0 | — |
case-19 | fail→pass | 4,082 | 1,664 | -59% | 1 | 1 | 0% | 596 | 633 | +6% | 0 | 0 | — |
case-20 | pass→pass | 6,053 | 2,625 | -57% | 1 | 1 | 0% | 1,114 | 843 | -24% | 0 | 0 | — |
case-21 | pass→pass | 14,209 | 15,740 | +11% | 1 | 1 | 0% | 2,931 | 3,748 | +28% | 0 | 0 | — |
case-22 | pass→pass | 13,716 | 11,572 | -16% | 1 | 1 | 0% | 2,873 | 2,895 | +1% | 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 +50 percentage points is the difference between those two pass rates over the 21 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.