Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Ship via Taiwan home delivery carriers — 黑貓宅急便, 宅配通, 新竹物流, 郵局. Use when setting up home delivery, choosing carriers by region / parcel size, integrating label printing, or handling redelivery. Do NOT use for CVS pickup (`tw-ecom-logistics-cvs`). STATUS: SKELETON — body pending.
.claude/skills/asgard-ai-platform-tw-ecom-logistics-home/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-12 | ✗→✓ | ▲ Improved | -6% | 0% |
| case-05 | ✓→✗ | ▼ Worse | 16% | 0% |
| case-02 | ✓→✓ | ＝ Same ✓ | 11% | 0% |
| case-03 | ✓→✓ | ＝ Same ✓ | -15% | 0% |
| case-04 | ✓→✓ | ＝ Same ✓ | -7% | 0% |
> STATUS: SKELETON — body pending.
tw-ecom-logistics-cvstw-ecom-logistics-cold-chainTODO: carrier pricing by 材積 vs weight, regional coverage differences, SLA promises.
TODO: carrier pick by region / weight / 時效.
TODO: label API, batch upload, tracking webhook.
TODO: 5-6 pitfalls (材積-based upcharge, 離島 surcharge opacity, tracking update lag, failed delivery handling, fragile-item insurance).
TODO.
TODO.
tw-ecom-logistics-cvs_Last verified: 2026-04_
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 57,001 | 51,582 | -10% | 1 | 1 | 0% | 3,703 | 3,987 | +8% | 0 | 0 | — |
case-02 | pass→pass | 19,805 | 45,910 | +132% | 1 | 1 | 0% | 2,449 | 2,712 | +11% | 0 | 0 | — |
case-03 | pass→pass | 8,328 | 6,921 | -17% | 1 | 1 | 0% | 1,371 | 1,167 | -15% | 0 | 0 | — |
case-04 | pass→pass | 21,816 | 52,367 | +140% | 1 | 1 | 0% | 3,601 | 3,358 | -7% | 0 | 0 | — |
case-05 | pass→fail | 81,721 | 45,679 | -44% | 1 | 1 | 0% | 2,611 | 3,026 | +16% | 0 | 0 | — |
case-06 | pass→pass | 20,390 | 28,268 | +39% | 1 | 1 | 0% | 2,718 | 4,849 | +78% | 0 | 0 | — |
case-07 | pass→pass | 13,047 | 12,394 | -5% | 1 | 1 | 0% | 2,508 | 2,486 | -1% | 0 | 0 | — |
case-08 | pass→pass | 17,568 | 13,186 | -25% | 1 | 1 | 0% | 2,605 | 2,391 | -8% | 0 | 0 | — |
case-09 | pass→pass | 19,517 | 18,027 | -8% | 1 | 1 | 0% | 2,561 | 2,561 | 0% | 0 | 0 | — |
case-10 | pass→pass | 17,651 | 16,367 | -7% | 1 | 1 | 0% | 2,476 | 3,145 | +27% | 0 | 0 | — |
case-11 | pass→pass | 9,973 | 6,324 | -37% | 1 | 1 | 0% | 1,347 | 1,293 | -4% | 0 | 0 | — |
case-12 | fail→pass | 15,023 | 13,886 | -8% | 1 | 1 | 0% | 2,265 | 2,124 | -6% | 0 | 0 | — |
case-13 | pass→pass | 14,108 | 16,273 | +15% | 1 | 1 | 0% | 2,381 | 2,974 | +25% | 0 | 0 | — |
case-14 | pass→pass | 19,944 | 24,460 | +23% | 1 | 1 | 0% | 2,726 | 4,124 | +51% | 0 | 0 | — |
case-15 | pass→pass | 14,526 | 12,654 | -13% | 1 | 1 | 0% | 2,180 | 2,140 | -2% | 0 | 0 | — |
case-16 | pass→pass | 15,922 | 14,367 | -10% | 1 | 1 | 0% | 2,463 | 2,224 | -10% | 0 | 0 | — |
case-17 | pass→pass | 14,718 | 21,030 | +43% | 1 | 1 | 0% | 2,102 | 3,370 | +60% | 0 | 0 | — |
case-18 | pass→pass | 18,457 | 20,422 | +11% | 1 | 1 | 0% | 2,923 | 3,328 | +14% | 0 | 0 | — |
case-19 | pass→pass | 17,358 | 19,144 | +10% | 1 | 1 | 0% | 2,681 | 2,773 | +3% | 0 | 0 | — |
case-20 | pass→pass | 15,860 | 16,533 | +4% | 1 | 1 | 0% | 2,688 | 2,976 | +11% | 0 | 0 | — |
case-21 | pass→pass | 10,195 | 15,455 | +52% | 1 | 1 | 0% | 1,715 | 2,368 | +38% | 0 | 0 | — |
case-22 | pass→pass | 15,778 | 15,218 | -4% | 1 | 1 | 0% | 2,269 | 2,557 | +13% | 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 0 percentage points is the difference between those two pass rates over the 22 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
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.