Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Integrate 綠界 (ECPay) for Taiwan e-commerce — credit card, ATM, CVS 代碼, 超商取貨付款, and CheckMacValue signature. Use when ECPay is the chosen PSP, setting up logistics-payment combined flows (超取+COD), computing or verifying CheckMacValue, or handling ECPay callbacks. Do NOT use for gateway selection or comparison (see tw-payment-integration). Do NOT use for non-ECPay providers. STATUS: SKELETON — body pending.
.claude/skills/asgard-ai-platform-tw-ecom-payment-ecpay/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-05 | ✗→✓ | ▲ Improved | -8% | 0% |
| case-01 | ✓→✗ | ▼ Worse | 11% | 0% |
| case-02 | ✓→✓ | ＝ Same ✓ | -12% | 0% |
| case-03 | ✓→✓ | ＝ Same ✓ | 32% | 0% |
| case-04 | ✓→✓ | ＝ Same ✓ | 83% | 0% |
> STATUS: SKELETON — body pending.
tw-ecom-logistics-cvsTODO: ECPay positioning, CheckMacValue signature, the 超取+COD combined model.
TODO: when ECPay's combined flow beats separating payment vs shipping.
TODO: merchant setup, first txn, COD-at-pickup reconciliation.
TODO: 5-6 pitfalls (CheckMacValue param ordering / URL encoding gotchas, COD settlement offset, 超取 failure refund flow, cross-sandbox credential leakage, 境外 card compatibility).
TODO.
TODO.
tw-payment-integrationtw-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 | pass→fail | 47,638 | 50,383 | +6% | 1 | 1 | 0% | 2,923 | 3,248 | +11% | 0 | 0 | — |
case-02 | pass→pass | 50,352 | 17,791 | -65% | 1 | 1 | 0% | 3,187 | 2,790 | -12% | 0 | 0 | — |
case-03 | pass→pass | 11,409 | 43,212 | +279% | 1 | 1 | 0% | 1,930 | 2,538 | +32% | 0 | 0 | — |
case-04 | pass→pass | 4,055 | 5,968 | +47% | 1 | 1 | 0% | 637 | 1,164 | +83% | 0 | 0 | — |
case-05 | fail→pass | 6,488 | 4,471 | -31% | 1 | 1 | 0% | 1,052 | 967 | -8% | 0 | 0 | — |
case-06 | pass→pass | 11,830 | 9,377 | -21% | 1 | 1 | 0% | 1,950 | 2,054 | +5% | 0 | 0 | — |
case-07 | pass→pass | 8,773 | 7,464 | -15% | 1 | 1 | 0% | 1,530 | 1,485 | -3% | 0 | 0 | — |
case-08 | pass→pass | 4,978 | 3,613 | -27% | 1 | 1 | 0% | 858 | 948 | +10% | 0 | 0 | — |
case-09 | pass→pass | 12,241 | 38,791 | +217% | 1 | 1 | 0% | 1,825 | 1,891 | +4% | 0 | 0 | — |
case-10 | pass→pass | 6,530 | 4,130 | -37% | 1 | 1 | 0% | 1,085 | 1,095 | +1% | 0 | 0 | — |
case-11 | pass→pass | 5,728 | 3,798 | -34% | 1 | 1 | 0% | 973 | 904 | -7% | 0 | 0 | — |
case-12 | pass→pass | 3,844 | 3,666 | -5% | 1 | 1 | 0% | 665 | 883 | +33% | 0 | 0 | — |
case-13 | pass→pass | 4,712 | 2,919 | -38% | 1 | 1 | 0% | 803 | 854 | +6% | 0 | 0 | — |
case-14 | pass→pass | 8,524 | 9,938 | +17% | 1 | 1 | 0% | 1,553 | 1,888 | +22% | 0 | 0 | — |
case-15 | pass→pass | 7,164 | 6,205 | -13% | 1 | 1 | 0% | 1,283 | 1,369 | +7% | 0 | 0 | — |
case-16 | pass→pass | 7,803 | 6,925 | -11% | 1 | 1 | 0% | 1,436 | 1,305 | -9% | 0 | 0 | — |
case-17 | fail→fail | 10,803 | 8,693 | -20% | 1 | 1 | 0% | 2,062 | 1,961 | -5% | 0 | 0 | — |
case-18 | pass→pass | 6,853 | 6,130 | -11% | 1 | 1 | 0% | 1,177 | 1,326 | +13% | 0 | 0 | — |
case-19 | pass→pass | 7,474 | 8,436 | +13% | 1 | 1 | 0% | 1,411 | 1,503 | +7% | 0 | 0 | — |
case-20 | pass→pass | 4,655 | 3,534 | -24% | 1 | 1 | 0% | 806 | 822 | +2% | 0 | 0 | — |
case-21 | pass→pass | 6,098 | 7,530 | +23% | 1 | 1 | 0% | 1,002 | 1,595 | +59% | 0 | 0 | — |
case-22 | pass→pass | 7,035 | 7,970 | +13% | 1 | 1 | 0% | 1,088 | 1,765 | +62% | 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.