Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when listing, showing, searching, creating, testing, or removing MaxCompute UDFs and UDF resources.
.claude/skills/aliyun-udf/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | -43% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -67% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -67% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -69% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -32% | 0% |
Use this workflow for MaxCompute UDF lifecycle work.
bashmcs udf list mcs udf search KEYWORD mcs udf show NAME mcs udf create NAME --inline-python script.py mcs udf test NAME --args '1, "abc"' mcs udf remove NAME mcs udf resource list mcs udf resource show NAME mcs udf resource remove NAME
mcs udf test --args accepts SQL literals only: numeric values, NULL, TRUE/FALSE, and quoted strings. Do not pass identifiers, function calls, expressions, or semicolon-separated fragments as test arguments.
For ordinary SQL querying, do not manage UDFs unless the question requires a function that is absent or broken.
See references/udf.md.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 8,208 | 4,500 | -45% | 1 | 1 | 0% | 1,940 | 737 | -62% | 0 | 0 | — |
case-02 | fail→fail | 5,139 | 4,455 | -13% | 1 | 1 | 0% | 953 | 429 | -55% | 0 | 0 | — |
case-03 | fail→pass | 6,205 | 5,020 | -19% | 1 | 1 | 0% | 626 | 356 | -43% | 0 | 0 | — |
case-04 | fail→pass | 5,949 | 1,159 | -81% | 1 | 1 | 0% | 989 | 326 | -67% | 0 | 0 | — |
case-05 | fail→pass | 6,357 | 1,581 | -75% | 1 | 1 | 0% | 1,106 | 362 | -67% | 0 | 0 | — |
case-06 | fail→pass | 5,961 | 1,357 | -77% | 1 | 1 | 0% | 995 | 312 | -69% | 0 | 0 | — |
case-07 | fail→pass | 3,126 | 1,142 | -63% | 1 | 1 | 0% | 511 | 346 | -32% | 0 | 0 | — |
case-08 | fail→pass | 8,892 | 3,521 | -60% | 1 | 1 | 0% | 1,540 | 449 | -71% | 0 | 0 | — |
case-09 | fail→pass | 7,451 | 3,373 | -55% | 1 | 1 | 0% | 1,322 | 806 | -39% | 0 | 0 | — |
case-10 | pass→pass | 7,019 | 5,894 | -16% | 1 | 1 | 0% | 1,341 | 1,190 | -11% | 0 | 0 | — |
case-11 | pass→pass | 8,101 | 1,820 | -78% | 1 | 1 | 0% | 1,501 | 395 | -74% | 0 | 0 | — |
case-12 | pass→pass | 4,434 | 2,104 | -53% | 1 | 1 | 0% | 869 | 540 | -38% | 0 | 0 | — |
case-13 | fail→pass | 13,552 | 4,129 | -70% | 1 | 1 | 0% | 2,375 | 931 | -61% | 0 | 0 | — |
case-14 | pass→pass | 9,781 | 4,125 | -58% | 1 | 1 | 0% | 1,766 | 919 | -48% | 0 | 0 | — |
case-15 | fail→pass | 8,844 | 2,577 | -71% | 1 | 1 | 0% | 1,594 | 647 | -59% | 0 | 0 | — |
case-16 | fail→pass | 7,115 | 1,919 | -73% | 1 | 1 | 0% | 1,324 | 444 | -66% | 0 | 0 | — |
case-17 | fail→pass | 7,023 | 1,827 | -74% | 1 | 1 | 0% | 1,252 | 438 | -65% | 0 | 0 | — |
case-18 | fail→pass | 5,274 | 1,913 | -64% | 1 | 1 | 0% | 838 | 458 | -45% | 0 | 0 | — |
case-19 | fail→pass | 3,308 | 1,331 | -60% | 1 | 1 | 0% | 520 | 368 | -29% | 0 | 0 | — |
case-20 | fail→pass | 4,776 | 1,505 | -68% | 1 | 1 | 0% | 659 | 390 | -41% | 0 | 0 | — |
case-21 | fail→pass | 9,817 | 1,717 | -83% | 1 | 1 | 0% | 1,510 | 465 | -69% | 0 | 0 | — |
case-22 | fail→pass | 14,964 | 3,793 | -75% | 1 | 1 | 0% | 1,572 | 406 | -74% | 0 | 0 | — |
case-23 | fail→pass | 6,816 | 2,019 | -70% | 1 | 1 | 0% | 1,125 | 472 | -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. 23 cases were attempted. The headline lift of +74 percentage points is the difference between those two pass rates over the 23 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.