Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when the user wants to run SQL against Alibaba Cloud MaxCompute (ODPS), inspect MaxCompute schemas, build or maintain a semantic package, enrich package semantics, manage MaxCompute UDFs, or record verified/failed SQL memory through the `mcs` CLI.
.claude/skills/aliyun-maxcompute-semantic/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | -39% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -60% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -62% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -55% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -78% | 0% |
This installed file is a discovery stub. Before using mcs for real work, load the runtime workflow that matches the task:
| Intent | Load | | --- | --- | | Answer a data question / write SQL / inspect schema for a query | mcs skill get query | | Query when mcs show reports no semantic package | mcs skill get query | | Build / refresh / onboard a semantic package | mcs skill get build | | Review / apply annotation suggestions, semantic-package proposals, or enrich semantics | mcs skill get enrich | | Create, link, edit, or diagnose profiles | mcs skill get onboarding | | Record or recall verified / failed SQL | mcs skill get memory | | Create or manage MaxCompute UDFs | mcs skill get udf | | File an upstream bug / issue against this skill | mcs skill get report-issue |
When answering a data question, producing SQL, inspecting schema for a query, or recovering from a query error, never run mcs build or mcs package propose. If mcs show or mcs status reports no build data / no semantic package, load mcs skill get query and use its cold-start workflow with live mcs meta commands. mcs sql review still runs syntax / dialect checks without a package; use those issues, but expect semantic hints and coverage to be skipped.
Run mcs build only when the user explicitly asks to build, refresh, onboard, or maintain a semantic profile.
bashmcs skill catalog # list available runtime workflows mcs skill get query # load the query workflow mcs skill get query --full # include references
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 11,321 | 4,948 | -56% | 1 | 1 | 0% | 2,173 | 668 | -69% | 0 | 0 | — |
case-02 | fail→fail | 9,301 | 5,941 | -36% | 1 | 1 | 0% | 460 | 642 | +40% | 0 | 0 | — |
case-03 | fail→fail | 7,358 | 4,876 | -34% | 1 | 1 | 0% | 1,453 | 714 | -51% | 0 | 0 | — |
case-04 | fail→pass | 24,425 | 4,695 | -81% | 1 | 1 | 0% | 1,994 | 1,226 | -39% | 0 | 0 | — |
case-05 | fail→pass | 10,108 | 2,436 | -76% | 1 | 1 | 0% | 1,966 | 790 | -60% | 0 | 0 | — |
case-06 | fail→pass | 10,125 | 1,891 | -81% | 1 | 1 | 0% | 1,676 | 644 | -62% | 0 | 0 | — |
case-07 | fail→pass | 7,487 | 1,920 | -74% | 1 | 1 | 0% | 1,453 | 650 | -55% | 0 | 0 | — |
case-08 | fail→pass | 15,361 | 1,790 | -88% | 1 | 1 | 0% | 2,697 | 589 | -78% | 0 | 0 | — |
case-09 | fail→pass | 5,879 | 2,836 | -52% | 1 | 1 | 0% | 976 | 912 | -7% | 0 | 0 | — |
case-10 | fail→pass | 13,743 | 3,770 | -73% | 1 | 1 | 0% | 2,151 | 1,072 | -50% | 0 | 0 | — |
case-11 | fail→pass | 12,548 | 5,610 | -55% | 1 | 1 | 0% | 1,884 | 1,097 | -42% | 0 | 0 | — |
case-12 | fail→pass | 9,455 | 2,099 | -78% | 1 | 1 | 0% | 1,524 | 621 | -59% | 0 | 0 | — |
case-13 | fail→pass | 14,423 | 1,239 | -91% | 1 | 1 | 0% | 2,322 | 581 | -75% | 0 | 0 | — |
case-14 | fail→pass | 16,445 | 2,652 | -84% | 1 | 1 | 0% | 2,770 | 768 | -72% | 0 | 0 | — |
case-15 | fail→pass | 9,772 | 2,008 | -79% | 1 | 1 | 0% | 1,727 | 601 | -65% | 0 | 0 | — |
case-16 | fail→pass | 8,744 | 2,618 | -70% | 1 | 1 | 0% | 1,458 | 786 | -46% | 0 | 0 | — |
case-17 | pass→pass | 7,323 | 5,980 | -18% | 1 | 1 | 0% | 1,206 | 1,417 | +17% | 0 | 0 | — |
case-18 | pass→pass | 6,873 | 7,336 | +7% | 1 | 1 | 0% | 1,410 | 1,816 | +29% | 0 | 0 | — |
case-19 | pass→pass | 6,533 | 5,124 | -22% | 1 | 1 | 0% | 1,149 | 1,069 | -7% | 0 | 0 | — |
case-20 | fail→pass | 4,835 | 2,483 | -49% | 1 | 1 | 0% | 668 | 751 | +12% | 0 | 0 | — |
case-21 | fail→pass | 10,143 | 3,523 | -65% | 1 | 1 | 0% | 1,608 | 1,038 | -35% | 0 | 0 | — |
case-22 | fail→pass | 11,096 | 1,946 | -82% | 1 | 1 | 0% | 1,883 | 648 | -66% | 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 19 counted toward the lift figure. The other 3 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 +73 percentage points is the difference between those two pass rates over the 19 comparable cases. 2 cases got worse with the skill loaded, and they are 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.