Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Set up a Superset open-source contribution — fork and clone superset-sh/superset, run local dev setup, and follow the repo's contribution rules through to a merge-ready PR. Use when the user wants to contribute to Superset, fix a Superset bug themselves, or prepare a PR against superset-sh/superset.
.claude/skills/superset-sh-contribute/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | -15% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 6% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -15% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -28% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -51% | 0% |
Take the user from "I want to fix/build X in Superset" to a merge-ready PR that follows the repo's rules. If the checked-out repo has CONTRIBUTING.md / DEVELOPMENT.md / AGENTS.md, those files are authoritative — read them and prefer them over this summary.
gh auth status, then fork and clone: gh repo fork superset-sh/superset --clone (or add a fork remote to an existing clone)./.superset/setup.local.sh once (configures per-workspace ports, app identity, local services, and a seeded dev account — no external credentials needed), then bun run devAGENTS.md and follow it.main; one change per PR — unrelated finds become a second PRbun run lint:fix, then verify bun run lint exits clean (CI fails on warnings too), bun run typecheck, bun run testfeat(desktop): ..., fix(web): ...) — PRs are squash-merged with the title as the commit subjectgh pr create against superset-sh/superset main, fill in the PR template honestly (what you ran, what you clicked, what's covered by tests), and report the PR URL back to the user.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-02 | fail→pass | 22,649 | 16,670 | -26% | 1 | 1 | 0% | 3,967 | 3,383 | -15% | 0 | 0 | — |
case-01 | fail→fail | 2,690 | 6,869 | +155% | 1 | 1 | 0% | 252 | 687 | +173% | 0 | 0 | — |
case-03 | fail→pass | 9,920 | 9,955 | +0% | 1 | 1 | 0% | 1,968 | 2,085 | +6% | 0 | 0 | — |
case-04 | fail→fail | 6,307 | 3,419 | -46% | 1 | 1 | 0% | 1,057 | 1,058 | +0% | 0 | 0 | — |
case-05 | fail→pass | 9,142 | 4,254 | -53% | 1 | 1 | 0% | 1,578 | 1,347 | -15% | 0 | 0 | — |
case-06 | fail→pass | 8,802 | 3,330 | -62% | 1 | 1 | 0% | 1,571 | 1,130 | -28% | 0 | 0 | — |
case-07 | fail→pass | 10,847 | 2,264 | -79% | 1 | 1 | 0% | 1,854 | 913 | -51% | 0 | 0 | — |
case-08 | fail→pass | 7,243 | 2,063 | -72% | 1 | 1 | 0% | 1,299 | 876 | -33% | 0 | 0 | — |
case-09 | fail→pass | 4,562 | 1,823 | -60% | 1 | 1 | 0% | 732 | 761 | +4% | 0 | 0 | — |
case-10 | fail→pass | 6,534 | 1,883 | -71% | 1 | 1 | 0% | 1,154 | 824 | -29% | 0 | 0 | — |
case-11 | pass→pass | 11,664 | 4,253 | -64% | 1 | 1 | 0% | 1,862 | 1,225 | -34% | 0 | 0 | — |
case-21 | fail→fail | 8,303 | 4,320 | -48% | 1 | 1 | 0% | 1,280 | 1,213 | -5% | 0 | 0 | — |
case-12 | pass→pass | 10,317 | 5,115 | -50% | 1 | 1 | 0% | 1,666 | 1,336 | -20% | 0 | 0 | — |
case-13 | pass→pass | 14,752 | 6,572 | -55% | 1 | 1 | 0% | 2,228 | 1,492 | -33% | 0 | 0 | — |
case-14 | fail→pass | 13,663 | 10,168 | -26% | 1 | 1 | 0% | 2,130 | 2,086 | -2% | 0 | 0 | — |
case-15 | pass→pass | 9,234 | 3,823 | -59% | 1 | 1 | 0% | 1,443 | 1,143 | -21% | 0 | 0 | — |
case-20 | pass→pass | 2,776 | 1,863 | -33% | 1 | 1 | 0% | 394 | 793 | +101% | 0 | 0 | — |
case-16 | pass→pass | 4,410 | 1,968 | -55% | 1 | 1 | 0% | 702 | 817 | +16% | 0 | 0 | — |
case-17 | pass→pass | 10,128 | 5,325 | -47% | 1 | 1 | 0% | 1,601 | 1,435 | -10% | 0 | 0 | — |
case-18 | fail→pass | 7,401 | 2,951 | -60% | 1 | 1 | 0% | 1,252 | 1,059 | -15% | 0 | 0 | — |
case-19 | pass→pass | 5,345 | 1,774 | -67% | 1 | 1 | 0% | 848 | 847 | -0% | 0 | 0 | — |
case-22 | fail→fail | 15,160 | 6,697 | -56% | 1 | 1 | 0% | 2,680 | 979 | -63% | 0 | 0 | — |
case-23 | fail→fail | 8,778 | 9,551 | +9% | 1 | 1 | 0% | 1,489 | 2,094 | +41% | 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, and 21 counted toward the lift figure. The other 2 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 +43 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.