Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Open a pull request adding a local Open Design plugin to the Open Design community catalog using gh CLI.
.claude/skills/nexu-io-od-plugin-contribute-open-design/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | 7% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 77% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -71% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -57% | 0% |
| case-16 | ✗→✓ | ▲ Improved | -50% | 0% |
Use this workflow when the active project contains a copied plugin folder and the user wants to propose it for the OpenDesign community catalog.
plugin_context_path is the copied plugin folder relative to the project working directory.curl -sS -X POST "$OD_DAEMON_URL/api/projects/$OD_PROJECT_ID/plugins/contribute-open-design" -H 'content-type: application/json' -d '{"path":"<plugin_context_path>"}'
ok is true, report the PR URL, branch name when present in the log, and any useful validation summary.message, code, and useful log lines. When authentication is missing, tell the user to run gh auth login --hostname github.com.Keep the pull request focused. Do not modify unrelated OpenDesign files unless a manifest validation issue requires a tiny supporting change.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 37,412 | 67,437 | +80% | 1 | 1 | 0% | 1,157 | 721 | -38% | 0 | 0 | — |
case-02 | fail→fail | 79,601 | 6,541 | -92% | 1 | 1 | 0% | 2,424 | 638 | -74% | 0 | 0 | — |
case-03 | fail→fail | 5,011 | 5,987 | +19% | 1 | 1 | 0% | 172 | 569 | +231% | 0 | 0 | — |
case-04 | fail→pass | 10,018 | 37,511 | +274% | 1 | 1 | 0% | 1,497 | 1,595 | +7% | 0 | 0 | — |
case-05 | pass→pass | 8,706 | 4,044 | -54% | 1 | 1 | 0% | 1,414 | 1,019 | -28% | 0 | 0 | — |
case-06 | pass→pass | 9,417 | 3,600 | -62% | 1 | 1 | 0% | 1,622 | 891 | -45% | 0 | 0 | — |
case-07 | fail→pass | 2,137 | 2,891 | +35% | 1 | 1 | 0% | 267 | 473 | +77% | 0 | 0 | — |
case-08 | fail→pass | 11,589 | 1,757 | -85% | 1 | 1 | 0% | 1,776 | 509 | -71% | 0 | 0 | — |
case-09 | pass→pass | 8,747 | 2,748 | -69% | 1 | 1 | 0% | 1,241 | 645 | -48% | 0 | 0 | — |
case-10 | fail→pass | 11,537 | 3,355 | -71% | 1 | 1 | 0% | 2,017 | 858 | -57% | 0 | 0 | — |
case-11 | pass→pass | 6,358 | 3,657 | -42% | 1 | 1 | 0% | 945 | 861 | -9% | 0 | 0 | — |
case-12 | pass→pass | 8,773 | 9,012 | +3% | 1 | 1 | 0% | 1,385 | 1,682 | +21% | 0 | 0 | — |
case-13 | pass→pass | 12,814 | 8,756 | -32% | 1 | 1 | 0% | 1,957 | 1,626 | -17% | 0 | 0 | — |
case-14 | pass→pass | 2,561 | 32,508 | +1169% | 1 | 1 | 0% | 319 | 584 | +83% | 0 | 0 | — |
case-15 | pass→pass | 13,294 | 6,522 | -51% | 1 | 1 | 0% | 2,041 | 1,267 | -38% | 0 | 0 | — |
case-16 | fail→pass | 11,350 | 33,988 | +199% | 1 | 1 | 0% | 1,834 | 924 | -50% | 0 | 0 | — |
case-17 | pass→pass | 2,437 | 2,124 | -13% | 1 | 1 | 0% | 382 | 622 | +63% | 0 | 0 | — |
case-18 | pass→pass | 6,496 | 2,563 | -61% | 1 | 1 | 0% | 1,038 | 567 | -45% | 0 | 0 | — |
case-19 | pass→pass | 11,119 | 2,538 | -77% | 1 | 1 | 0% | 1,535 | 632 | -59% | 0 | 0 | — |
case-20 | fail→pass | 12,757 | 2,522 | -80% | 1 | 1 | 0% | 2,109 | 682 | -68% | 0 | 0 | — |
case-21 | pass→pass | 2,209 | 1,822 | -18% | 1 | 1 | 0% | 303 | 513 | +69% | 0 | 0 | — |
case-22 | pass→pass | 9,535 | 3,629 | -62% | 1 | 1 | 0% | 1,450 | 845 | -42% | 0 | 0 | — |
case-23 | pass→pass | 4,988 | 2,517 | -50% | 1 | 1 | 0% | 810 | 686 | -15% | 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 +26 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.