Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Guided scenario for creating an Open Design plugin folder that can be installed into My plugins.
.claude/skills/nexu-io-od-plugin-authoring/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-13 | ✗→✓ | ▲ Improved | 26% | 0% |
| case-01 | ✗→✓ | ▲ Improved | -36% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 16% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -20% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 32% | 0% |
Use this scenario when the user wants to create their own OpenDesign plugin.
Produce a folder named generated-plugin/ in the active project workspace. At minimum, the folder must contain:
SKILL.md with frontmatter and clear agent instructions.open-design.json with valid plugin metadata, od.kind, mode, task kind, capabilities, and inputs when needed.Add examples/, assets/, or other supporting files only when they help the plugin be used or reviewed.
docs/plugins-spec.md and the schema at docs/schemas/open-design.plugin.v1.json.SKILL.md as the canonical behavior description. open-design.json should describe how OpenDesign installs, applies, and presents that behavior.textgenerated-plugin/ SKILL.md open-design.json examples/ assets/
The examples/ and assets/ directories are optional. Do not create empty directories just to match the sketch.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-12 | pass→pass | 12,082 | 4,276 | -65% | 1 | 1 | 0% | 1,703 | 1,038 | -39% | 0 | 0 | — |
case-13 | fail→pass | 11,046 | 12,722 | +15% | 1 | 1 | 0% | 2,393 | 3,027 | +26% | 0 | 0 | — |
case-01 | fail→pass | 19,096 | 11,223 | -41% | 1 | 1 | 0% | 3,993 | 2,554 | -36% | 0 | 0 | — |
case-02 | fail→pass | 18,123 | 21,128 | +17% | 1 | 1 | 0% | 3,929 | 4,571 | +16% | 0 | 0 | — |
case-03 | fail→pass | 23,697 | 17,181 | -27% | 1 | 1 | 0% | 4,894 | 3,895 | -20% | 0 | 0 | — |
case-04 | pass→fail | 11,551 | 13,525 | +17% | 1 | 1 | 0% | 1,886 | 2,942 | +56% | 0 | 0 | — |
case-05 | pass→fail | 15,018 | 18,852 | +26% | 1 | 1 | 0% | 2,266 | 3,738 | +65% | 0 | 0 | — |
case-06 | pass→pass | 10,837 | 10,199 | -6% | 1 | 1 | 0% | 1,735 | 2,063 | +19% | 0 | 0 | — |
case-07 | fail→fail | 16,343 | 11,131 | -32% | 1 | 1 | 0% | 3,441 | 2,554 | -26% | 0 | 0 | — |
case-08 | fail→pass | 17,805 | 34,093 | +91% | 1 | 1 | 0% | 2,997 | 3,970 | +32% | 0 | 0 | — |
case-09 | fail→pass | 19,602 | 17,573 | -10% | 1 | 1 | 0% | 4,234 | 3,682 | -13% | 0 | 0 | — |
case-10 | fail→fail | 14,882 | 47,608 | +220% | 1 | 1 | 0% | 2,656 | 3,546 | +34% | 0 | 0 | — |
case-11 | fail→fail | 33,619 | 12,865 | -62% | 1 | 1 | 0% | 7,685 | 2,935 | -62% | 0 | 0 | — |
case-14 | fail→pass | 9,676 | 12,691 | +31% | 1 | 1 | 0% | 1,758 | 2,788 | +59% | 0 | 0 | — |
case-15 | pass→pass | 24,960 | 16,451 | -34% | 1 | 1 | 0% | 5,422 | 3,496 | -36% | 0 | 0 | — |
case-16 | pass→pass | 11,599 | 15,460 | +33% | 1 | 1 | 0% | 2,508 | 3,631 | +45% | 0 | 0 | — |
case-17 | pass→pass | 15,979 | 18,480 | +16% | 1 | 1 | 0% | 2,900 | 4,147 | +43% | 0 | 0 | — |
case-18 | pass→pass | 13,059 | 13,114 | +0% | 1 | 1 | 0% | 2,186 | 2,779 | +27% | 0 | 0 | — |
case-19 | pass→pass | 18,666 | 23,987 | +29% | 1 | 1 | 0% | 3,829 | 5,139 | +34% | 0 | 0 | — |
case-20 | pass→pass | 9,889 | 2,943 | -70% | 1 | 1 | 0% | 1,525 | 816 | -46% | 0 | 0 | — |
case-21 | fail→pass | 8,819 | 1,544 | -82% | 1 | 1 | 0% | 1,450 | 600 | -59% | 0 | 0 | — |
case-22 | fail→pass | 9,985 | 2,201 | -78% | 1 | 1 | 0% | 1,626 | 762 | -53% | 0 | 0 | — |
case-23 | fail→fail | 22,539 | 14,330 | -36% | 1 | 1 | 0% | 4,637 | 3,194 | -31% | 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 +30 percentage points is the difference between those two pass rates over the 23 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.