Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Phase 1 sample plugin synthesizing a SKILL.md frontmatter for backwards-compat tests.
.claude/skills/nexu-io-sample-plugin/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 226% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -46% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 114% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 70% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 47% | 0% |
This is the SKILL.md half of the Phase 1 e2e fixture. The companion open-design.json sidecar carries the canonical OpenDesign plugin manifest fields; this file proves the SKILL-only adapter path stays honest when an install lacks an explicit sidecar (just delete open-design.json to test the legacy compat tier).
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 10,360 | 54,399 | +425% | 1 | 1 | 0% | 1,482 | 4,837 | +226% | 0 | 0 | — |
case-02 | fail→pass | 40,115 | 50,650 | +26% | 1 | 1 | 0% | 8,232 | 4,456 | -46% | 0 | 0 | — |
case-03 | fail→fail | 60,138 | 31,232 | -48% | 1 | 1 | 0% | 4,690 | 6,560 | +40% | 0 | 0 | — |
case-04 | fail→pass | 38,035 | 13,794 | -64% | 1 | 1 | 0% | 1,246 | 2,664 | +114% | 0 | 0 | — |
case-05 | fail→pass | 9,080 | 11,166 | +23% | 1 | 1 | 0% | 1,243 | 2,115 | +70% | 0 | 0 | — |
case-06 | fail→fail | 16,500 | 27,748 | +68% | 1 | 1 | 0% | 3,049 | 5,530 | +81% | 0 | 0 | — |
case-07 | fail→fail | 4,413 | 35,873 | +713% | 1 | 1 | 0% | 613 | 1,095 | +79% | 0 | 0 | — |
case-08 | fail→pass | 12,782 | 17,150 | +34% | 1 | 1 | 0% | 1,958 | 2,883 | +47% | 0 | 0 | — |
case-09 | fail→fail | 42,690 | 10,424 | -76% | 1 | 1 | 0% | 8,222 | 1,659 | -80% | 0 | 0 | — |
case-10 | fail→fail | 11,974 | 8,082 | -33% | 1 | 1 | 0% | 2,174 | 1,494 | -31% | 0 | 0 | — |
case-11 | fail→pass | 40,107 | 16,306 | -59% | 1 | 1 | 0% | 8,220 | 2,760 | -66% | 0 | 0 | — |
case-18 | fail→fail | 39,390 | 8,267 | -79% | 1 | 1 | 0% | 2,185 | 1,479 | -32% | 0 | 0 | — |
case-12 | fail→fail | 3,423 | 10,992 | +221% | 1 | 1 | 0% | 428 | 2,017 | +371% | 0 | 0 | — |
case-13 | fail→pass | 12,888 | 22,071 | +71% | 1 | 1 | 0% | 2,230 | 3,883 | +74% | 0 | 0 | — |
case-14 | fail→pass | 5,370 | 14,188 | +164% | 1 | 1 | 0% | 763 | 2,448 | +221% | 0 | 0 | — |
case-15 | fail→fail | 31,263 | 26,937 | -14% | 1 | 1 | 0% | 5,870 | 5,328 | -9% | 0 | 0 | — |
case-16 | fail→fail | 8,189 | 9,232 | +13% | 1 | 1 | 0% | 1,313 | 1,656 | +26% | 0 | 0 | — |
case-17 | fail→fail | 12,921 | 14,565 | +13% | 1 | 1 | 0% | 1,859 | 2,574 | +38% | 0 | 0 | — |
case-19 | fail→fail | 36,277 | 14,682 | -60% | 1 | 1 | 0% | 5,478 | 2,379 | -57% | 0 | 0 | — |
case-20 | fail→fail | 8,200 | 6,885 | -16% | 1 | 1 | 0% | 1,135 | 1,274 | +12% | 0 | 0 | — |
case-21 | pass→pass | 10,248 | 9,315 | -9% | 1 | 1 | 0% | 1,887 | 1,857 | -2% | 0 | 0 | — |
case-22 | pass→pass | 17,794 | 14,527 | -18% | 1 | 1 | 0% | 2,504 | 2,305 | -8% | 0 | 0 | — |
case-23 | pass→fail | 15,362 | 14,781 | -4% | 1 | 1 | 0% | 2,567 | 2,781 | +8% | 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. 1 case got worse with the skill loaded, and it is 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.