Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Default reference pipeline for the figma-migration taskKind — figma-extract → token-map → generate → critique.
.claude/skills/nexu-io-od-figma-migration/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | -3% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 86% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -37% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -45% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -40% | 0% |
Spec §1 / §10.1 / §21.3.1 / §23.3.3: the canonical figma-migration flow. The pipeline is sequenced so each stage's output is the next stage's input — figma-extract writes figma/tree.json, token-map writes token-map/colors.json (etc.), and generate reads both before producing the HTML artifact.
jsonc{ "stages": [ { "id": "extract", "atoms": ["figma-extract"] }, { "id": "tokens", "atoms": ["token-map"] }, { "id": "generate", "atoms": ["file-write", "live-artifact"] }, { "id": "critique", "atoms": ["critique-theater"], "repeat": true, "until": "critique.score>=4 || iterations>=3" } ] }
The scenario expects two oauth-prompt / form surfaces from the plugin layer:
figma-oauth — oauth.route='connector', connectorId='figma',persists at the project tier so multi-conversation work doesn't re-prompt.
file-pick — kind='form', captures the Figma file URL on thefirst turn so figma-extract knows where to read.
The daemon auto-derives the figma-oauth surface when od.connectors.required[].id='figma' is declared and the connector isn't yet authorised (spec §10.3.1 implicit oauth-prompt rule).
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-02 | fail→pass | 7,478 | 5,690 | -24% | 1 | 1 | 0% | 1,232 | 1,199 | -3% | 0 | 0 | — |
case-01 | fail→fail | 9,898 | 20,928 | +111% | 1 | 1 | 0% | 1,673 | 1,320 | -21% | 0 | 0 | — |
case-03 | fail→pass | 6,761 | 11,749 | +74% | 1 | 1 | 0% | 1,090 | 2,031 | +86% | 0 | 0 | — |
case-04 | fail→pass | 26,027 | 4,626 | -82% | 1 | 1 | 0% | 1,453 | 911 | -37% | 0 | 0 | — |
case-05 | fail→pass | 14,199 | 3,712 | -74% | 1 | 1 | 0% | 2,008 | 1,104 | -45% | 0 | 0 | — |
case-06 | fail→pass | 12,006 | 4,663 | -61% | 1 | 1 | 0% | 1,822 | 1,085 | -40% | 0 | 0 | — |
case-21 | pass→pass | 13,990 | 9,746 | -30% | 1 | 1 | 0% | 2,135 | 2,337 | +9% | 0 | 0 | — |
case-07 | fail→pass | 10,285 | 3,221 | -69% | 1 | 1 | 0% | 1,398 | 907 | -35% | 0 | 0 | — |
case-08 | fail→fail | 8,052 | 8,093 | +1% | 1 | 1 | 0% | 1,190 | 1,110 | -7% | 0 | 0 | — |
case-09 | fail→pass | 18,958 | 3,573 | -81% | 1 | 1 | 0% | 2,209 | 1,028 | -53% | 0 | 0 | — |
case-10 | fail→pass | 16,626 | 4,157 | -75% | 1 | 1 | 0% | 2,544 | 989 | -61% | 0 | 0 | — |
case-11 | pass→pass | 14,219 | 3,842 | -73% | 1 | 1 | 0% | 2,368 | 1,167 | -51% | 0 | 0 | — |
case-12 | fail→pass | 22,074 | 10,723 | -51% | 1 | 1 | 0% | 2,326 | 1,962 | -16% | 0 | 0 | — |
case-13 | fail→pass | 16,918 | 10,166 | -40% | 1 | 1 | 0% | 2,017 | 1,899 | -6% | 0 | 0 | — |
case-14 | fail→pass | 9,978 | 4,488 | -55% | 1 | 1 | 0% | 1,504 | 1,165 | -23% | 0 | 0 | — |
case-15 | fail→fail | 7,454 | 6,401 | -14% | 1 | 1 | 0% | 1,122 | 1,645 | +47% | 0 | 0 | — |
case-16 | fail→pass | 18,812 | 6,195 | -67% | 1 | 1 | 0% | 3,589 | 1,339 | -63% | 0 | 0 | — |
case-17 | pass→pass | 11,123 | 3,494 | -69% | 1 | 1 | 0% | 1,457 | 859 | -41% | 0 | 0 | — |
case-18 | pass→pass | 19,145 | 11,527 | -40% | 1 | 1 | 0% | 2,552 | 2,151 | -16% | 0 | 0 | — |
case-19 | pass→pass | 12,955 | 9,987 | -23% | 1 | 1 | 0% | 1,724 | 2,296 | +33% | 0 | 0 | — |
case-20 | pass→pass | 18,655 | 15,828 | -15% | 1 | 1 | 0% | 2,716 | 2,760 | +2% | 0 | 0 | — |
case-22 | pass→pass | 8,007 | 10,704 | +34% | 1 | 1 | 0% | 1,654 | 2,424 | +47% | 0 | 0 | — |
case-23 | fail→fail | 23,211 | 2,483 | -89% | 1 | 1 | 0% | 1,711 | 769 | -55% | 0 | 0 | — |
case-24 | fail→pass | 7,315 | 3,401 | -54% | 1 | 1 | 0% | 1,225 | 1,061 | -13% | 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. 24 cases were attempted, and 23 counted toward the lift figure. The other 1 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 +54 percentage points is the difference between those two pass rates over the 23 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.