Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Keep registry base and radix trees in sync when editing shared UI
.claude/skills/kunanonj-cursor-plugin-shadcn-rule-registry-bases-parity/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | 172% | 0% |
| case-23 | ✓→✗ | ▼ Worse | -81% | 0% |
| case-25 | ✓→✗ | ▼ Worse | -51% | 0% |
| case-16 | ✓→✓ | ＝ Same ✓ | -38% | 0% |
| case-01 | ✗→✗ | ＝ Same ✗ | -46% | 0% |
apps/v4/registry/bases/base and apps/v4/registry/bases/radix are parallel registries. Anything that exists in both trees for the same purpose (preview blocks, mirrored examples, shared card layouts, etc.) must stay in sync.
bases/base/..., apply the same behavioral and visual change to the matching path under bases/radix/... (and the reverse).@/registry/bases/base/ui/* vs @/registry/bases/radix/ui/*, or Base UI vs Radix component props).blocks/preview/** — preview cards and blocksui/* components when both exist for the same componentAfter edits, briefly confirm both trees were updated (or state why one side is intentionally unchanged).
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 5,417 | 4,876 | -10% | 1 | 1 | 0% | 926 | 502 | -46% | 0 | 0 | — |
case-02 | fail→fail | 5,467 | 5,592 | +2% | 1 | 1 | 0% | 256 | 548 | +114% | 0 | 0 | — |
case-03 | fail→fail | 4,057 | 4,595 | +13% | 1 | 1 | 0% | 186 | 512 | +175% | 0 | 0 | — |
case-04 | fail→pass | 3,753 | 7,035 | +87% | 1 | 1 | 0% | 686 | 1,864 | +172% | 0 | 0 | — |
case-05 | fail→fail | 4,580 | 5,395 | +18% | 1 | 1 | 0% | 184 | 502 | +173% | 0 | 0 | — |
case-06 | fail→fail | 2,315 | 4,857 | +110% | 1 | 1 | 0% | 473 | 540 | +14% | 0 | 0 | — |
case-07 | fail→fail | 19,023 | 5,347 | -72% | 1 | 1 | 0% | 3,662 | 511 | -86% | 0 | 0 | — |
case-08 | fail→fail | 3,958 | 5,243 | +32% | 1 | 1 | 0% | 236 | 547 | +132% | 0 | 0 | — |
case-09 | fail→fail | 4,046 | 5,529 | +37% | 1 | 1 | 0% | 219 | 515 | +135% | 0 | 0 | — |
case-10 | fail→fail | 10,633 | 7,861 | -26% | 1 | 1 | 0% | 2,215 | 497 | -78% | 0 | 0 | — |
case-11 | fail→fail | 3,781 | 31,300 | +728% | 1 | 1 | 0% | 249 | 732 | +194% | 0 | 0 | — |
case-12 | fail→fail | 5,738 | 5,288 | -8% | 1 | 1 | 0% | 303 | 538 | +78% | 0 | 0 | — |
case-13 | fail→fail | 4,567 | 5,394 | +18% | 1 | 1 | 0% | 156 | 553 | +254% | 0 | 0 | — |
case-14 | fail→fail | 5,989 | 6,740 | +13% | 1 | 1 | 0% | 236 | 530 | +125% | 0 | 0 | — |
case-15 | fail→fail | 10,551 | 3,218 | -70% | 1 | 1 | 0% | 2,535 | 638 | -75% | 0 | 0 | — |
case-16 | pass→pass | 6,096 | 2,561 | -58% | 1 | 1 | 0% | 1,231 | 768 | -38% | 0 | 0 | — |
case-17 | fail→fail | 3,619 | 7,993 | +121% | 1 | 1 | 0% | 182 | 679 | +273% | 0 | 0 | — |
case-18 | fail→fail | 4,562 | 5,460 | +20% | 1 | 1 | 0% | 218 | 574 | +163% | 0 | 0 | — |
case-24 | fail→fail | 3,018 | 4,507 | +49% | 1 | 1 | 0% | 330 | 561 | +70% | 0 | 0 | — |
case-19 | fail→fail | 4,988 | 6,156 | +23% | 1 | 1 | 0% | 834 | 545 | -35% | 0 | 0 | — |
case-20 | fail→fail | 15,534 | 5,188 | -67% | 1 | 1 | 0% | 3,336 | 494 | -85% | 0 | 0 | — |
case-21 | fail→fail | 12,054 | 6,008 | -50% | 1 | 1 | 0% | 2,431 | 529 | -78% | 0 | 0 | — |
case-22 | fail→fail | 4,102 | 5,519 | +35% | 1 | 1 | 0% | 217 | 603 | +178% | 0 | 0 | — |
case-23 | pass→fail | 12,087 | 4,614 | -62% | 1 | 1 | 0% | 2,840 | 540 | -81% | 0 | 0 | — |
case-25 | pass→fail | 8,641 | 5,619 | -35% | 1 | 1 | 0% | 1,582 | 775 | -51% | 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. 25 cases were attempted, and 4 counted toward the lift figure. The other 21 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 -33 percentage points is the difference between those two pass rates over the 4 comparable cases. 10 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.