Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Implement secure Next.js Server Actions for mutations, forms, and optimistic UI. Use when building actions, form flows, auth checks, or revalidation after writes.
.claude/skills/hoangnguyen0403-nextjs-server-actions/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -26% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 36% | 0% |
| case-02 | ✓→✓ | = Same ✓ | -1% | 0% |
| case-03 | ✓→✓ | = Same ✓ | 11% | 0% |
| case-04 | ✓→✓ | = Same ✓ | -9% | 0% |
> !WARNING] > If project uses pages/ directory instead of App Router, ignore this skill entirely.
Build action files as secure server entrypoints, not as thin wrappers around unsafe form data.
actions.ts with 'use server'.z.object({ ... }).safeParse(...) or formData.get('title') -> validated shape.useFormStatus, useActionState, useTransition, or useOptimistic as needed. Typical form wiring is <form action={createPost}> or <form action={action}>, with useFormStatus() and disabled={pending} inside the submit child.FormData or arguments before processing.actions.ts to avoid closure bugs.redirect() in try/catch: redirect() throws; catching it suppresses redirect.When this skill applies, preserve the following domain terminology or equivalent concrete examples in the answer when relevant:
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 19,960 | 18,766 | -6% | 1 | 1 | 0% | 4,089 | 3,032 | -26% | 0 | 0 | — |
case-02 | pass→pass | 19,734 | 13,599 | -31% | 1 | 1 | 0% | 2,451 | 2,438 | -1% | 0 | 0 | — |
case-03 | pass→pass | 23,482 | 21,851 | -7% | 1 | 1 | 0% | 3,314 | 3,689 | +11% | 0 | 0 | — |
case-04 | pass→pass | 18,648 | 15,716 | -16% | 1 | 1 | 0% | 2,525 | 2,309 | -9% | 0 | 0 | — |
case-05 | pass→pass | 19,027 | 13,516 | -29% | 1 | 1 | 0% | 2,480 | 3,070 | +24% | 0 | 0 | — |
case-06 | pass→pass | 20,820 | 19,557 | -6% | 1 | 1 | 0% | 2,768 | 3,074 | +11% | 0 | 0 | — |
case-07 | pass→pass | 9,206 | 17,126 | +86% | 1 | 1 | 0% | 1,471 | 2,746 | +87% | 0 | 0 | — |
case-08 | pass→pass | 20,323 | 19,652 | -3% | 1 | 1 | 0% | 2,610 | 3,356 | +29% | 0 | 0 | — |
case-09 | fail→pass | 11,999 | 18,587 | +55% | 1 | 1 | 0% | 2,104 | 2,865 | +36% | 0 | 0 | — |
case-10 | pass→pass | 20,916 | 18,174 | -13% | 1 | 1 | 0% | 2,557 | 3,077 | +20% | 0 | 0 | — |
case-11 | pass→pass | 11,181 | 15,409 | +38% | 1 | 1 | 0% | 2,005 | 2,160 | +8% | 0 | 0 | — |
case-12 | pass→pass | 20,253 | 18,672 | -8% | 1 | 1 | 0% | 2,921 | 3,273 | +12% | 0 | 0 | — |
case-13 | pass→pass | 13,801 | 17,680 | +28% | 1 | 1 | 0% | 1,349 | 2,671 | +98% | 0 | 0 | — |
case-14 | pass→pass | 5,525 | 8,046 | +46% | 1 | 1 | 0% | 901 | 1,826 | +103% | 0 | 0 | — |
case-15 | pass→pass | 24,248 | 23,186 | -4% | 1 | 1 | 0% | 3,353 | 3,939 | +17% | 0 | 0 | — |
case-16 | pass→pass | 4,874 | 14,453 | +197% | 1 | 1 | 0% | 753 | 2,099 | +179% | 0 | 0 | — |
case-17 | pass→pass | 10,507 | 17,529 | +67% | 1 | 1 | 0% | 1,776 | 2,502 | +41% | 0 | 0 | — |
case-18 | pass→pass | 12,907 | 17,331 | +34% | 1 | 1 | 0% | 2,169 | 2,671 | +23% | 0 | 0 | — |
case-19 | pass→pass | 19,118 | 14,227 | -26% | 1 | 1 | 0% | 2,577 | 3,052 | +18% | 0 | 0 | — |
case-20 | pass→pass | 19,764 | 20,197 | +2% | 1 | 1 | 0% | 2,495 | 3,125 | +25% | 0 | 0 | — |
case-21 | pass→pass | 22,906 | 12,841 | -44% | 1 | 1 | 0% | 2,978 | 2,824 | -5% | 0 | 0 | — |
case-22 | pass→pass | 8,732 | 7,911 | -9% | 1 | 1 | 0% | 432 | 757 | +75% | 0 | 0 | — |
case-23 | pass→pass | 28,625 | 23,987 | -16% | 1 | 1 | 0% | 3,954 | 3,726 | -6% | 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 +9 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.