SKILL.md
$27
export const greetUser = actionClient
.inputSchema(z.object({ name: z.string().min(1) }))
.action(async ({ parsedInput: { name } }) => {
return { greeting: Hello, ${name}! };
});
## Chainable API OrdercreateSafeActionClient(opts?)
.use(middleware) // repeatable, adds pre-validation middleware
.metadata(data) // required if defineMetadataSchema is set
.inputSchema(schema, utils?) // Standard Schema or async factory function
.bindArgsSchemas([...]) // schemas for .bind() arguments (order with inputSchema is flexible)
.useValidated(middleware) // repeatable, adds post-validation middleware (requires inputSchema or bindArgsSchemas before it)
.outputSchema(schema) // validates action return value
.action(serverCodeFn, utils?) // creates SafeActionFn
.stateAction(serverCodeFn, utils?) // creates SafeStateActionFn (for useStateAction or React's useActionState)
Each method returns a new client instance — the chain is immutable.
## Entry Points
| Entry point | Environment | Exports |
|---|---|---|
| `next-safe-action` | Server | `createSafeActionClient`, `createMiddleware`, `createValidatedMiddleware`, `returnValidationErrors`, `flattenValidationErrors`, `formatValidationErrors`, `DEFAULT_SERVER_ERROR_MESSAGE`, error classes, all core types |
| `next-safe-action/hooks` | Client | `useAction`, `useOptimisticAction`, `useStateAction`, hook types |
| `next-safe-action/stateful-hooks` | Client | `useStateAction` (re-export from hooks for backward compatibility) |
## Supporting Docs
- [Client setup & configuration](./client-setup.md)
- [Input & output validation with Standard Schema](./input-output-validation.md)
- [Server error handling](./error-handling.md)
## Anti-Patterns
// BAD: Missing "use server" directive — action won't work
import { actionClient } from "@/lib/safe-action";
export const myAction = actionClient.action(async () => {});
// GOOD: Always include "use server" in action files
"use server";
import { actionClient } from "@/lib/safe-action";
export const myAction = actionClient.action(async () => {});
// BAD: Calling .action() without .metadata() when metadataSchema is defined
const client = createSafeActionClient({
defineMetadataSchema: () => z.object({ actionName: z.string() }),
});
client.action(async () => {}); // TypeScript error!
// GOOD: Always provide metadata before .action() when schema is defined
client
.metadata({ actionName: "myAction" })
.action(async () => {});
// BAD: Returning an error instead of throwing
export const myAction = actionClient
.inputSchema(z.object({ email: z.string().email() }))
.action(async ({ parsedInput }) => {
const exists = await db.user.findByEmail(parsedInput.email);
if (exists) {
return { error: "Email taken" }; // Not type-safe, not standardized
}
});
// GOOD: Use returnValidationErrors for field-level errors
import { returnValidationErrors } from "next-safe-action";
export const myAction = actionClient
.inputSchema(z.object({ email: z.string().email() }))
.action(async ({ parsedInput }) => {
const exists = await db.user.findByEmail(parsedInput.email);
if (exists) {
returnValidationErrors(z.object({ email: z.string().email() }), {
email: { _errors: ["Email is already in use"] },
});
}
return { success: true };
});
## Server Code Function Parameters
The function passed to `.action()` receives a single object:
.action(async ({
parsedInput, // validated input (typed from inputSchema)
clientInput, // raw client input (unknown)
bindArgsParsedInputs, // validated bind args tuple
bindArgsClientInputs, // raw bind args
ctx, // context from middleware chain
metadata, // metadata set via .metadata()
}) => {
// return data
});
For `.stateAction()`, a second argument is added:
.stateAction(async ({ parsedInput, ctx }, { prevResult }) => {
// prevResult is the previous SafeActionResult (structuredClone'd)
return { count: (prevResult.data?.count ?? 0) + 1 };
});