laguagu/claude-code-nextjs-skills / skills/ai-sdk-6
ai-sdk-6
VerifiedVercel AI SDK v6 development, for projects already on ai@6. Use when building or maintaining AI agents, chatbots, tool integrations, streaming apps, or structured output in a v6 codebase. New projects and ai@7 code use ai-sdk-7; an unknown version goes through ai-sdk. Covers ToolLoopAgent, useChat, generateText, streamText, tool approval, provider tools, MCP integration, and Output patterns.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 5.1 KB |
| references/agents.md | markdown | 1.0 KB |
| references/core-functions.md | markdown | 1.1 KB |
| references/examples.md | markdown | 787 B |
| references/mcp.md | markdown | 1009 B |
| references/middleware.md | markdown | 1.0 KB |
| references/tools.md | markdown | 1.9 KB |
| references/ui-hooks.md | markdown | 1.9 KB |
| references/workflows.md | markdown | 831 B |
references/tools.md
View on GitHub# AI SDK 6 tool contracts
Define typed input with `inputSchema`; keep authorization in `execute` and
server policy. Tool descriptions and model instructions cannot authorize an
operation. A tool without `execute` can be fulfilled by the client or an
external executor; it is not automatically a server action.
V6 manual approval uses tool-level `needsApproval` and the UI's
`addToolApprovalResponse`. Do not replace it with SDK 7's `toolApproval`.
Bind browser-round-tripped approval to trusted server state for sensitive
operations. `experimental_toolApprovalSecret` (HMAC-signed approvals) is
accepted by `generateText`/`streamText` from 6.0.202 and by `ToolLoopAgent`
settings and `prepareCall` from 6.0.272. Before 6.0.202, approved tool calls
replayed from client history ran without schema re-validation or an approval
re-check, so a forged approval executed: treat 6.0.202 as the security floor.
Approval does not replace tenant/role checks or idempotency.
Render the installed union, including `input-streaming`, `input-available`,
`approval-requested`, `approval-responded`, `output-available`,
`output-denied` and `output-error`. Typed parts use `tool-{name}`; runtime
schemas use dynamic tool parts. Partial input is not completed input.
Continue with `sendAutomaticallyWhen: lastAssistantMessageIsCompleteWithToolCalls`
or `...WithApprovalResponses` when all client tool outputs or approvals are
ready; avoid a custom loop that resubmits indefinitely. Client-executed tools
answer from `onToolCall`: check `toolCall.dynamic` first for type narrowing,
and call `addToolOutput({ tool, toolCallId, output })` (or `state:
'output-error'` with `errorText`) without awaiting it there, which can deadlock.
Provider-executed tools have provider-owned execution and permission semantics.
See the installed tool source and
[v6 tool calling guide](https://github.com/vercel/ai/blob/ai%406.0.297/content/docs/03-ai-sdk-core/15-tools-and-tool-calling.mdx).
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |