vercel-labs/vercel-plugin
60 skills418k installs11.2k starsGitHub
Install
npx skills add https://github.com/vercel-labs/vercel-pluginSkills in this repo
1Agent Browseragent-browser is a Vercel plugin skill that exposes a browser automation command line interface so coding agents can behave like a human tester on the web. Solo builders shipping Next.js, Vite, Nuxt, or other local dev stacks use it when curl is not enough—you need to see layout, click through flows, or capture proof that localhost actually rendered. The skill’s metadata ties into common repo layouts such as e2e folders, Playwright and Cypress configs, and dev-server bash patterns, so invocation often follows naturally when a server starts or when the user asks to automate the browser. It supports the full loop from navigation and form fills to screenshots and structured extraction, with optional chaining into agent-browser-verify for a explicit visual gut-check message when a localhost or 127.0.0.1 URL appears. It is not a replacement for a full CI matrix; it is fast agent-driven smoke and exploration on demand. Expect shell and browser permissions and respect sites’ terms when scraping production data.404kinstalls2NextjsThe nextjs skill packages Vercel-aligned Next.js App Router expertise for solo builders using Claude Code, Cursor, or Codex on real repos. It activates when you are scaffolding routes, debugging hydration or caching surprises, wiring Server Actions, or planning how data should flow between server and client components. Metadata ties the skill to concrete file globs—app/**, next.config.*, tsconfig—and shell habits like next build and create-next-app, so agents surface it in the right workspace without you re-pasting framework rules. Built-in validate hooks catch legacy Pages Router exports early, which saves painful compile cycles. Whether you are launching a marketing site, a SaaS dashboard, or an API-backed product, the skill keeps architecture decisions consistent with current Next.js docs and typical Vercel deploy paths, reducing accidental anti-patterns before they reach ship.4.5kinstalls3React Best PracticesReact Best Practices is Vercel Engineering’s agent-oriented performance playbook for React and Next.js codebases (v1.0.0, January 2026). Solo builders and small teams install it when they want Claude, Cursor, or Codex to refactor or generate UI code against the same priorities humans would use: kill request waterfalls first, shrink bundles second, then apply incremental advanced patterns. The document spans 40+ rules in eight categories, each with explanations, before/after examples, and impact notes so automation does not trade clarity for speed. Use it during active frontend work in Build and again before release when you are hardening Ship performance. It does not replace product design or backend architecture, but it sharply reduces inconsistent agent output on Next.js apps, marketing sites, and SaaS dashboards where Core Web Vitals and TTFB matter for conversion.819installs4ShadcnThe shadcn skill gives solo builders expert agent guidance for shadcn/ui across CLI setup, installing primitives, composing production interfaces, custom registries, and Tailwind theming. Use it when you run npx shadcn init or add, edit components.json, work under components/ui or monorepo app paths, or troubleshoot theme and component failures. It encodes Vercel plugin metadata with priority, official ui.shadcn.com doc links, and validate rules—such as warning when base-ui is chosen while AI Elements expect Radix—so you avoid subtle type and API mismatches. The skill fits SaaS and web product builds where you want copy-paste-owned components instead of opaque UI kits. Intermediate complexity reflects monorepos, registry customization, and migration commands. It does not replace visual brand strategy in validate or launch distribution work; it accelerates consistent, accessible UI implementation so you can ship faster and review in ship with fewer ad-hoc Tailwind experiments.439installs5Deployments Cicddeployments-cicd is expert agent guidance for Vercel’s deployment lifecycle and the CI files that drive it. Solo builders hit this when wiring GitHub Actions, GitLab, or Bitbucket pipelines to preview and production URLs, promoting a tested preview, or rolling back a bad production deploy. The skill encodes bash patterns for vercel CLI commands and filesystem patterns for vercel.json and monorepo apps/*/vercel.json layouts. A recommended validate hook steers you away from hand-rolled node-cron schedules toward Vercel Cron Jobs so scheduled tasks run on the platform. Use it during Ship when hardening release automation and during Operate when inspecting deployments or recovering from incidents. It assumes you want Vercel-native workflows rather than generic container-only deploy scripts.356installs6TurbopackTurbopack is an agent skill for solo and indie builders shipping Next.js apps who want the default Vercel bundler configured correctly instead of fighting slow dev servers or stale Webpack setups. It activates when you enable Turbopack, optimize hot module replacement, debug bundler build failures, or compare Turbopack with Webpack behavior. The skill encodes expert knowledge from turbo.build and Next.js architecture docs: how HMR should feel instant at scale, how Next.js 16 expects top-level turbopack configuration, and when to chain into the broader nextjs skill for webpack(config) migrations or experimental.turbopack renames. For a one-person team, that means fewer opaque compile errors and less time lost re-reading release notes. Use it during local build iteration before you treat production deploy issues as infrastructure problems.332installs7Ai Sdkai-sdk is expert agent guidance for the Vercel AI SDK—the TypeScript layer solo builders use to ship AI features on Vercel and elsewhere without relearning every provider API. It activates when you touch chat or completion API routes, shared lib/ai helpers, or dependencies on ai and @ai-sdk packages, and it points agents at official docs, reference, and sitemap for up-to-date patterns. Use it while prototyping a validated idea with a thin chat MVP, while building production streaming endpoints with tool calling, or when extending an agent with MCP-aware flows, structured JSON, embeddings, or reranking. The skill emphasizes correct project placement (App Router vs Pages, monorepo apps/* paths) so Cursor and Claude Code suggestions stay consistent with how Vercel documents the stack. It complements infrastructure skills for deployment but stays focused on application-level SDK usage rather than platform dashboards alone.297installs8Vercel Agentvercel-agent teaches solo builders how Vercel’s built-in Agent suite fits into everyday ship and operate work: AI-assisted pull request review, faster incident investigation with codebase and deployment context, and help installing related SDKs. It is aimed at indie teams already on Vercel who want fewer manual review passes and clearer debugging when previews or production misbehave. The skill recognizes CI files and bash usage of vercel agent, and it explicitly hands off to deployments-cicd when GitHub Actions wire VERCEL_TOKEN or vercel-action—so you treat review automation and preview deploys as one story. Use it when standing up review bots, understanding what Agent can see in your repo, or pairing operational anomalies with suggested fixes—not as a replacement for your own security review. It sits beside ai-sdk for application AI features while focusing on platform-native agent tooling.295installs9Chat SdkChat-sdk is a Vercel Chat SDK expert skill for solo builders who do not want a separate microservice per chat platform. It activates when you build multi-platform bots—Slack, Telegram, Microsoft Teams, Discord, Google Chat, GitHub, or Linear—from one application, typically under Next.js API routes such as `app/api/slack/**` or shared `lib/bot/**` modules. The skill walks through the Chat class, adapters, threading, rich cards and modals, streaming responses, and webhook verification so your agent does not reinvent incompatible handlers per vendor. Metadata points at AI SDK UI chatbot documentation and the Vercel ai-chatbot reference implementation, which helps you align UI streaming with inbound platform events. Use it when you are wiring bot endpoints, consolidating webhook folders, or extending an existing Vercel app with a second chat provider without forking business logic.292installs10Vercel CliVercel-cli is the Vercel platform terminal skill for solo builders who deploy and operate on Vercel without memorizing every subcommand. It applies when users ask to deploy a project, configure environment variables, connect domains, start local development, add a database, or install Marketplace integrations. The documented workflow loads the skill, maps the request through a decision tree, reads the matching `references/<topic>.md` file, and executes commands with agent-friendly defaults such as `--yes` in CI, authentication via `VERCEL_TOKEN`, and `vercel pull` when env sync matters. It also steers you toward `vercel curl` for protected previews rather than weakening security. For indies, that turns repetitive infra chat into repeatable, citeable steps from link → env → deploy → domain → integration, whether you ship a first MVP or adjust production settings later.292installs11Routing MiddlewareRouting Middleware is a Vercel plugin skill that steers solo builders and their agents through request interception at deploy time—before responses hit cache or static assets. You reach for it when adding auth gates, locale-based rewrites, A/B personalization, or canonical redirects that must live in middleware.ts, proxy.ts, or vercel.json rather than scattered in every page. The skill maps concrete file globs and bash patterns so Claude Code or Cursor activates on the right touchpoints in monorepos and multi-app layouts. It also encodes validation hints, including the Next.js 16 rename from middleware.ts to proxy.ts and when to hand off to the nextjs skill for migration-safe patterns. For a one-person team shipping on Vercel, this reduces foot-guns around NextResponse imports, runtime mismatches, and platform config that are expensive to debug only after production traffic hits the wrong region or auth cookie.286installs12Vercel Functionsvercel-functions is a Vercel-focused agent skill for solo builders shipping APIs and server logic on Next.js or compatible frameworks. It activates when you touch api routes, app route handlers, or deployment config and need expert defaults for modern Web Request/Response handlers instead of legacy Pages Router patterns. The skill encodes official docs priorities around runtimes, edge versus serverless tradeoffs, streaming, scheduled jobs, and performance tuning so your agent does not guess from stale training data. Built-in validate patterns turn common footguns—default exports, NextApiRequest types, raw OpenAI or Anthropic clients in routes—into actionable fixes, including a recommended handoff to ai-sdk for provider abstraction. For indie teams on Vercel, it compresses hours of docs spelunking into checklist-driven edits aligned with vercel.json and monorepo layouts, while still supporting ship-phase debugging through logs and dev workflows implied by metadata bash patterns.281installs13Ai Gatewayai-gateway helps solo builders centralize LLM access on Vercel so one integration path covers OpenAI, Anthropic, Google, and other providers with routing, failover, and spend visibility. The skill is meant for moments when you are adding @ai-sdk/gateway, choosing model strings, or replacing brittle per-vendor keys in serverless environments. Its validate layer catches hyphenated version mistakes, missing provider prefixes, and outdated model slugs that agents often hallucinate from old training cutoffs. Guidance pushes toward OIDC token flows through the Vercel CLI rather than manually rotated gateway keys when your deployment supports it. For indie SaaS and agent products, this reduces vendor lock-in at the HTTP boundary while keeping observability aligned with Vercel’s AI Gateway docs. Pair it with env-vars when .env still holds raw OPENAI_API_KEY or ANTHROPIC_API_KEY lines, since the plugin chains those detections into gateway-oriented auth advice.279installs14Env Varsenv-vars gives solo builders a precise playbook for Vercel environment variables—from which .env file loads in dev versus CI to how vercel env commands sync preview and production without leaking secrets into git. The skill triggers on .env paths and common bash invocations so your agent does not treat all dotenv files as interchangeable. It explains OIDC-oriented flows where Vercel can supply short-lived credentials, which pairs naturally with AI workloads that otherwise tempt developers to paste long-lived provider keys. When validation surfaces raw OPENAI_API_KEY, ANTHROPIC_API_KEY, or GOOGLE_API_KEY entries, the plugin chains into ai-gateway guidance to steer you toward gateway auth patterns. For indie teams, this is the difference between a green local build and a red production deploy caused by missing scoped variables or example files that never document required keys. Use it during launch week and ongoing operate iterations whenever you rotate secrets or onboard a new collaborator who needs a safe .env.example template.270installs15Vercel StorageVercel Storage is an agent skill from the Vercel plugin that helps solo and indie builders choose, configure, and implement data storage on Vercel deployments. It is aimed at Next.js and full-stack apps where you need files, low-latency config, Postgres, or Redis without leaving the Vercel ecosystem. Invoke it when you are adding lib/blob or edge-config modules, installing marketplace database packages, or debating Blob versus Edge Config versus Neon for a new feature. The skill surfaces priority-ranked official documentation and concrete path and bash install triggers so your coding agent stays aligned with Vercel’s supported patterns. That matters because the wrong storage choice creates rework on auth, caching, and edge runtime limits. Use it during backend and integration work; pair with auth and workflow skills when persistence touches user data or long-running jobs.266installs16WorkflowWorkflow is a Vercel plugin agent skill focused on the Workflow DevKit for builders who need reliability beyond a single serverless invocation. Solo developers use it when an API route, background job, or agent must pause for human approval, wait on webhooks, retry failed steps, or resume after deploys without losing state. The skill anchors on official Vercel workflow documentation and useworkflow.dev, with high-priority metadata so agents recognize workflow directories and package installs early. It fits multi-phase work: you implement steps and routes in Build, then rely on the same mental model when debugging stuck runs or scaling task volume in Operate. Compared to ad-hoc setTimeout chains or naive queue scripts, WDK-oriented guidance pushes toward explicit steps and durable semantics. Invoke when prompts mention vercel workflow, workflow devkit, or durable orchestration on Vercel.254installs17AuthAuth is a Vercel plugin skill that steers solo builders through user authentication on Next.js apps deployed to Vercel. It consolidates guidance for Clerk as a native Marketplace option alongside Descope and Auth0, covering middleware protection, dedicated sign-in and sign-up routes, and configuration files such as clerk.config and auth.ts. Use it when you are implementing login for a SaaS MVP, gating dashboards, or provisioning auth through Vercel’s marketplace rather than rolling custom sessions. The skill links Auth.js and official Next.js authentication documentation so agents do not invent deprecated Pages Router patterns. It is intentionally integration-focused: you get SDK install bash patterns and path triggers for common repo layouts. After auth is wired, storage and workflow skills often follow for per-user data and async jobs. Skip it if you already have a frozen auth stack documented and only need penetration testing.251installs18Runtime Cacheruntime-cache is a Vercel-focused agent skill for solo builders shipping serverless apps who need an explicit Runtime Cache layer on top of framework defaults. It explains how Vercel’s ephemeral per-region key-value store works, how to structure cache keys and tags, and how invalidation propagates when data changes across API routes, middleware, and build-time consumers. Use it when you are implementing response caching, cross-function shared state, or tag-driven purge flows rather than leaning only on Next.js cache components or CDN headers. The skill aligns with paths like lib/cache and installing @vercel/functions, and it nudges you away from raw Redis TCP clients that break in serverless unless you adopt HTTP-based Upstash patterns. It pairs naturally with next-cache-components for App Router caching and vercel-storage when you need durable or Redis-shaped storage instead of short-lived runtime cache entries.250installs19Verificationverification is a Vercel plugin skill for solo builders who start a dev server and need more than a green compile— they want proof that the feature actually works across UI, network, persistence, and configuration. It listens for dev-server startup and natural-language frustration or explicit verify phrases, then steers the agent through a full-story pass: what the user is trying to do, what should happen in the browser, which API or server action fires, what data should change, and whether responses match expectations including environment variables. It fits indie workflows where one person toggles between Next, Vite, or Astro locally and Vercel dev in preview parity. It is not a replacement for formal CI suites; it is a guided sanity check when integration bugs hide between layers. Use it when local behavior is ambiguous, regressions span client and server, or you want confidence before shipping or demoing to users.246installs20Next Cache Componentsnext-cache-components is an agent skill for solo builders on Next.js 16 who need current guidance on Cache Components, partial prerendering, and the use cache directive family. It explains how to mark server components and data fetches for caching, how cacheLife profiles control freshness, and how cacheTag and updateTag let you invalidate precisely after mutations instead of blunt full-route revalidation. The skill is aimed at App Router code under app/ or monorepo apps/* layouts and surfaces when you run next dev or next build or when prompts mention PPR, stale behavior, or migrating from unstable_cache. Validation rules flag deprecated unstable_cache calls and older cacheHandler configuration, steering refactors toward directive-based caching. It complements Vercel deployment concerns but stays framework-native rather than describing Vercel Runtime Cache KV. Use it when SEO-sensitive pages need static shells with dynamic holes, when API-backed RSC segments need tagged invalidation, or when upgrading a Next 14/15 codebase that still relies on unstable_cache helpers.238installs21Vercel SandboxVercel Sandbox is an agent skill that teaches solo builders how to run untrusted or AI-generated code inside ephemeral Firecracker microVMs on Vercel rather than on their own machine. It is aimed at indie hackers shipping coding agents, internal tools, or student-style playgrounds where arbitrary snippets must not touch the host filesystem or secrets. Use it when prompts mention @vercel/sandbox, isolated execution, code runners, or safely executing model output before you merge results into your repo. The skill prioritizes integration guidance—install lines, import patterns, and trigger phrases—so Claude Code, Cursor, or Codex can suggest the right package and mental model. It matters because one bad `eval` in production costs more than a few cents of sandbox runtime: isolation is the minimum bar for agent products. Pair it with your agent orchestration and deployment skills after you validate outputs.237installs22Next UpgradeNext Upgrade is an agent skill for solo builders maintaining Next.js apps who need a disciplined path from an older major to the latest release. It encodes when to run official codemods, how to bump next in package managers, and where to read breaking-change notes—so you do not rely on vague “update next” chat advice that leaves half-migrated App Router code. Use it during upgrades, between majors, or when grep shows legacy patterns that codemods should rewrite. The skill wires retrieval aliases and bash patterns so Cursor or Claude Code notice next.config and lockfile edits in the same session as migration work. When legacy Pages Router symbols remain, it points you at the companion nextjs skill for deeper routing and rendering guidance. The outcome is a repo that matches current Next defaults with fewer manual fix loops before you ship or launch.232installs23Bootstrapbootstrap is a workflow orchestrator skill from Vercel Labs for solo builders cloning a Next.js or full-stack repo that depends on Vercel-linked storage, auth, and managed integrations. It watches common signals—.env.example files, README setup sections, package.json db scripts, Drizzle and Prisma configs, and imports like Neon serverless, Upstash, Blob, Edge Config, and better-auth—then sequences vercel link, vercel integration add or install, vercel env pull, and only then database push, seed, migrate, or generate plus dev server startup. The goal is to prevent the classic failure mode of copying env files or running migrations before the project is linked and secrets are pulled. When sunset storage packages appear, it routes you to vercel-storage guidance instead of dead-end docs. Use when setting up or repairing a repository, not for day-to-day feature coding. Intermediate complexity: you need the Vercel CLI, a linked team/project, and comfort with env-local workflows.229installs24MarketplaceMarketplace is an agent skill for solo builders shipping on Vercel who need trustworthy steps for third-party add-ons instead of guessing install flows. It covers consuming marketplace services, authoring custom integrations, and managing provisioned resources with the Vercel integration CLI and official docs anchors. The skill emphasizes integration.json structure, retrieval aliases (marketplace, add-ons, third-party services), and when database or Clerk signals appear it points you toward the sibling Storage and Auth skills in the same plugin. Use it while you are wiring Neon Postgres, Clerk middleware, or other marketplace partners into a Next.js or full-stack app before you hard-code secrets or duplicate billing setup. Intermediate complexity assumes you already deploy on Vercel and can run CLI commands from the project root.226installs25Vercel Plugin EvalVercel-plugin-eval is a maintainer-facing agent skill for proving the Vercel Claude plugin behaves correctly under real sessions. It walks you through timestamped test directories under ~/dev/vercel-plugin-testing/, installing the plugin with npx add-plugin, and spawning WezTerm sessions that use interactive zsh and the claude x alias so PretoolUse hooks actually run. The skill forbids shortcuts that invalidate results: no --print mode, no dangerously-skip-permissions, no improvised eval scripts—only documented Bash steps in the conversation. You monitor debug logs, confirm deduplicated skill injection, and produce a coverage report across hook scenarios. Advanced complexity; aimed at plugin authors and power users validating regressions before wider release.224installs26Plugin AuditPlugin-audit is an agent skill for investigating vercel-plugin skill injection on real repositories after Claude Code has run. You locate conversation JSONL logs, parse every tool_use name and input, and replay matching through the same loadSkills and matchSkills functions the PretoolUse hook uses—fast and deterministic compared to shelling the hook per event. The workflow highlights coverage gaps (skills that should have fired but did not), pattern holes in bash and path rules, and stale plugin cache states that cause missed guidance on marketplace or deploy tasks. Multi-phase usefulness spans Ship review when debugging odd sessions and Build agent-tooling when tuning retrieval patterns. Intermediate-to-advanced; best when you already have logs from a named project path.214installs27Benchmark AgentsBenchmark-agents is an agent skill for operators and plugin authors who need to prove that skills and hooks actually run during real tool-calling sessions on Vercel’s stack. It walks you through creating eval directories, installing the plugin, spawning independent Claude Code panes, reading debug logs and claim dirs, inspecting generated source, and mining conversation logs for user corrections. The methodology insists on interactive evaluation because plugin hooks do not fire in print-only or scripted Claude invocations, so the skill is as much a discipline guide as a scenario catalog. Solo builders shipping agent products use it when a multi-system build (workflows, gateways, MCP, queues) must be validated before customers touch it. After failures, the documented path is update skills or hooks, run release, and re-spawn evals—closing the loop between platform features and agent behavior.212installs28ReleaseRelease is the vercel-plugin maintainer skill that turns a ready branch into a shippable plugin version without skipping checks. It starts with three parallel gates—TypeScript noEmit on hooks, the full Bun test suite, and validate for skill frontmatter plus manifest integrity—and halts immediately if anything fails. You then choose patch, minor, or major bump semantics with a clear table: fixes and docs are patch, new skills or hooks are minor, breaking hook API or skill map changes are major, defaulting to patch when the user only says release. Only .plugin/plugin.json holds the version; build refreshes compiled hooks and the generated skill manifest before a single commit and push. Solo builders shipping agent plugins use it when asked to release, ship, bump and push, or cut a release, and it pairs naturally after benchmark-agents or local fixes confirm behavior.211installs29Benchmark TestingBenchmark-testing is a workflow skill for maintainers and power users of vercel-plugin who need reproducible evidence that skill injection works across scenarios. It walks you through creating a base directory of numbered slug folders, installing the plugin into each with npx add-plugin, and launching Claude Code inside WezTerm with absolute cwd, nested-session guards, and project settings loaded explicitly. The emphasis on spawn versus split-pane, pacing between launches, and consistent prompt quoting prevents flaky benchmark runs that would mislead you about plugin quality. Solo builders validating agent behavior before recommending a plugin to a team, or plugin authors regression-testing injection rules, get an operational playbook rather than manual one-off tries. Expect shell, network for add-plugin, and terminal multiplexing literacy.209installs30Knowledge UpdateKnowledge-update is a meta skill bundled with vercel-plugin that fires at session start to correct model hallucinations about the Vercel platform as of the dated knowledge block in SKILL.md. It tells the agent to trust the document over training data for topics like Fluid Compute replacing recommended edge-only patterns, Python runtimes on Fluid Compute, and database offerings through the Marketplace rather than first-party Postgres or KV. Path patterns such as vercel.ts and phrase-based prompt signals help injection when you are editing deploy configuration or asking architecture questions. For solo builders shipping full-stack agents or APIs on Vercel, this reduces rework from outdated “edge-only middleware” or “Vercel is static-only” advice across Build, Ship, Launch, and Operate. Treat it as living release notes the plugin maintainers curate, not a substitute for reading current vercel.com/docs for your exact project.209installs31Benchmark E2eBenchmark E2E is a Vercel-plugin maintainer skill that wraps a Bun orchestrator around realistic sample projects so you can validate skill injection, dev-server startup, and conversation-log analysis in one shot. Indie builders and agent-tooling authors use it when they are evolving a Claude plugin and need evidence—not anecdotes—that injected skills still produce working apps. The default full run covers nine projects over roughly two to three hours; quick mode trims to three projects in under an hour. Each stage is contractual: create dirs and run traced Claude sessions, launch and poll dev servers, match JSONL sessions via run-manifest.json, then emit improvement recommendations. Advanced complexity: you need Bun, the vercel-plugin repo layout, and tolerance for long-running local benchmarks. It is meta tooling for eval loops, not a substitute for unit tests in your own SaaS product unless you fork the pattern.203installs32Benchmark SandboxBenchmark-sandbox (as exposed in the vercel-plugin tree) supports disciplined evaluation of agents inside isolated sandboxes: a structured logger writes either human-readable or NDJSON events to stderr while preserving stdout for a machine-readable RunSummary. Solo builders and skill authors use that split when comparing prompts, skills, or plugin versions without leaking credentials into CI logs. Secret redaction scans for Anthropic key prefixes and generic long tokens tied to API_KEY-style variables, which matters when benchmarks replay real env configs. The skill sits in the agent-eval toolchain—pair it when you need reproducible runs and parseable telemetry instead of one-off curl tests. Prism lists it for builders who treat skills like products and want evidence before shipping. If you only need unit tests for pure functions, a normal test runner suffices; choose this when sandbox runners and event-shaped benchmarks are the deliverable.196installs33Next Forgenext-forge is Vercel’s opinionated SaaS starter packaged as agent expert guidance for solo builders who want a full Turborepo layout instead of hand-rolling auth, billing, and observability. The skill activates when your tree matches pnpm workspaces, multiple apps under apps/, shared packages like database, auth, payments, analytics, and observability, or when you run next-forge init. Agents learn where environment keys live, how Biome and TypeScript configs chain, and how to change product behavior without breaking cross-package imports. It spans scaffolding a new product (Validate) through ongoing feature work (Build) and pre-release hardening (Ship) because the same monorepo conventions apply at each step. Complexity is advanced: many third-party services and env contracts. Use it whenever you are inside a next-forge repo or deliberately choosing that template for a solo SaaS launch on Vercel-aligned tooling.173installs34Vercel FirewallVercel-firewall is an expert agent skill for solo builders and small teams who host APIs and web apps on Vercel and need platform-level protection without becoming a full-time security engineer. It consolidates guidance on the Vercel Firewall product: automatic DDoS mitigation, the Vercel WAF with custom and managed rules, IP allowlists and blocks, rate limiting, bot protection, system bypass for trusted automation, and Attack Mode when you are under active abuse. The skill is written for moments when you are tightening rules before a launch spike, or when production metrics show scrapers, credential stuffing, or volumetric traffic you cannot absorb at origin. It expects you to reason in Vercel’s control plane and CLI (`vercel firewall`), not to replace application-auth design. Use it alongside normal ship-time security reviews; during Operate it becomes the playbook for iterating rules without taking the site offline. Prism lists it under Security so builders comparing edge WAF skills can find Vercel-native flows instead of generic OWASP checklists that ignore Vercel’s managed edge.103installs35Agent Browser Verifyagent-browser-verify is a Vercel plugin skill that hooks automated browser verification to local dev server lifecycles. When you start Next, Vite, Nuxt, or package-manager dev scripts, the agent can launch agent-browser to confirm the app actually renders instead of a blank or error screen. It checks console errors, validates important UI elements, and returns a clear pass or fail so you do not stack features on a broken baseline. Solo builders benefit because agents often claim success without seeing the browser; this skill encodes a lightweight visual regression gate aligned with Anthropic sub-agent documentation linked in metadata. It complements manual testing during Build and formal checks in Ship. It is not a replacement for unit tests or production monitoring. Install when your workflow already uses Claude Code-style dev patterns and you want automatic screenshots and error scans on dev startup or when the user reports loading failures.69installs36TurborepoTurborepo is Vercel’s build orchestrator for JS/TS monorepos: it models tasks in turbo.json, runs them in parallel, and reuses work through hashing and caching. This skill gives agents opinionated guidance when you are standing up a monorepo, tuning cache keys and outputs, turning on remote caching, or shrinking CI to affected packages only. It aligns with current docs, flags the v2 rename of pipeline to tasks, and can nudge related fixes when legacy Vercel storage imports show up. Solo builders shipping multi-package SaaS, CLIs, or internal tools benefit most when build times or flaky CI become the bottleneck—not when they run a single-package app with no workspace layout.64installs37Ai Elementsai-elements is a Vercel plugin skill that steers your coding agent through the AI Elements library—pre-built React components for AI-native interfaces layered on shadcn/ui and the Vercel AI SDK. Solo builders shipping a SaaS copilot or in-app agent use it when prompts mention chat UIs, message lists, tool invocation UI, token streaming, or reasoning panels instead of rebuilding primitives from scratch. The skill encodes install commands across npm, pnpm, bun, and yarn, bash patterns for npx ai-elements and shadcn registry adds, and glob rules so agents notice components/ai-elements and chat-related folders. It points to official SDK documentation for tool calling and sitemap-backed discovery, which helps keep implementations aligned with current APIs. Expect integration guidance rather than a single generated artifact: you still own routing, auth, and backend tools, but front-of-house AI UX becomes faster and more consistent.62installs38SwrSWR is a Vercel Labs agent skill that steers solo builders through client-side data fetching in React using the SWR library. Invoke it when you add useSWR hooks, shared fetchers under lib/ or hooks/, or when you need caching, background revalidation, mutations, optimistic UI, pagination, or infinite loading without reinventing patterns. The skill encodes detection signals—file globs, import paths, package install commands, and natural-language phrases like stale-while-revalidate—so coding agents attach the right guidance early. It prioritizes keeping UI synchronized with server state after creates and updates, and points to official SWR documentation for edge cases. You still need a working API and error boundaries; this skill focuses on hook composition and cache semantics, not backend design. Ideal for Next.js or Vite React dashboards where snappy lists and detail views matter.58installs39Vercel QueuesVercel Queues Skill guides solo builders through Vercel’s durable event streaming queue in public beta, including topics, consumer groups, automatic retries, and delayed delivery. It triggers when repositories contain lib/queues layouts, workflow TypeScript modules, or API routes under app/api/workflow, or when dependencies pull in @vercel/queue or @vercel/workflow. Builders adopt it for async processing that should not block HTTP responses, fan-out patterns that multicast work to multiple consumers, and event-driven architectures aligned with Workflow DevKit. The skill references operational cost at roughly sixty cents per million operations so indie projects can reason about spend before committing. It complements conformance and workflow collaboration docs linked in metadata. Use during backend build when webhooks, background enrichment, email sends, or agent orchestration steps need reliable at-least-once handling on Vercel without self-managing Redis or SQS.56installs40Json Renderjson-render is a Vercel plugin agent skill that tells your coding agent how to render AI chat responses in React and Next.js apps that use the AI SDK message model. Solo builders reach for it when a custom chat UI shows broken partial streams, wrong tool-invocation layouts, or legacy message.content shapes that no longer match UIMessage v6. The skill scopes work to chat and message component folders and emphasizes parts-based rendering, tool result presentation, and streaming-friendly UI states. It is not a generic markdown tutorial; it is procedural knowledge for agent-facing interfaces where safety and live updates matter. When the agent detects migration debt or hand-rolled markdown for model output, chainTo rules load ai-sdk or ai-elements so you fix the root SDK and component patterns instead of patching symptoms. Use it during Build whenever the conversation UI is on the critical path for your agent product or embedded copilot.55installs41V0 Devv0-dev is Vercel Labs guidance for v0—the AI UI generator tied to GitHub and Vercel deployment workflows. Solo builders install it when they want structured help choosing v0 CLI versus SDK flows, generating components from prompts, and folding v0 output into a real Next.js or React repo without breaking streaming or design system conventions. The skill documents docs URLs, import and bash patterns for discovery, and automatic chain suggestions to shadcn for registry theming and to AI SDK v6 when chat or streaming APIs show up in generated files. It fits builders who treat v0 as the first pass on marketing pages or dashboards, then harden with your stack. Skip it if you are not using v0 or Vercel-aligned React UI at all.55installs42Cron JobsThe cron-jobs skill is a Vercel-focused reference for solo builders adding, editing, or debugging scheduled tasks declared in vercel.json. It keeps you on the platform-native path: cron entries invoke serverless function routes on a schedule instead of bolting node-cron, cron-parser, croner, or node-schedule into a long-lived process that will not survive serverless timeouts. Metadata points at Vercel docs and sitemap retrieval, and path patterns help agents find the right vercel.json in monorepos. When your handler smells like polling with setInterval or setTimeout loops, the skill steers you toward the Workflow DevKit skill for durable execution. When imports show classic npm cron packages, it hands off to vercel-functions so the HTTP entrypoint is correct. Use it whenever you are scheduling backups, digest emails, cache warming, or quota resets on Vercel without reinventing infrastructure cron.54installs43ObservabilityObservability is a Vercel-plugin agent skill that turns the platform’s observability stack into actionable guidance for solo builders shipping Next.js and full-stack apps on Vercel. It focuses on Drains (shipping logs, traces, speed insights, and web analytics to external sinks), first-party Web Analytics and Speed Insights packages, runtime log inspection, custom events, OpenTelemetry wiring, and dashboard-oriented monitoring workflows. Use it when you are adding or fixing instrumentation in instrumentation.ts, app layouts, or _app entry files, when users report slow pages or broken flows you cannot reproduce locally, or when you need a consistent export path for analytics and traces without bolting on ad-hoc logging. The skill maps common repo layouts—including monorepo apps/* paths—and CLI habits like vercel logs and @vercel/analytics installs so your agent proposes changes that match Vercel’s documented observability model rather than generic APM advice.54installs44Vercel Apivercel-api is a Vercel-plugin agent skill that teaches coding agents how to reach real platform state through the Vercel MCP server and REST API instead of hallucinating project IDs or deployment URLs. It covers authenticating and structuring calls for projects, deployments, environment variables, domains, logs, and documentation discovery, with bash and path patterns that fire when you add @vercel/sdk or wire MCP via .mcp.json and mcp.vercel.com. Solo builders use it to let Claude Code, Cursor, or Codex promote previews, patch env vars, or pull deployment events during a fix session. The skill intentionally chains forward: when your session touches vercel.deployments.create or project updates, it nudges the deployments-cicd skill for promotion and preview URL workflows; when MCP setup dominates, it points to ai-sdk for MCP client integration. Treat it as the live-control layer on top of Vercel hosting—skill package guidance, not a hosted MCP server by itself.53installs45Investigation ModeInvestigation Mode is a Vercel-oriented agent skill that acts as an orchestrated debugging coordinator when solo builders hit dead air: deploys that never finish, pages that spin or stay white, or requests that time out with no obvious stack trace. It is built for Claude Code and similar agents attached to Next.js or Vercel-hosted projects, and it fires on natural frustration signals in chat rather than requiring you to name every diagnostic command. The workflow walks through runtime logs first, then workflow or CI run status, then browser-level verification, and finally deployment and environment context—reporting what it found after each leg so you can decide the next fix. That structure matters because indie builders often bounce between tailing logs, re-running builds, and refreshing localhost without a checklist, which wastes tokens and time. Use it when you are blocked on “is it my code, my env, or the platform?” and you want a repeatable escalation path before rewriting features blindly.50installs46Vercel ServicesVercel Services is an integration skill from the Vercel plugin that guides solo builders through deploying more than one runtime inside a single Vercel project. It is aimed at monorepo layouts where a Python FastAPI or Go service lives next to a Next.js or Vite frontend, without splitting into separate hosting accounts prematurely. The skill references official Vercel Services docs and encodes detection patterns—backend folders, pyproject.toml, go.mod, uvicorn commands, and vercel dev with local linking—so agents propose the right service boundaries and config files. For indie SaaS founders, that means one deploy surface for API plus UI while you are still iterating. Use it when conversation or repo signals mention backend and frontend together, Python or Go on Vercel, or multi-service monorepos. It is narrower than generic Vercel hosting skills because it focuses on service composition rules rather than marketing or analytics setup.44installs47Vercel FlagsVercel-flags is an agent skill from the Vercel plugin that steers solo builders through Vercel’s feature-flags platform: defining flags in conventional files like flags.ts, wiring the Flags SDK for Next.js or SvelteKit, and using the unified dashboard for gradual rollouts and experiments. It is aimed at indie SaaS teams who deploy on Vercel and want safer releases without a separate flag vendor sprawl. Invoke it when you are adding toggles, rolling features to a percentage of users, or aligning runtime evaluation with Vercel’s Flags Explorer rather than hard-coding env switches. The skill points to official Vercel feature-flag docs and flags-sdk.dev so agents install the right packages and respect path and import conventions your repo may already use.41installs48Sign In With Vercelsign-in-with-vercel is a Vercel Labs plugin skill that walks solo builders through using Vercel as an OAuth 2.0 and OpenID Connect identity provider for end-user authentication. It is meant when you are adding login to a Vercel-hosted app and want account-based sign-in without standing up a separate IdP from scratch. Metadata points agents at official Vercel docs, common route globs, and retrieval aliases like vercel oauth and vercel login. The skill intentionally composes with others: if the codebase already uses NextAuth or Auth.js, it loads auth guidance for provider and middleware patterns; when OAuth secrets are present, it surfaces env-vars guidance for vercel env and secure configuration. Builders on Next.js App Router or Pages who need a concrete identity-provider integration—not generic auth theory—should invoke it while scaffolding auth API routes and login UX.38installs49CmsThe Vercel cms skill gives solo builders a single playbook for content-driven sites: pick a headless CMS, model content, embed or host the studio, and connect Next.js on Vercel with preview and revalidation so editors see changes before production. It leans on Vercel’s CMS solution docs and recognizes common file layouts—sanity.config, lib/sanity, api/draft, api/revalidate, and embedded studio routes—so agents install the right packages and scaffold the integration instead of guessing per vendor. Sanity’s native Marketplace path gets first-class treatment alongside Contentful and other majors. Use it when you are building marketing sites, blogs, or product marketing layers that must stay fast on the edge while non-developers ship copy. After build, the same webhooks and preview flows support launch content ops without a separate skill.32installs50PaymentsPayments is a Vercel plugin agent skill that walks solo and indie builders through Stripe integration on modern Next.js deployments. It is meant when you are implementing paid tiers, one-time checkout, or recurring subscriptions and need consistent guidance on Marketplace wiring, server routes for webhooks, and client-side Elements or Checkout. The skill aligns file paths and package installs with Stripe’s documented flows so your agent edits the right API handlers and lib modules instead of inventing structure. It emphasizes webhook verification, session creation, and subscription lifecycle concerns that block shipping if done ad hoc. Builders on Claude Code, Cursor, or Codex who already chose Vercel get opinionated, doc-linked steps rather than a generic payment tutorial.31installs51EmailEmail is a Vercel plugin agent skill that steers implementation of outbound mail on Vercel using Resend and React Email. Solo builders adding welcome messages, password resets, receipts, or notification hooks get structured guidance on installing dependencies, verifying sending domains, structuring template components, and exposing secure API routes under conventional `app/api/send` or `lib/resend` paths. The skill activates when your repo matches email-related file patterns or package installs, reducing guesswork around marketplace-native Resend setup versus ad-hoc SMTP. It emphasizes transactional patterns suitable for serverless route handlers on Vercel rather than full marketing automation suites. Use it while building backend integrations so templates, env vars, and send endpoints stay consistent with Vercel and Resend documentation links embedded in the skill metadata.30installs52GeistGeist is an agent skill that helps solo builders adopt Vercel’s default Geist typography family in Next.js projects without guessing import paths or layout conventions. It is aimed at indie devs shipping SaaS or marketing sites on the App Router who want interfaces that match Vercel’s aesthetic out of the box. Invoke it when you are adding or swapping fonts in layout.tsx, wiring CSS variables in globals.css, or aligning Tailwind theme fonts with Geist Sans, Mono, or Pixel. The skill encodes path and import heuristics so the agent focuses on the right files in single-app and apps/* monorepo layouts. It also documents when to hand off to broader Next.js guidance for next/font optimization. Use it early in UI polish so typography stays consistent before you layer components and marketing pages.29installs53SatoriSatori is an integration-focused agent skill from the Vercel plugin that teaches how to use Vercel’s Satori library—and the higher-level @vercel/og helpers—to turn HTML and CSS into SVG for social preview images. Indie builders shipping Next.js apps on the App Router or Pages router use it when adding opengraph-image routes, dynamic OG endpoints, or branded share cards without maintaining brittle PNG templates. The skill ties file conventions, import patterns, and package install commands to real paths such as app/**/opengraph-image and pages/api/og so agents suggest code that matches your repo layout. It matters for Build-time UI work and Launch-time SEO because correct 1200×630 previews improve click-through when you distribute on X, LinkedIn, or Slack. Expect constraints around supported CSS subsets and runtime choice (including wasm) rather than full browser rendering.28installs54Ai Generation PersistenceAI generation persistence is an agent skill for solo builders shipping Vercel-style AI apps who need every LLM or media generation recorded—not ephemeral stream-only output. It encodes Vercel AI SDK practices: generate a unique ID, persist prompts and results to a database or Blob store, link each run to a stable URL users can reopen, and log token cost alongside the record. The skill targets Next.js route handlers under generate/generations APIs and dynamic chat or generate pages, with import signals for `ai`, `@ai-sdk/*`, `@vercel/blob`, and ID libraries. Use it when prompt phrases mention saving chat history, generation history, or persisting AI output to a database. It complements streaming UX by making generations auditable, shareable, and billable without building persistence ad hoc in chat.25installs55MicroMicro is a Vercel plugin agent skill for solo builders who still touch the legacy micro npm package—small async HTTP servers and API endpoints—or who inherit repos that import micro and call send on the response object. It explains when micro fits historically and how it differs from current Vercel serverless patterns. Built-in validation recommends migrating legacy imports to modern route handlers that return Response.json, with an explicit chainTo vercel-functions skill and upgrade rationale. Retrieval intents cover creating microservices, HTTP servers, and API endpoints with micro. Use during Build when extending old services or during Ship review when an agent scans for deprecated patterns before production deploy. Intermediate complexity assumes comfort with Node HTTP and Vercel deployment context.25installs56Geistdocsgeistdocs is an expert agent skill for Vercel’s Geistdocs documentation stack: Next.js plus Fumadocs for MDX pages, centralized config in geistdocs.tsx, and optional AI chat, internationalization, and reader feedback. Solo builders launching OSS libraries, SaaS APIs, or agent skill catalogs use it when npx @vercel/geistdocs init, wiring content/docs/**/*.mdx, or tuning site metadata beats hand-rolling a docs framework. The skill maps official preview docs, GitHub source, sitemap URLs, and concrete file patterns so agents do not hallucinate Fumadocs APIs. It spans build-time site setup and launch-ready llms.txt/SEO surfaces, making it a practical default for indie teams who want Vercel-shaped docs without enterprise doc platforms.24installs57Styled Jsxstyled-jsx is Vercel’s scoped CSS-in-JS library for React and Next.js: you colocate real CSS in components using `<style jsx>`, with optional dynamic values, global rules, and the `styled-jsx/css` helper for external sheets. It is aimed at solo builders and small teams who want styling that stays local to a component instead of fighting global stylesheets or generated class utilities. Reach for this skill when you are wiring the babel plugin, adding package installs, or debugging hover states, themes, and responsive rules inside Next.js pages and shared UI. It matters because scoped CSS reduces accidental overrides in growing codebases while keeping the familiar CSS authoring model agents can edit reliably.3installs58Edge Runtimeedge-runtime is a Vercel plugin skill that keeps solo builders from shipping Edge Functions and Middleware that silently assume a full Node.js process. When you add edge-runtime packages, edit edge-runtime.config files, or import from edge-runtime, the skill surfaces constraints that matter on Vercel’s lightweight isolate runtime: no fs, child_process, net, dns, CommonJS require, or eval—and practical alternatives such as fetch-backed I/O and KV storage. It fits indie SaaS and API projects that need global low-latency routes, auth middleware at the edge, or A/B gates without maintaining a separate worker platform. Use it while implementing backend routes and again when hardening before deploy so CI or agent-assisted review catches forbidden imports before production. Intermediate complexity reflects needing familiarity with ESM, Vercel’s deployment model, and the difference between Node serverless and edge isolates. It complements generic backend skills by encoding Vercel-specific footguns as severity-ranked validation messages.2installs59Nccncc is a Vercel plugin agent skill that makes solo builders fluent in @vercel/ncc—the Node.js Compiler Collection that collapses an entry file and its dependency graph into a single bundle. It is aimed at indie developers shipping serverless functions who do not want to upload entire node_modules folders, authors publishing self-contained CLI binaries, and teams shrinking Docker layers. The skill explains installation (global or devDependency), basic ncc build with output directories, watch mode for iterative compile cycles, and ncc run for quick local execution without writing dist files. Guidance extends to CLI option literacy so agents recommend minification, external packages, and asset handling appropriately. Priority metadata and import/bash pattern hooks help Cursor and Claude Code surface the skill when package manifests reference @vercel/ncc or when build scripts invoke ncc build, keeping advice aligned with Vercel’s bundling model rather than generic webpack tutorials.2installs60StreamdownStreamdown is Vercel’s streaming-aware Markdown renderer for React apps that show live AI output. Solo and indie builders shipping chat sidebars, copilots, or in-app assistants often hit flicker, broken lists, and half-finished code fences when they pipe partial strings through classic react-markdown. This skill walks through installation, Tailwind v3 and v4 scanning setup, and the minimal ChatMessage pattern that passes accumulated content plus an isAnimating flag so unterminated blocks degrade gracefully instead of breaking layout. It is aimed at anyone already on npm, pnpm, bun, or yarn who wants a focused integration guide rather than ad-hoc parsing hacks. Use it during frontend build when you are replacing or comparing react-markdown in a streaming pipeline, wiring design-system styles, or aligning component file paths with agent-assisted edits. The metadata also flags import and bash patterns so coding agents can find the right files after you add the dependency.1installs