Generated Computers
Locals: an Apple Silicon engine that runs local LLMs up to 3x faster. A product of Generated Computers.
翻譯準備中
個人 GitHub 帳號下公開儲存庫的擁有者可以登入並驗證所有權,自行管理介紹,無需管理員核准。
登入並產生驗證檔案,在 30 分鐘內將其提交至此儲存庫的預設分支。其他分支或分叉儲存庫中的檔案無效。然後在下方檢查檔案,即可無需管理員核准直接管理專案。請保留檔案以維持管理權限。
如需登入,將前往登入頁面。登入後請回到此處,再次點擊所選的驗證按鈕。
在專案資料夾中執行 /nomorevibe verify,即可驗證所有權並自行管理介紹。
一分鐘完成驗證Nexity is an open source system for starting and growing AI-native websites on Next.js + Sanity.
翻譯準備中
取自儲存庫 README
正在載入完整 README。
檢視完整 README ↗Nexity is an open source system for starting and growing AI-native websites on Next.js + Sanity. Its starter, base/, is a pnpm monorepo with two packages:
web/: the Next.js 16 app (App Router, Tailwind CSS 4, TypeScript), with the Sanity Studio embedded at /admin.studio/: a standalone Sanity Studio with the same schemas. Schemas live here and web/ imports them.This repository holds the starter, a demo site built on it (theme-meridian/) and, in .agents/, the agent layer that works with them: skills and read-only agents (reviewers, auditors) for any coding agent (Claude Code, Codex, Cursor and others).
| Version | |
|---|---|
| Next.js | 16.4 (App Router, Turbopack) |
| React | 19.3 |
| Sanity Studio | 6.20 (next-sanity 13.3, @sanity/client 8.11) |
| Tailwind CSS | 4.3, with shadcn/ui |
| TypeScript | 7.0 |
| Zod | 4.6 |
| Oxlint / Oxfmt | 1.89 / 0.74 |
| c15t (cookie consent) | 3.1 (@c15t/nextjs) |
| Resend | 6.34 |
| Node.js | 22.12 or newer |
| pnpm | 12.15 |
Versions are the minimums in each starter's package.json; both starters use the same ones.
| Entry | What you get | How to start |
|---|---|---|
base/ | Unbranded starter. Same features as Meridian, neutral look (default Tailwind palette, Inter). Starts on an empty dataset and shows a Next steps screen until you add the first content. | npm create sanity@latest -- --template Pagepro/nexity/base |
theme-meridian/ | Demo site for Meridian, a fictional B2B brand: full design, product mocks and seed content. | npm create sanity@latest -- --template Pagepro/nexity/theme-meridian, then pnpm seed |
.agents/ | Agent skills and read-only agents for setting up a project, creating sections, fixing links, writing alt texts and auditing Sanity use. | Copy or symlink into your project's .agents/ (see Agent layer) |
sanity init creates or picks a Sanity project and dataset, downloads the starter folder, writes web/.env.local and studio/.env with the project id and dataset, creates the API tokens, adds the localhost CORS origins and installs the dependencies. Then:
pnpm dev # http://localhost:3000 (site) and http://localhost:3000/admin (Studio)
Locally, the standalone Studio at http://localhost:3333 (also started by pnpm dev) reloads faster than /admin. Production serves only the Studio embedded at /admin.
In theme-meridian/, run pnpm seed to import the Meridian dataset (pages, posts, case studies, images) into the dataset you picked. The seed replaces documents with the same IDs, so use a fresh dataset.
Each starter has its own README.md and docs/ with setup details, environment variables and deployment notes.
base | theme-meridian |
|---|---|
Both starters have the same features. They differ only in look and seed content. The full list for editors and marketers is in each starter's docs/FEATURES.md.
/admin on the same domain, click-to-edit visual preview, live draft preview, tag-based revalidation on publish (no rebuilds), and a Sections Gallery with sample content for every block.robots.txt with a single IS_INDEXABLE launch switch; schema.org JSON-LD; pre-launch GROQ checks for missing or out-of-range titles and descriptions (docs/guides/before-release.md).llms.txt, and every page as Markdown (add .md to any URL).sanity.blueprint.ts sets up CORS origins and the revalidation webhook from env vars.AGENTS.md (how to build), DESIGN.md (how it looks) and BRAND.md (what it says) give any coding agent the project's rules, plus skills and read-only agents in .agents/. See Agent layer.Every coding agent working on a Nexity project reads the same files. Nothing in them is tied to one vendor.
| File | Role | Read by |
|---|---|---|
AGENTS.md (root, web/, studio/) | How to build: architecture, commands, conventions, the step-by-step checklists | Codex, Cursor and most agents natively; Claude Code through CLAUDE.md, a one-line @AGENTS.md import |
DESIGN.md | How it looks: tokens in the frontmatter, then section anatomy and do's and don'ts | Every skill that touches what a visitor sees, and styling-fidelity-reviewer |
BRAND.md | What it says (a filled-in example in Meridian, a template in base): audience, voice, messaging, the invented world of names and numbers | create-section, project-setup, generate-alts, product-mock |
.agents/skills/<name>/SKILL.md | Repeatable workflows (set up a project, create a section, fix links, write alts, audit Sanity use) in the open Agent Skills format | Codex from .agents/skills/; Claude Code from .claude/skills/, a symlink to it |
.agents/agents/<name>.md | Read-only reviewers and auditors with a fixed checklist and output format | Claude Code as subagents (.claude/agents/ symlink); any other agent reads the file and walks the checklist |
In this repository the symlinks are already in place, so Claude Code and Codex both pick up the skills when you open it. The skills and agents are plain files, not a plugin, and sanity init --template does not bring them along, so copy or symlink the ones you want into your project. Keep .agents/ as the one copy and point Claude Code at it with two relative symlinks:
# without a clone: fetch one skill, or all agents
npx degit Pagepro/nexity/.agents/skills/create-section .agents/skills/create-section
npx degit Pagepro/nexity/.agents/agents .agents/agents
# from your project root, with this repository cloned next to it
mkdir -p .agents/skills .agents/agents
# copy a skill
cp -R ../nexity/.agents/skills/create-section .agents/skills/
# or symlink it, so it follows updates to the clone
ln -s "$(realpath ../nexity/.agents/skills/create-section)" .agents/skills/create-section
# agents used by the skills
cp ../nexity/.agents/agents/*.md .agents/agents/
# Claude Code reads .claude/; point it at the same files
mkdir -p .claude && ln -s ../.agents/skills .claude/skills && ln -s ../.agents/agents .claude/agents
project-setup, create-section, sanity-module-catalog, pr-writeup, work-humanizer, product-mock. product-mock works only in theme-meridian. pr-writeup needs work-humanizer next to it.link-fix, generate-alts, sanity-optimizations.brand-kit, site-contradictions, blog-post-quality-review, internal-links, blog-post-update (needs blog-post-quality-review, source-review and work-humanizer next to it). They install as the Claude Code plugin marketing-skills (/plugin marketplace add Pagepro/nexity, then /plugin install marketing-skills@nexity) or with npx skills add Pagepro/nexity; marketing-setup then walks a marketer through the rest. Start from MARKETING-SKILLS-README.md..agents/agents/: the reviewers content-model-architect, styling-fidelity-reviewer, sanity-schema-reviewer and groq-parity-reviewer (run by create-section) and link-reviewer (run by link-fix and create-section); the auditors sanity-query-auditor, sanity-image-auditor and sanity-delivery-auditor (run by sanity-optimizations); and alt-text-writer (run by generate-alts).work-humanizer comes from phuryn/work-humanizer (MIT, license file kept in the skill folder)..agents/README.md describes each skill and agent; docs/guides/skills.md in either starter covers installing them.
.nvmrc in each starter)packageManager field; corepack enable picks it up)See CONTRIBUTING.md. To report a vulnerability, see SECURITY.md.
MIT © 2026 Pagepro