This skill scaffolds and customizes a production-ready SaaS website on top of the TencentEdgeOne/saas-starter template (Next.js 14 + TypeScript + Tailwind + Supabase + Stripe + AI), and prepares it for deployment to EdgeOne Pages. It should be used when the user wants to start a new SaaS project, AI SaaS, or AI tool site from scratch — e.g. "build a SaaS site on EdgeOne Pages", "scaffold an AI SaaS with saas-starter", "帮我用 saas-starter 搭一个 SaaS", "初始化一个 AI 图片生成 SaaS 部署到 EdgeOne Pages", "create a new AI SaaS project". The skill first asks whether the user has their own product Prompt. If yes, it parses the Prompt to derive brand, features, color palette, and copy; if no, it runs a short built-in questionnaire (product direction / color / payments on-off / AI on-off / i18n). It then uses `npx degit TencentEdgeOne/saas-starter` to bootstrap, applies customizations, generates `.env.local`, and guides local run. Do NOT trigger when the user already has a running project and only wants to add a single function/API
This skill scaffolds and customizes a production-ready SaaS website on top of the TencentEdgeOne/saas-starter template (Next.js 14 + TypeScript + Tailwind + Supabase + Stripe + AI), and prepares it for deployment to EdgeOne Pages. It should be used when the user wants to start a new SaaS project, AI SaaS, or AI tool site from scratch — e.g. "build a SaaS site on EdgeOne Pages", "scaffold an AI SaaS with saas-starter", "帮我用 saas-starter 搭一个 SaaS", "初始化一个 AI 图片生成 SaaS 部署到 EdgeOne Pages", "create a new AI SaaS project". The skill first asks whether the user has their own product Prompt. If yes, it parses the Prompt to derive brand, features, color palette, and copy; if no, it runs a short built-in questionnaire (product direction / color / payments on-off / AI on-off / i18n). It then uses `npx degit TencentEdgeOne/saas-starter` to bootstrap, applies customizations, generates `.env.local`, and guides local run. Do NOT trigger when the user already has a running project and only wants to add a single function/API (use edgeone-pages-dev). Do NOT trigger for pure deployment of an existing project (use edgeone-pages-deploy). Do NOT trigger for non-SaaS static sites, blogs, or portfolios.
metadata
{"author":"edgeone-pages-saas","version":"1.0.0"}
EdgeOne Pages SaaS Scaffolding Skill
Scaffold and customize a SaaS website from the TencentEdgeOne/saas-starter template, then hand it off to the deploy skill.
When to use this skill
User wants to start a new SaaS site (AI SaaS, AI tool, membership site, etc.) on EdgeOne Pages
User mentions the saas-starter template explicitly
User wants a Next.js + Supabase + Stripe SaaS foundation with credits / auth / payments / AI
Do NOT use for:
Adding a single API/function to an existing project → use edgeone-pages-dev
Deploying an already-built project → use edgeone-pages-deploy
Plain static sites, blogs, landing pages without SaaS features
How to use this skill (for the coding agent)
Follow the main flow below top-to-bottom. Load references/*.md only when the flow tells you to — do NOT preload everything.
⛔ Critical Rules (never skip)
ALWAYS ask the user about their own Prompt FIRST, before doing anything else. This is the entry gate of the skill.
Never fabricate the user's product — if information is missing and the user has no Prompt, run the built-in questionnaire in references/user-prompt-intake.md.
Never rewrite the template's source tree wholesale. Apply targeted edits only — see references/customization.md for the exact files to touch.
Use npx degit TencentEdgeOne/saas-starter <project> to bootstrap (no .git history). Do not run git clone unless degit is unavailable.
Never commit .env.local or any real API keys to git. Make sure .gitignore covers it (the template already does).
DO NOT deploy inside this skill. Deployment belongs to edgeone-pages-deploy. In Step 6, BEFORE telling the user "you can deploy now", MUST verify that edgeone-pages-deploy skill is installed locally (check whether ~/.codebuddy/skills/edgeone-pages-deploy/ — or its Windows equivalent %USERPROFILE%\.codebuddy\skills\edgeone-pages-deploy\ — exists). If not installed, follow the guided-install branch in Step 6.3 instead of just leaving a text hint.
The template's package.json uses next dev (NOT edgeone pages dev). Only suggest edgeone pages dev when the user actually needs to develop/debug Edge Functions, Cloud Functions, or KV locally — and then guide via edgeone-pages-dev skill.
Confirm destructive operations before running (removing unused modules, deleting locale files, rewriting config). Show the plan, then proceed after user says go.
Main Flow
Step 1 — Gate question: does the user have a Prompt?
This is mandatory and happens FIRST. Use the IDE's question UI (e.g. ask_followup_question) to ask:
开始之前先问一下：你有没有自己准备好的 Prompt / 产品描述？
有 — 贴进来，我会据此定位品牌 / 功能 / 配色 / 文案
没有 — 我会用内置问答引导你，问几个关键问题
完全默认 — 直接按模板默认（AI 图片生成 SaaS，紫色主题，中英双语）落地，后续再改
Branch:
User provides a Prompt → read references/user-prompt-intake.md § "Prompt 解析" and extract a structured spec
User chooses the questionnaire → read references/user-prompt-intake.md § "内置问答" and run through the questions
User chooses default → skip intake, use default spec in references/user-prompt-intake.md § "默认规格"
Output of this step: a Spec object with at least the following fields:
(rare) remove /auth routes, strip Supabase; warn the user this breaks credits/admin
features.credits = false
remove credits UI, simplify DB schema from database_init.sql
features.aiImage = false
remove src/app/ai/, remove @ai-sdk/* deps from package.json, remove FAL_API_KEY / FIREWORKS_API_KEY from env
features.blog = false
remove src/app/blog/, content/blog/
features.admin = false
remove src/app/admin/, remove ADMIN_EMAILS env
features.cms = false
remove cms/contentful/, Contentful deps, contentful:export script, CMS env vars
features.i18n (locales)
prune dictionaries/ to requested locales
Apply edits one module at a time and report progress. If a destructive change has downstream impact (e.g. removing auth kills credits), call it out and ask for confirmation.
Step 4 — Environment variables
Read references/env-setup.md and:
Copy .env.example → .env.local
Based on the enabled features, produce a filtered env list (only keys the user actually needs)
For each required key, either:
Ask the user to paste the value, or
Guide them to the provider's dashboard and wait
Write the collected values into .env.local
Remind: never commit .env.local
Optional Supabase schema step: if features.auth or features.credits is on, point the user to database_init.sql and the Supabase SQL Editor (one-paragraph summary is enough — this skill doesn't manage DB execution).
Step 5 — Local verification
Run (or tell the user to run):
npm run dev # starts Next.js on http://localhost:3000
The template's default dev command is next dev (port 3000). If the user wants to debug Edge Functions / Cloud Functions / KV Storage on EdgeOne Pages locally, they should switch to edgeone pages dev (port 8088) — and load the edgeone-pages-dev skill for guidance.
Ask the user to open the URL and confirm:
Homepage renders
Language switcher works (if i18n enabled)
Auth signup works (if Supabase env filled)
If something breaks, diagnose common causes:
Missing env var → check .env.local matches the filtered list from Step 4
Supabase schema not initialized → run database_init.sql
Port conflict → change port via next dev -p <port>
