Scaffold New App

Scaffold a New App

Meta-skill that asks the right questions, picks a stack, and routes to the appropriate scaffold-* skill. Stops the "decision paralysis" tax of starting a new project.

When to use

• Starting any new project where the stack isn't pre-decided.
• Helping someone else start a project and they don't know what to pick.
• Replacing a hacked-together prototype with a production-shaped scaffold.

When NOT to use

• Project is already underway — adding scaffolding to an existing app is surgery, not scaffolding.
• The user has already declared a stack ("scaffold a Rails app") — go straight to that scaffold.
• Constrained environment (regulated, customer-mandated stack) — follow the constraint.

The decision tree

Walk through these questions with the user, in order. Stop at the first answer that determines the stack.

Q1: What is this project for?

| Answer | Likely stack | |---|---| | B2B SaaS, multi-tenant | → scaffold-saas-starter | | Indie SaaS / B2C app | → scaffold-fullstack-app | | Internal tool | → scaffold-fullstack-app (or low-code if it's truly simple) | | API-only service | → scaffold-go-microservice (or fastapi-production if Python team) | | CLI tool | → scaffold-cli-tool | | Mobile app | → react-native-expo (cross-platform) or native-only stack | | Static marketing site | → Astro or plain Next.js (minimal) | | Background data pipeline | → dbt-data-modeling for SQL, custom Python/Go for everything else | | ML inference API | → fastapi-production |

Q2: Who's the team?

| Answer | Adjustment | |---|---| | Solo founder | Optimize for speed-to-deploy → indie SaaS scaffold + Vercel + Neon | | Small team (2–5) | Same indie scaffold; pin tooling versions | | Medium team (6–20) | SaaS scaffold; consider monorepo with apps/* packages | | 20+ engineers | Get a senior architect; this skill isn't for you |

Q3: What's the geographic / compliance footprint?

| Answer | Adjustment | |---|---| | India primary | Razorpay over Stripe; consult india-dpdp-act skill | | EU primary | GDPR posture from day 1; data residency in EU | | US/global | Stripe; consider pci-dss-checklist for payment-handling scope | | HIPAA | Specialist review needed; baseline scaffold + hipaa-audit skill | | None / dev tool | Skip compliance skills |

Q4: What's the deployment target?

| Answer | Adjustment | |---|---| | Vercel + Neon | Default for Next.js scaffolds | | Cloudflare | cloudflare-workers scaffold instead | | Railway / Fly | Same scaffolds; tweak deploy config | | Self-hosted Kubernetes | scaffold-go-microservice + kubernetes-helm | | Lambda | Probably wrong choice for a new app — challenge it |

Q5: Will this need real-time features?

| Answer | Adjustment | |---|---| | No | Default scaffolds | | Chat / collaborative | Add Durable Objects or Pusher; consider Cloudflare Workers | | Live dashboards | Server-Sent Events from API (cheaper than WebSocket) | | Multi-cursor / OT | This is a specialist build; out of scope for scaffolding |

The output

After the questions, this skill produces:

1. A stack decision summary — the actual stack picked, written down in docs/stack.md.
2. The route to the right scaffold-* skill — which then emits the project.
3. A list of follow-on skills the user will likely need — e.g. razorpay-integration, clerk-auth, sentry-monitoring.

When defaults are wrong

Push back on the user when their request implies a poor fit:

• "I want to build a SaaS in Rust" — challenge: do you actually have Rust expertise? Productivity hit is significant for non-CPU-bound web work.
• "I want to use [obscure framework]" — challenge: who else on the team can maintain this?
• "I want microservices from day 1" — challenge: do you have the team size to operate them? Default to a modular monolith.
• "I want both Stripe and Razorpay wired" — challenge: do you have the geographic split to justify this maintenance load?
• "I want my own auth" — challenge: have you used Clerk/Auth0? Building auth is six weeks you don't get back.

The right answer is sometimes "I want X anyway, here's why" — that's fine. But surface the trade-off; don't silently scaffold a regrettable choice.

Anti-patterns

• Scaffolding without asking — produces wrong stack 60% of the time.
• Asking 20 questions — most decisions can be defaulted; only 3–5 actually drive the stack.
• Defaulting to the latest fashion — "Use the new shiny" is a costly default. Recommend boring tech with track record.
• Picking microservices for "future scale" — you don't have scale; you have an extra week of ops setup.
• Picking a backend language nobody on the team knows — productivity loss outweighs theoretical benefits.
• Not committing the stack decision to the repo — six months later, nobody remembers why you picked X.
• Skipping the compliance question for an India/EU/healthcare app — late compliance work is 10x harder than early.

Verify it worked

• [ ] User answered Q1; you've picked the matching scaffold-* skill.
• [ ] Adjustments from Q3 (compliance) and Q4 (deploy target) are applied.
• [ ] docs/stack.md exists in the repo with the decision and rationale.
• [ ] User has a list of follow-on skills they'll likely invoke.
• [ ] Push-back happened if the request implied a poor fit.
• [ ] The scaffold runs on the user's machine without manual intervention beyond pnpm install.
• [ ] The user can describe the stack in one sentence ("Next.js App Router + Postgres RLS + Clerk + Razorpay + Brevo, on Vercel + Neon").

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: MaheshAwasare
• Source: MaheshAwasare/claude-skills-pro
• License: MIT

Install and usage instructions live in the source repository linked above.