build_skill

Overview

This skill helps non-technical founders build SaaS features using AI coding tools, choose the right tool, write effective prompts, and iterate on AI-generated code. It focuses on practical workflows: scoping, handing a spec to the AI, testing, giving targeted feedback, and stopping after 2–4 iterations when the feature meets acceptance criteria.

How this skill works

The skill guides you through tool selection (Lovable for fast MVPs, Claude Code for context-heavy work, Replit for learning, Cursor for power users) and provides clear prompt templates for each tool. It prescribes a build workflow: start with a scoped spec, ask the AI to implement, test happy paths and edge cases, give specific feedback using a short template, and iterate until the feature meets the defined stop conditions.

When to use it

• You need to build a new feature but can’t code it yourself.
• You have an existing codebase and want AI to add context-aware changes.
• You want to learn by doing with AI-assisted coding environments.
• You must decide which AI coding tool best fits your project stage.
• You need a practical review and iteration loop for AI-produced code.

Best practices

• Always start with a scoped spec (1-3 hour chunk) before asking AI to build.
• Pick the tool to match the situation: Lovable for MVPs, Claude Code for context, Replit to learn, Cursor for advanced needs.
• Test the happy path and edge cases on mobile and desktop—don’t just run the code.
• Give feedback in a specific template: What I tried; Expected; Got.
• Limit iterations to 2–4 rounds and stop when happy path, edge cases, and mobile work.

Example use cases

• Ask Lovable to scaffold a minimal MVP signup and onboarding flow.
• Use Claude Code to add a feature to an existing React+TypeScript project with references to src/components.
• Use Replit to build a simple frontend while learning React patterns.
• Use Cursor for power-user tweaks and complex integrations in an existing repo.
• Break a large feature into chunks like 'user auth flow' or 'dashboard with 4 metrics' and iterate.