# Sergio Bershadsky AI Replit Prompt > Replit Prompt Generator - Skill: `tomevault-io/sergio-bershadsky-ai-replit-prompt` (Agent Skill, multi-file: 2 files) - Install (CLI): `npx skillmds@latest add tomevault-io/sergio-bershadsky-ai-replit-prompt` - Raw SKILL.md: https://api.skillmd.com/api/skills/tomevault-io/sergio-bershadsky-ai-replit-prompt/raw - Safety review: pending (external: skill-scanner PASS, skillspector PASS) - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: AI & ML - Author: tomevault-io (https://skillmd.com/u/tomevault-io) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/tomevault-io/sergio-bershadsky-ai-replit-prompt --- # Replit Prompt Generator Transform requirements into optimized prompts for Replit Agent that maximize AI understanding and minimize iterations. ## When to Use - User wants to build something with Replit Agent - User needs to convert an idea into a Replit-ready prompt - User wants to improve an existing Replit prompt - User is preparing to start a new Replit project ## Core Principles (Replit's Official Guidelines) ### 1. Checkpoint — Structure Iteratively Break large goals into smaller, testable steps. Use checkpoints between phases. ### 2. Debug — Provide Detailed Context Include exact error messages, relevant code snippets, and steps already attempted. ### 3. Discover — Ask for Suggestions Request tool/library recommendations when unsure about approach. ### 4. Experiment — Iterate on Prompts Refine requests by adjusting wording or adding detail. ### 5. Instruct — Use Positive Goals Frame as "do this" rather than "don't do that". ### 6. Select — Focused Context Provide relevant context only; avoid overwhelming with unrelated information. ### 7. Show — Concrete Examples Reduce ambiguity with code snippets, mockups, or reference apps. ### 8. Simplify — Concise Language Use direct, short sentences. Avoid jargon. ### 9. Specify — Define Outputs State expected outputs, constraints, data formats, and edge cases. ### 10. Test — Plan First Outline features, data structures, and user flows before prompting. ## Prompt Structure Template When generating prompts for Replit, use this structure: ```markdown ## Project Overview [1-2 sentences: What is being built and its core purpose] ## Tech Stack - Frontend: [specific framework/library] - Backend: [specific framework/language] - Database: [specific database] - Authentication: [specific method] - Additional: [any other required tools/APIs] ## Core Features 1. [Feature 1]: [specific description with expected behavior] 2. [Feature 2]: [specific description with expected behavior] 3. [Feature 3]: [specific description with expected behavior] ## UI/UX Requirements - Design style: [modern/minimal/material/etc.] - Color scheme: [specific colors or "neutral professional"] - Layout: [sidebar/top-nav/dashboard/etc.] - Responsive: [yes/no, breakpoints if specific] ## Data Model [Describe main entities and their relationships] - Entity1: field1, field2, field3 - Entity2: field1, field2 (relates to Entity1) ## User Flows 1. [Flow 1]: Step A → Step B → Step C → Result 2. [Flow 2]: Step A → Step B → Result ## Constraints - [Constraint 1: e.g., "Must work offline"] - [Constraint 2: e.g., "No external API calls except X"] ## Success Criteria - [ ] [Specific testable criterion 1] - [ ] [Specific testable criterion 2] - [ ] [Specific testable criterion 3] ``` ## Procedure ### Step 1: Gather Requirements Ask the user about: 1. **What** they want to build (core functionality) 2. **Who** will use it (target users) 3. **How** it should work (key user flows) 4. **Why** (problem being solved) If user provides vague input, ask clarifying questions: - "What should happen when a user clicks X?" - "Should data persist between sessions?" - "Any specific design/framework preferences?" ### Step 2: Extract Technical Details Determine from user input or ask about: - Preferred tech stack (or let Replit decide) - Authentication requirements - Data storage needs - Third-party integrations - Performance requirements ### Step 3: Structure the Prompt Transform gathered information into the structured template above. **Bad prompt transformation:** ``` ❌ Bad: "Make a website for my business" ✅ Good: "Create a portfolio website for a freelance photographer with: - Home page with hero image carousel - Gallery page with filterable photo categories - Contact form with email integration - About page with bio and testimonials - Tech: React, TailwindCSS, EmailJS for form - Design: Clean, minimal, lots of whitespace" ``` ### Step 4: Add Specificity Enhance the prompt with: - **Concrete examples**: "Like Stripe's dashboard layout" - **Specific behaviors**: "Click 'Add' opens modal, not new page" - **Edge cases**: "If cart is empty, show 'Start shopping' CTA" - **Validation rules**: "Email must be valid format, password min 8 chars" ### Step 5: Present Draft Show the optimized prompt to the user: ``` ## Replit Agent Prompt **Optimized for:** [Build Mode / Plan Mode] **Estimated complexity:** [Low / Medium / High] --- [Generated prompt content] --- **Recommendations:** 1. [Tip for using this prompt effectively] 2. [Suggested follow-up prompts after initial build] Ready to copy to Replit? ``` Wait for user confirmation or modification requests. ### Step 6: Provide Iteration Guidance After delivering the prompt, advise on: - Using Plan Mode first for complex features - Setting up checkpoints between phases - Follow-up prompts for refinements ## Output Format ``` ## Replit Agent Prompt **Mode:** Build / Plan **Complexity:** Low / Medium / High --- [Structured prompt following template] --- ## Next Steps 1. **Initial Build**: Copy prompt to Replit Agent 2. **After MVP**: [Suggested enhancement prompt] 3. **Testing**: [What to verify after build] ## Checkpoint Strategy Phase 1: [Description] → Checkpoint Phase 2: [Description] → Checkpoint Phase 3: [Description] → Checkpoint ``` ## Bad vs Good Prompt Examples | Issue | Bad Prompt | Good Prompt | |-------|-----------|-------------| | Vague | "Fix my code" | "The login function in auth.js throws 'undefined user' on line 42 when email contains '+'" | | No scope | "Make a website" | "Create a 3-page portfolio: Home (hero, skills), Projects (grid with filters), Contact (form)" | | Negative | "Don't make it slow" | "Implement lazy loading for images, paginate lists to 20 items" | | No detail | "Add animation" | "Fade in hero text on page load over 0.5s, slide in cards from bottom on scroll" | | Overwhelming | "Build the entire backend" | "Implement user signup/login with JWT and a /api/profile GET endpoint" | ## Rules 1. **ALWAYS use positive framing** — "Do X" not "Don't do Y" 2. **ALWAYS include tech stack** — Even if "let Replit decide" 3. **ALWAYS define success criteria** — Testable checkboxes 4. **ALWAYS break into phases** — For anything beyond simple apps 5. **NEVER use vague terms** — No "nice", "good", "fast", "modern" without specifics 6. **NEVER assume context** — Replit Agent has no prior knowledge ## Additional Resources See reference files for detailed patterns: - **`references/examples.md`** — Full prompt examples for common app types - **`references/tech-stacks.md`** — Recommended tech stack combinations --- > Converted and distributed by [TomeVault](https://tomevault.io/claim/sergio-bershadsky) — claim your Tome and manage your conversions.