---
name: prompt-better
summary: Improve rough AI prompts into clear PRDs, build prompts, debug prompts, review prompts, and acceptance criteria for AI coding workflows.
description: Use this skill whenever the user asks to improve a prompt, create a PRD, make a Codex/Claude Code/Manus/Base44/Google AI Studio build prompt, turn an idea into requirements, write a debugging prompt, or review a prompt before using it with an AI builder.
---

# Prompt Better Skill

## Purpose
Turn rough ideas into precise, usable prompts for AI coding and product-building tools. Prioritize clarity, constraints, acceptance criteria, honest data handling, and copy-paste-ready output.

## Core Workflow
1. Identify the user's goal in one sentence.
2. Classify the requested output:
   - PRD
   - Build prompt
   - Debug prompt
   - Design prompt
   - Review/evaluator prompt
   - Deployment prompt
   - General prompt improvement
3. Add missing structure:
   - Goal
   - Context
   - Requirements
   - Constraints
   - Output format
   - Acceptance criteria
   - Non-goals / out-of-scope items
4. Preserve the user's intent and language, but make it clearer and more specific.
5. Do not invent facts, APIs, live data, pricing, laws, benchmarks, or tool capabilities.
6. If data is fictional or placeholder, label it clearly as sample data.
7. Return a final copy-paste-ready prompt.

## PRD-First Rule
For app, SaaS, dashboard, website, AI agent, or coding projects, recommend starting with a PRD before writing code unless the task is very small.

A strong PRD should include:
- Product name
- Product summary
- Product goal
- Target users
- Core problem
- User stories
- Main features
- Screens or tabs
- User flows
- Data requirements
- Design requirements
- Technical requirements
- MVP scope
- Out-of-scope items
- Acceptance criteria
- Future features

## Prompt Quality Rubric
Score prompts from 1-10 across:
- Goal clarity
- Context quality
- Requirements
- Constraints
- Output format
- Examples/references
- Acceptance criteria
- Safety and honesty
- Technical precision
- Actionability

If any score is below 8, rewrite that part.

## Build Prompt Template
Use this when creating a coding prompt:

You are a senior product-minded full-stack developer.

Build [app name].

## Goal
[goal]

## Users
[target users]

## Main Features
[features]

## User Flow
[flow]

## Tech Stack
[stack]

## Data
[data model and sample data rules]

## Design
[visual style]

## File Structure
[folders and components]

## Rules
[constraints and non-goals]

## Acceptance Criteria
[how to verify success]

## Final Output
[commands, files, setup notes]

## Debug Prompt Template
Use this when the user gives an error:

I am getting this error:
[paste full error]

What I was doing:
[action]

Expected behavior:
[expected]

Actual behavior:
[actual]

Please:
1. Identify the root cause.
2. Fix only the related files.
3. Preserve existing working features.
4. Explain the change simply.

## Style Rules
- Be direct.
- Do not over-ask questions.
- Make reasonable assumptions when safe and label them.
- Prefer structured sections and checklists.
- Make final prompts copy-paste-ready.
- Protect the user from fake live data or unsupported claims.
- Keep the MVP focused before suggesting advanced features.