KhazP/vibe-coding-prompt-template

Templates and workflow for generating PRDs, Tech Designs, and MVP and more using LLMs for AI IDEs

Vibe‑Coding Prompt Template – What It Is

Vibe‑Coding is a structured, five‑step workflow that helps you use an AI coding assistant (Claude, Gemini, ChatGPT, Cursor, VS Code Copilot, etc.) to turn an idea into a working MVP. The repository does not contain a runnable application; instead it provides:

Ready‑to‑copy prompts (part1‑deepresearch.md, part2‑prd‑mvp.md, …) that you paste into any chat‑based AI tool.
A small CLI (npx vibeworkflow) that can scaffold the necessary “agent files” (AGENTS.md, MEMORY.md, agent_docs/…) inside your own project folder.
Detailed documentation that walks a user through the five phases:
1. Deep Research – validate the idea with market data.
2. Product Requirements (PRD) – define the MVP scope.
3. Technical Design – pick a stack, deployment target, and AI model.
4. Agent Files – generate the instruction set the AI will follow.
5. Build – iteratively code, review, and verify with the AI.
Guidance on advanced topics such as memory management, multi‑agent orchestration, model‑selection matrices, and AI‑safety checklists.

The repo also lists a few real‑world projects that were built using this workflow (e.g., vibeworkflow.app, moneyvisualiser.com).

Who Might Use It

Solo developers or small teams who want a repeatable process for AI‑assisted coding without writing their own prompt engineering scaffolding.
Educators / learners looking for a concrete example of how to structure interactions with LLM‑powered IDEs.
Product managers who need a clear, documented hand‑off from idea to code that can be audited.

How It Works (High‑Level)

Copy a prompt file (e.g., part1-deepresearch.md) into your chat window.
The AI asks a few clarifying questions and produces a research document.
Repeat with the PRD and Technical Design prompts, each time feeding the previous output as context.
Run npx vibeworkflow (or paste the “agent file” prompts) to generate a set of markdown files that act as the AI’s “contract” (AGENTS.md, MEMORY.md, etc.).
In your AI‑enabled IDE, tell the assistant to read AGENTS.md and start building the MVP feature‑by‑feature, pausing for review after each chunk.

What the Repository Contains

Item	Purpose
`part*.md` files	Prompt templates for each planning step.
`AGENTS.md` (generated)	Central instruction set that tells the AI how to behave across the project.
`MEMORY.md` (generated)	Persistent, file‑based memory to avoid over‑loading the chat context.
`agent_docs/` (generated)	Human‑readable docs (tech stack, testing strategy, etc.) that the AI can reference.
`npx vibeworkflow` CLI	Small Node‑based helper that auto‑creates the above files based on the PRD/TechDesign you placed in `docs/`.
`README.md` (this file)	Full walkthrough, tool recommendations, safety guidelines, and links to example projects.

Why It Might Be Useful

Consistency – By codifying the planning and hand‑off steps, you reduce the chance of the AI “drifting” or forgetting earlier decisions.
Tool‑agnostic – Works with any chat‑based LLM (Claude, Gemini, ChatGPT) and with a variety of IDE extensions (Cursor, VS Code Copilot, Claude Code, etc.).
Safety‑first – Includes checklists for model selection, cost limits, prompt‑injection protection, and artifact‑first memory, encouraging responsible AI use.
Rapid prototyping – The workflow is designed to be completed in a few hours, making it suitable for hackathons or quick MVPs.

Getting Started (Quick Steps)

Install Node (if you don’t have it).
Open a new folder for your app.
Run npx vibeworkflow through your AI‑enabled IDE (or copy the prompts manually). The CLI will create the docs/ and agent_docs/ scaffolding.
Paste the part1‑deepresearch.md prompt into your chosen AI chat and follow the interaction.
Repeat for parts 2 and 3, saving the outputs into docs/.
Run npx vibeworkflow doctor to verify the generated agent files.
Give the AI its first command (e.g., “Read AGENTS.md, propose a Phase 1 plan, wait for my approval, then build.”) and iterate.

Limitations

The repository does not provide a ready‑made codebase; you still need to write or generate the actual application code via the AI.
Success depends on the capabilities of the underlying LLM and the quality of the prompts you feed it.
It is geared toward web or lightweight app projects; it explicitly warns against using the workflow for safety‑critical or heavily regulated systems.

License & Contributions

Licensed under the MIT License.
Pull requests are welcomed; the author encourages community‑driven improvements to the prompts and documentation.

In short: the vibe‑coding‑prompt‑template repo is a practical, open‑source guide and CLI for orchestrating AI‑assisted software development, turning a high‑level idea into a concrete MVP through a repeatable, safety‑aware workflow.