Analysis updated 2026-05-18
Turn a rough product idea into a Markdown PRD a coding AI can act on directly.
Reorganize an existing PRD's features into MVP, Next, Later, and won't-do buckets.
Get a plain-language comparison of technical approaches mapped to your feature list.
Produce acceptance criteria written as precondition, user action, and observable result.
| 1killermouse/vibe-coding-prd-skill | 0311119/free_registertool | 18597990650-lab/multi-agent-game | |
|---|---|---|---|
| Stars | 24 | 24 | 24 |
| Language | — | Python | Python |
| Setup difficulty | easy | hard | moderate |
| Complexity | 1/5 | 4/5 | 3/5 |
| Audience | pm founder | developer | developer |
Figures from each repo's GitHub metadata at analysis time.
Clone and copy into your Codex skills folder, then restart Codex, trigger phrases are in Chinese.
This project is a skill for Codex or Claude Code that turns messy requirements into a clean, engineering-ready document called a PRD, short for product requirements document. It takes whatever you feed it, whether that is a rough idea, an existing PRD, notes on competing products, or a mix of all three, and reorganizes it into a Markdown file that a coding AI can act on directly without guessing what you actually meant. The README explains that ordinary PRDs usually mix business background, presentation language, and vague descriptions, which forces a coding AI to guess at scope, interactions, and how to know a feature is actually done. This skill fixes that by walking through four required checkpoints before it will produce anything final. First, it defines the requirement itself: who the users are, what situation they are in, what problem they have, and what is explicitly out of scope. Second, it sorts features into what ships now, what comes next, what comes later, and what will not be built at all, so the plan does not keep growing without limit. Third, it settles on a technical approach, comparing options in plain terms and mapping each feature to how it would actually be built. Fourth, it sketches the basic screens, the main user paths, and rough wireframes, so everyone agrees on how the product behaves before code gets written. The skill will not produce a finished PRD until a person has confirmed all four checkpoints. Along the way, it can read requirements typed directly, or pull content from a linked Feishu document if that companion skill is available. It rewrites acceptance criteria as a clear chain of a starting condition, a user action, and an observable result, rather than vague language like "good experience." For AI-based products it adds sections on agent workflows, prompts, tools, and quality checks, for more conventional products it adds sections on permissions, data, interfaces, and error handling. Installing it means cloning the repository and copying its skill folder into your Codex skills directory, then restarting Codex. Trigger phrases are written in Chinese, and the project does not state a license.
A Codex or Claude Code skill that turns messy requirements or ideas into a clear, engineering-ready PRD through four required confirmation steps.
Setup difficulty is rated easy, with roughly 5min to a first successful run.
Mainly pm founder.
This repo across BitVibe Labs
Verify against the repo before relying on details.