[Proposal] Official "AI System Prompt" / Skills Standard for Vue ecosystem

[asdlem]

Description

I am proposing that the Vue team publishes an official "AI Skills" / System Prompt definition file (e.g., ai-best-practices.md or similar).

Context & Motivation

Recently, Vercel released an official repository agent-skills, which defines best practices for AI Agents (like ChatGPT, Claude, Cursor) when writing React/Next.js code.

Currently, when developers ask AI to write Vue code, the results are often inconsistent:

1. Legacy Mix: AI often mixes Vue 2 Options API with Vue 3 Composition API.
2. Outdated Syntax: It may not strictly use <script setup> or standard TypeScript patterns unless heavily prompted.
3. Lack of Standardization: Every developer has to write their own "System Prompt" to correct the AI.

The Proposal

I suggest adding a dedicated section or a raw Markdown file in the official documentation repo that serves as a "Single Source of Truth" for AI Context.

This file should explicitly state guidelines for LLMs, such as:

• "Always use Composition API with <script setup lang='ts'>."
• "Use ref over reactive by default (if that is the recommended stance)."
• "Prefer Pinia for state management."
• "Use standard naming conventions..."

Benefits

• Vendor Integration: Tools like Cursor, Copilot, or V0 can pull this official definition to ensure they generate idiomatic Vue code.
• Community: Developers can simply point their AI agents to this official URL to get the best results.
• Modernization: It forces the AI ecosystem to align with the latest Vue 3.x standards.

Reference (React example): Vercel React Best Practices Skill

[skirtles-code]

While I'm not very familiar with this myself, there is ongoing work on AI support at:

• https://github.com/vuejs-ai/skills

This is currently a community project, but Evan has suggested it may become official once it's ready:

... we want to wait until it’s a bit more mature and tested before declaring it as official.

(Quote source: https://discord.com/channels/325477692906536972/1469571233011208285)

For anyone interested in seeing this progress faster, I'm sure they'd appreciate some help testing it out!