Replies: 6 comments 5 replies
-
|
Thank you for the RFC @onmax I think adding more AI native feature within Nuxt + the ecosystem is a must have from nowadays. Though, I have troubles thinking of a I would start first with the module metadata: export default defineNuxtConfig({
modules: ['npm-agentskills/nuxt'],
agents: {
skills: {
targets: ['claude', 'cursor']
},
mcp: 'https://nuxt.com/mcp',
llms: 'https://nuxt.com/llms.txt'
}
}) |
Beta Was this translation helpful? Give feedback.
-
I like this idea. I'm curious what the expected skill content is for these libraries? Is it a way to provide high-level documentation to the agent for how to use the library effectively? |
Beta Was this translation helpful? Give feedback.
-
|
I think the RFC from Cloudlfare is a much better solution for discoverability and distribution |
Beta Was this translation helpful? Give feedback.
-
|
I have created https://nuxt-skill.onmax.me/ - a new strategy to work with skills. I believe this strategy adheres better to the Nuxt philosophy: simple, composable, and great DX! Check it out :) |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
-
RFC: Bundling Agent Skills in Nuxt Modules
Note
Update: I have created https://nuxt-skill.onmax.me/ - a new strategy to work with skills. I believe this strategy adheres better to the Nuxt philosophy: simple, composable, and great DX! Check it out :)
Abstract
We propose a standard method for Nuxt modules to distribute "Agent Skills" (AI coding assistant instructions) alongside their source code. By adding an
agentsfield topackage.json, modules can define skills that are automatically discovered, version-locked, and exported to the user's AI agent configuration (Claude, Cursor, Copilot, etc.).The Problem
Currently, agent skills for Nuxt are distributed separately from the modules they support (e.g., in a centralized
nuxt-skillsrepository). This causes several issues:The Solution
We propose bundling skills directly inside the npm package. When a user installs a Nuxt module, they get the corresponding skills automatically.
1. The
package.jsonConventionModule authors define skills in their
package.jsonusing theagentsproperty. This namespace allows for defining skills now and other agent capabilities later.{ "name": "@nuxthub/core", "version": "0.10.4", "agents": { "skills": [ { "name": "nuxthub", "path": "./skills/nuxthub" } ] }, "files": ["dist", "skills"] }2. Directory Structure
Skills follow the agentskills.io format:
Warning
The
agentspackage.json field is a proposed convention, not yet part of the official agentskills.io specification. We've proposed it to the broader ecosystem in agentskills/agentskills#81.Implementation
We have created a reference implementation:
npm-agentskills.Usage in Nuxt (Automatic)
Add the module to your config. Skills are discovered and exported on
nuxi prepareornuxi dev. We use theagentskey to configure the module:Usage via CLI (Non-Nuxt)
The CLI binary has been renamed to
agentsfor brevity:npx agents export --target claude npx agents listFuture Extensibility
Following feedback from atinux (Nuxt creator), we have renamed the configuration key from
agentskillstoagents. This ensures the ecosystem remains open to future AI implementations, such as the Model Context Protocol (MCP) orllms.txtstandards.While this RFC focuses on skills, the
agentsnamespace allows the configuration to evolve without breaking changes. A future configuration might look like this:Supported Agents
All exports are project-local:
.claude/skills/.github/skills/.cursor/skills/.codex/skills/.opencode/skill/.agents/skills/.goose/skills/Real-World Example
See the implementation in
@onmax/nuxt-better-auth:Next Steps
References
npm-agentskills- Reference implementationBeta Was this translation helpful? Give feedback.
All reactions