---

Bring senior-grade Aurelia v2 expertise into every AI agent session. This package bundles eight router-routed skills — one thin router (aurelia-expert) and seven pillar skills (aurelia-foundation, aurelia-runtime, aurelia-component-library, aurelia-largespa, aurelia-migration, aurelia-plugin, aurelia-ecosystem) — so any compatible agent picks the right pillar on its own and never confuses Aurelia 1 with Aurelia 2.

Works with OpenCode, Claude Code, and any agent that supports the skill tool or can invoke CLI-installed skills via bunx / npx.

Quick start

Three install paths — pick whichever fits your workflow:

# 1. CLI install (any agent) — Bun preferred
bunx aurelia-expert install

# 2. Cross-agent install via skills.sh
npx skills add expert-vision-software/aurelia-expert --skill aurelia-expert -a opencode

# 3. OpenCode auto-install — add to your opencode.json
#    { "plugin": ["aurelia-expert"] }

That's it — skills are available immediately. The CLI defaults to local scope (writes to ./.opencode/skills/). Pass --scope global to install once for all your projects (~/.config/opencode/skills/).

Bundled skills

Skill	Leading word	What it does
aurelia-expert	branch	Router-only. Reads the prompt, classifies it as one of seven branches, hands off to the named pillar.
aurelia-foundation	scaffold	First-decision reference: philosophy, hello-world quickstart, custom-element naming, lifecycle hooks, AI-assisted scaffolding.
aurelia-runtime	resolve	Component wiring: resolve() DI, @route, <au-viewport>, AppTasks, custom events, cross-feature orchestration.
aurelia-component-library	assemble	Styled, variable-driven component library on Aurelia v2 + Tailwind: design-token layer, shared/components/ui/ layout, ui- element prefix, component anatomy, greenfield deploy + migrate-existing with a v1-source gate.
aurelia-largespa	slice	Enterprise SPA structure: pages/, features/, shared/, hierarchical Agents.md, thin-page orchestrators, Model/DTO boundary. Focal point.
aurelia-migration	lift	v1 → v2 lift table (eight removed APIs), AUR0xxx debugging, .to-view performance, DeepWiki citation protocol.
aurelia-plugin	package	Distributable Aurelia v2 plugins: register(container) contract, .customize() options with DI.createInterface tokens, global resource registration, AppTask hooks, npm distribution (peerDependencies, dual ESM/CJS, ?raw templates), and an extract procedure (library / feature-slice / shared-dir → plugin).
aurelia-ecosystem	wire	First-party plugins and runtime ecosystem: @aurelia/fetch-client, @aurelia/validation, @aurelia/dialog, @aurelia/state, @aurelia/i18n, forms (model.bind/matcher.bind/submit.trigger:prevent), Vitest + @aurelia/testing component tests, and SSR/prerendering (aurelia2-ssr, takeover, hydration, sitemap/robots).

All skills load on demand. Only metadata (name + description) sits in every session at startup; the full skill body loads only when the agent decides it's relevant — zero overhead when not in use.

Examples

🧭 Router dispatches by branch

You: "Inject a service into a custom element with constructor parameters."

The aurelia-expert router classifies the prompt as resolve and hands off to aurelia-runtime. The pillar answers in its own voice:

Use resolve(IFoo) as the default value:
constructor(private readonly userService: IUserService = resolve(IUserService)) {}
Tests inject mocks through the constructor.
@inject is gone in v2 — never write it.

🏗️ Scaffold a foundation

You: "Start a new Aurelia 2 app and create a custom element called user-profile."

Routes to aurelia-foundation → quickstart + components references:

1. Scaffold: bunx makes aurelia my-app --template vite-ts
2. Component name: user-profile (kebab-case, contains a hyphen)
3. Files: src/components/user-profile.ts + user-profile.html
4. @customElement({ name: 'user-profile', dependencies: [] })
5. Bindable inputs: @bindable name: string;

🧩 Slice a large SPA

You: "We're starting a customer-portal SPA with billing, support, and reporting features."

Routes to aurelia-largespa → feature-first layout + orchestrator + Agents.md hierarchy:

src/
├── pages/                ← thin navigators (no business logic)
│   └── billing/index.ts
├── features/             ← one folder per business capability
│   ├── billing/
│   │   ├── components/
│   │   ├── services/     ← I-prefixed interfaces; resolve(), never concrete
│   │   ├── models/       ← *.model.ts; Model.fromDTO() at the API boundary
│   │   └── index.ts      ← registers local custom elements + DI bindings
│   └── support/
├── shared/               ← globally registered; no business logic
│   ├── auth/
│   └── http/
└── Agents.md             ← root precedence surface

🔁 Lift a v1 codebase

You: "I'm porting an au-northwind reference app to Aurelia 2."

Routes to aurelia-migration → v1-removals + DeepWiki citations:

The v1 → v2 lift is mechanical. Eight APIs changed.
Run this table once per file (reference/v1-removals.md):

  PLATFORM.moduleName('order-list')   →  no wrapper needed; native bundler import
  @inject(HttpClient, ILogger)        →  constructor(private readonly svc = resolve(HttpClient))
  configureRouter(config)             →  @route([...])
  <router-view>                       →  <au-viewport>
  <compose viewModel="...">           →  <au-compose component.bind="...">
  .delegate on a custom event         →  .trigger (throws AUR0713 at compile time)
  activate() / deactivate()           →  canLoad / loading / canUnload / unloading
  inline style="width: ${value}%"     →  width.style="value + '%'"

Requirements

Component	Notes
Bun ≥ 1.0	Required for the CLI installer, test suite, and OpenCode plugin runtime.
Aurelia 2.x project	Optional. Skills fire on any prompt that mentions Aurelia — no project scaffold needed to browse them.
OpenCode, Claude Code, or any skill-compatible agent	The eight skills surface in the agent's <available_skills> list once installed.

Installation

1. CLI install (any agent)

# Local scope (default): installs to ./.opencode/skills/
bunx aurelia-expert install

# Global scope: installs to ~/.config/opencode/skills/
bunx aurelia-expert install --scope global

# Same commands work with npx
npx aurelia-expert install

# Check / remove
bunx aurelia-expert status
bunx aurelia-expert uninstall --scope local

Non-interactive by design — no prompts, CI-friendly. The CLI writes a .version marker under skills/aurelia-expert/ so subsequent runs are no-ops.

2. Cross-agent install via skills.sh

npx skills add expert-vision-software/aurelia-expert --skill aurelia-expert -a opencode

skills.sh is the cross-agent installer. The command above targets OpenCode; swap -a opencode for claude-code or any other agent it supports.

3. OpenCode plugin auto-install

Add aurelia-expert to your opencode.json plugin array:

{
  "$schema": "https://opencode.ai/config.json",
  "plugin": ["aurelia-expert"]
}

OpenCode installs the package on next session start, then plugin.ts#config() auto-copies the eight skills into .opencode/skills/ (idempotent — checks .version marker).

For local development against a checkout of this repo, reference the directory directly:

{
  "$schema": "https://opencode.ai/config.json",
  "plugin": ["file:///absolute/path/to/aurelia-expert"]
}

Scope choice

We ship the package unscoped (aurelia-expert, not @expert-vision-software/aurelia-expert). Reasons:

• bunx aurelia-expert install works directly — no bunx @org/... syntax to remember.
• npx aurelia-expert install works directly.
• The name is unique enough on the public npm registry.

If a future conflict forces a rename, we'll publish @expert-vision-software/aurelia-expert and document the migration.

Why this plugin?

• v1 vs v2 contamination is impossible — the six forbidden v1 APIs (.delegate → AUR0713, <router-view>, PLATFORM.moduleName, configureRouter, <compose>, activate/deactivate) are pinned as prohibitions across every pillar. @inject is DEPRECATED — still valid, resolve() is preferred.
• Predictable routing — every skill carries a leading-word (branch, scaffold, resolve, assemble, slice, lift, package, wire); the router classifies in one pass and stops.
• Hierarchical Agent instructions — aurelia-largespa teaches how to author Agents.md files at root + directory scope, so project conventions override the skill defaults.
• Markdown-only bundle — pure skill content, no runtime dependencies, zero overhead until the agent decides to load a skill.

Development

# Clone and install
git clone https://github.com/Expert-Vision-Software/aurelia-expert.git
cd aurelia-expert
bun install

# Type-check (no emit — this package is published as .ts sources)
bun run check

# Run the test suite (skills + installer)
bun test

# Smoke-test the CLI against the local checkout
bunx . install --scope local
bunx . status

prepublishOnly runs bun run check && bun test automatically before any npm publish. The CI release workflow (.github/workflows/release.yml) extracts release notes from CHANGELOG.md, creates a GitHub Release, then runs npm publish --provenance --access public (requires the NPM_TOKEN secret and id-token: write permission for provenance).

See CONTRIBUTING.md for the file layout, architecture decisions, and authoring conventions.

Acknowledgments

• OpenCode — plugin architecture, skill loader, config schema
• skills.sh — cross-agent installer
• The Aurelia team — https://docs.aurelia.io is the canonical source of truth for every claim in this package

---