AI Coding Agent Skills - Commerce Engine Docs

The commercengine/skills repository teaches AI coding agents the same implementation contracts documented here. The ce-seo and ce-ai skills encode deployment decisions, framework mounting, bundle boundaries, cart semantics, and failure modes that are easy to miss from API signatures alone. These are coding-agent skills. They are different from @commercengine/ai, which exposes runtime WebMCP tools to agents visiting the storefront.

Install the skills

Agent Skills CLI

npx skills add commercengine/skills

Claude Code plugin marketplace

/plugin marketplace add commercengine/skills
/plugin install ce@commercengine/skills

Manual Claude Code install

git clone https://github.com/commercengine/skills ~/.claude/skills/commercengine

The repository follows the Agent Skills format and can also be consumed by tools that support compatible skill/plugin manifests.

`ce-seo`

Use ce-seo for metadata, Product/ProductGroup JSON-LD, canonical URLs, robots, XML sitemaps, llms.txt, Markdown mirrors, preview noindex policy, deployment-mode selection, and CMS-owned storefront routes. The skill starts with the key decision:

Does a server receive requests after deployment?
├─ yes → mount one SEO request adapter
└─ no  → run one SEO asset prebuild

For server mode, the skill enables robots: true and sitemap: true explicitly because both request-handler options default to off. It also distinguishes the package route default (/products/:slug) from the starter convention (productBase: "/product") and treats indexable as an override on top of deployment detection, not a required production flag. Example prompts:

Use ce-seo to replace the hand-written SEO in this Next.js storefront with @commercengine/seo.

Use ce-seo to add JSON-LD, llms.txt, sitemaps, and Markdown mirrors to this Astro static site.

`ce-ai`

Use ce-ai for WebMCP registration, catalog/cart tools, Hosted Checkout bridges, navigation, session capability reporting, diagnostics, policy/FAQ tools, and custom AI modules. It preserves the runtime safety boundary:

Agent may:   search → inspect → configure cart → open checkout
Human does:  credentials → address/payment → final purchase

Example prompts:

Use ce-ai to let browser agents search products and add items to the real cart in this Next.js app.

Use ce-ai to review our custom agent tools for accidental customer-data or payment exposure.

Start with `/ce` when unsure

The Commerce Engine router can combine the specialist skills for a broader request:

/ce make this storefront production-ready for search engines and browser agents

The router knows about ce-setup, ce-catalog, ce-cart-checkout, ce-seo, ce-ai, and the framework skills.

What skills add beyond API docs

Topic	API surface alone	Skill guidance
SEO serving	exposes handlers	choose server middleware vs static prebuild first
client bundle	types compile	keep SEO instance server-bound; share only pure config
static Markdown	generator exists	do not mount `.md` catch-all routes in static apps
breadcrumbs	JSON-LD helper exists	resolve URLs through SEO route resolvers
AI registration	function exists	mount early, diagnose registration, abort on cleanup
cart quantity	mutation method exists	distinguish add vs absolute set; honor promo/free lines
WebMCP safety	tool schema exists	never add address/payment/order-placement tools

Starter projects as executable reference

The skills intentionally point agents back to tark-ai/ce-starter-projects. All 15 starters use the published packages across React, Next.js, TanStack Start, Astro, and SvelteKit. Use the starters when an agent needs to answer questions such as:

Where should this code live in this framework?
Is this module safe to import into the browser?
How should generic social tags yield to page-specific metadata?
When does WebMCP registration mount and clean up?
Which directory receives generated static SEO files?

Keep docs, skills, and starters aligned

For Commerce Engine maintainers, these surfaces have different roles:

Package source and tests define behavior.
Starter projects demonstrate complete application wiring.
Agent skills encode decision rules and pitfalls for coding agents.
These docs explain the supported public workflow to developers.

When behavior changes, update all four together.

AI-assisted development

General Commerce Engine coding-agent resources.

Skills repository

Browse the complete skill source.

​Install the skills

​Agent Skills CLI

​Claude Code plugin marketplace

​Manual Claude Code install

​ce-seo

​ce-ai

​Start with /ce when unsure

​What skills add beyond API docs

​Starter projects as executable reference

​Keep docs, skills, and starters aligned