À propos
Cette compétence fournit les connaissances essentielles du domaine du commerce de détail et du e-commerce aux développeurs qui construisent des solutions de vitrine pour les PME. Elle formalise les concepts clés, les écueils courants et les modèles de données requis, tels que Produit-Variantes et le stock sensible aux canaux, pour éviter les implémentations naïves. Utilisez-la lors de la spécification de fonctionnalités impliquant des catalogues, des stocks, des tarifications ou la récupération de paniers.
Installation rapide
Claude Code
Recommandénpx skills add avelikiy/great_cto -a claude-code/plugin add https://github.com/avelikiy/great_ctogit clone https://github.com/avelikiy/great_cto.git ~/.claude/skills/vertical-retailCopiez et collez cette commande dans Claude Code pour installer cette compétence
Documentation
Retail & e-commerce — spec it like someone who's run a store
The SMB retail buyer already pays Shopify ($39–399/mo + 2.9%), BigCommerce, Wix, Ecwid, or WooCommerce. They are not naive — so the spec can't be either. A storefront that "has products and a cart" is table stakes; the value is in the parts those platforms do badly. Read this before writing the catalog/inventory/pricing/cart sections of any retail ARCH or PLAN doc.
1. Domain vocabulary (use these exact words)
- SKU vs variant — a variant is one buyable configuration (Red / Large); its SKU is the unique code that variant ships and is counted under. A "product" is the parent; you stock, price, and sell variants, not products.
- Multichannel / omnichannel — selling across several channels (own storefront, Amazon, eBay, in-store POS, Instagram). Omnichannel additionally means one inventory pool behind all of them. Channel-awareness is the whole game for SMB inventory.
- Reorder point — stock level that triggers a purchase order = (avg daily demand × lead time in days) + safety stock. Lead time = supplier days from order to receipt. Safety stock = buffer for demand/lead-time variance. Reordering without all three is wrong.
- COGS (cost of goods sold) and margin = (price − COGS) / price. Landed cost = unit cost + freight + duties + handling; margin must use landed cost, not invoice cost.
- ATS / available to sell = on-hand − allocated (reserved by open orders) − safety stock. Customers buy against ATS, never raw on-hand.
- Backorder vs preorder — backorder = out of stock now, will refill (sell against incoming PO). Preorder = not released yet, future availability date. Different fulfillment promises.
- Cart abandonment rate = 1 − (completed checkouts / carts created); industry ~70%.
- AOV (average order value) and conversion rate = orders / sessions. The two levers pricing/promotions move.
- Fulfillment — pick/pack/ship. Dropship = supplier ships direct, seller never holds stock (so "stock" is the supplier's ATS feed, not yours).
- MAP (minimum advertised price) — supplier-imposed price floor; a pricing rule must respect it or the seller loses the brand.
2. Non-obvious domain rules
- Shopify owns the storefront — don't fight it head-on. A me-too checkout loses. The wedge is the platforms' weak spots: multichannel inventory + reorder, and cart recovery. Spec the storefront as competent-and-owned, and put the differentiation in the other three.
- Variants explode combinatorially. options (Size × Color × Material) multiply: 5×8×3 = 120 variants per product. The data model, UI, and import flow must assume hundreds of variants per product, each with its own SKU / price / stock — not a flat product list.
- Inventory must be channel-aware. The same SKU is sold on storefront + Amazon + POS; stock must decrement across all and sync back, or you oversell. Single-channel inventory is the most common naive failure and the strongest wedge.
- Pricing rules interact with floors. A promotion or demand-based rule must clamp to a margin floor and MAP. A rule that can price below landed-cost margin is a bug, not a discount.
3. What a naive build gets wrong
- Products without a variant model — a flat
product { price, stock }table. Breaks the instant the seller stocks two sizes. Variants are core, not an add-on. - Single-channel inventory — stock that lives only in the storefront, no sync across Shopify / Amazon / POS. Guarantees overselling for any real SMB.
- Reorder without lead-time / safety-stock — "reorder when stock < 10" stocks out during the supplier lead time. Must use reorder-point math.
- Cart recovery that ignores suppression / consent — emailing/SMSing without consent, or after unsubscribe/purchase, is illegal (CAN-SPAM / TCPA / GDPR) and burns deliverability. Honor suppression + quiet hours.
- Pricing that ignores the margin floor — a promo engine that can sell below cost, or below MAP.
4. Must-model entities
| Entity | Key fields |
|---|---|
| Product | id, title, option axes (e.g. Size, Color) — the parent |
| Variant | product_id, option values (Red/L), SKU, price, COGS/landed cost — one per option combo |
| InventoryLevel | variant_id, channel/location, on_hand, allocated, safety_stock → derive ATS |
| ReorderRule | variant_id, reorder_point, reorder_qty, lead_time_days, supplier |
| PricingRule | scope (variant/collection), trigger (demand/margin/schedule), action, margin_floor, MAP |
| AbandonedCart | cart_id, customer, line items, value, abandoned_at, recovery state, consent/suppression |
The Variant option matrix and the channel-keyed InventoryLevel are the two that naive specs collapse — keep them explicit.
5. Per-product notes (wedge + the one domain thing)
- storefront (content) — catalog, checkout, themes; a store the seller owns. Wedge: owned channel + SEO (it must rank — see [[local-seo]]). The one thing: the Product→Variant model and clean indexable URLs. Don't out-engineer Shopify's checkout; match it and move on.
- inventory (crud) — track stock across channels, auto-reorder before stockout. This is the underserved-by-Shopify wedge. The one thing: channel-aware InventoryLevel + reorder-point math (lead time + safety stock). Get this right and the product justifies itself.
- pricing (dashboard) — rules-based pricing + promotions reacting to demand/margin. Wedge: margin-aware automation SMBs do by hand. The one thing: every rule clamps to margin floor + MAP.
- cart-recovery (crm) — win back abandoned carts via timed email/SMS. Wedge: recovering the ~70% that abandon. The one thing: consent + suppression + timing — defer the messaging mechanics to [[lifecycle-messaging]].
6. Compliance (light — defer the heavy parts)
- Sales tax nexus — economic nexus thresholds vary by US state (post-Wayfair); the seller may owe tax in states they've never shipped to. Note it in the spec; defer the actual calc/filing to billing. Don't hand-roll tax.
- Email / SMS consent — cart recovery needs prior consent (CAN-SPAM / TCPA / GDPR), honored unsubscribe, and quiet-hours/suppression. Defer the delivery + consent machinery to [[lifecycle-messaging]]; the spec just states the requirement.
- PCI — checkout uses Stripe-hosted elements so card data never touches our servers (SAQ-A scope). State that intent; defer the scope proof to pci-reviewer.
Output
When applied, contribute a Retail domain section to the ARCH/PLAN/DESIGN doc:
## Retail domain
- model: Product→Variant (option matrix, per-variant SKU/price/stock) · channel-aware InventoryLevel (ATS = on_hand − allocated − safety_stock)
- reorder: reorder_point = avg_demand × lead_time + safety_stock (not "< N")
- pricing: every rule clamps to margin_floor + MAP (margin on landed cost)
- cart-recovery: consent + suppression + timing → [[lifecycle-messaging]]
- wedge: multichannel inventory + reorder, cart recovery (don't fight Shopify's storefront/checkout)
- compliance: tax nexus → billing · consent → [[lifecycle-messaging]] · PCI Stripe-hosted (SAQ-A) → pci-reviewer
- migration: catalog/variant/stock import path → [[migration-ready-schema]]
Dépôt GitHub
Frequently asked questions
What is the vertical-retail skill?
vertical-retail is a Claude Skill by avelikiy. Skills package instructions and resources that Claude loads on demand, so Claude can perform vertical-retail-related tasks without extra prompting.
How do I install vertical-retail?
Use the install commands on this page: add vertical-retail to Claude Code as a plugin, or clone its repository into your skills directory, then restart Claude so it picks up the skill.
What category does vertical-retail belong to?
vertical-retail is in the Meta category, tagged ai and design.
Is vertical-retail free to use?
Yes. vertical-retail is listed on AIMCP and free to install. It runs inside Claude, so no separate service account is required to use the skill itself.
Compétences associées
Cette compétence propose une configuration éprouvée en production pour Content Collections, un outil axé sur TypeScript qui transforme des fichiers Markdown/MDX en collections de données typées de manière sûre avec une validation Zod. Utilisez-la lors de la création de blogs, de sites de documentation ou d'applications Vite + React riches en contenu pour garantir la sécurité de typage et la validation automatique du contenu. Elle couvre tout, de la configuration du plugin Vite et de la compilation MDX à l'optimisation des déploiements et la validation des schémas.
Cette compétence permet aux développeurs de créer des applications avec la plateforme de marchés prédictifs Polymarket, incluant l'intégration d'API pour le trading et les données de marché. Elle fournit également une diffusion de données en temps réel via WebSocket pour surveiller les transactions en direct et l'activité du marché. Utilisez-la pour mettre en œuvre des stratégies de trading ou pour créer des outils traitant les mises à jour de marché en direct.
Cette compétence aide les développeurs à créer des plugins OpenCode qui s'interconnectent avec plus de 25 types d'événements tels que les commandes, les fichiers et les opérations LSP. Elle fournit la structure du plugin, les spécifications de l'API événementielle et les modèles d'implémentation pour les modules JavaScript/TypeScript. Utilisez-la lorsque vous avez besoin d'intercepter, de surveiller ou d'étendre le cycle de vie de l'assistant IA OpenCode avec une logique personnalisée pilotée par les événements.
SGLang est un framework de service LLM haute performance spécialisé dans la génération rapide et structurée pour les workflows JSON, regex et agentiques grâce à son cache de préfixe RadixAttention. Il offre une inférence nettement plus rapide, particulièrement pour les tâches avec des préfixes répétés, ce qui le rend idéal pour les sorties complexes et structurées ainsi que les conversations multi-tours. Choisissez SGLang plutôt que des alternatives comme vLLM lorsque vous avez besoin d'un décodage contraint ou que vous construisez des applications avec un partage étendu de préfixes.
