400 SKUs.14% return rate.A creative agency that takes five days to brief.
Generic AI writers treat every variant as a fresh description and produce twelve versions of the same paragraph for a t-shirt that comes in twelve colors. Platform tools (Shopify Magic, Amazon’s built-in AI) are locked to their platform. Subscription tools charge $49-99 a month for narrower output.
The Ecommerce Skills Pack is six Claude skills built for the operators who actually run catalogs. Variant-aware. Platform-specific. Composable. One-time $89.
Ecommerce is the most AI-curious vertical and the most poorly served by it.
The workflows that run a catalog are messy in a specific way. Variants explode SKU counts. Returns hide margin problems behind customer-service noise. Supplier emails get answered by people who don’t care about your terms. Inventory is too important to model on Sunday night but never gets touched the rest of the week. Creative briefs sit in agency Slack threads for five days. Reviews accumulate faster than anyone can respond.
The generic AI tools weren’t built for this. They produce 12 versions of the same paragraph for a 12-variant SKU. They treat returns as text classification with no margin context. They write supplier emails that sound like cover letters. They do not compose with each other. The Ecommerce Skills Pack was built for operators by operators, on the June 2026 Anthropic Agent Skills format, with composition between skills as a first-class feature.
Average catalog size at which generic AI writers stop being useful and start producing variant copy that reads identical to every other description in your store.
Industry-average return rate eating margin. The return-reason classifier turns the noise into a prioritized fix list — and the supplier negotiation skill turns the fix list into leverage.
Versus $49-99 monthly recurring for Copy.ai, Jasper, or Helium 10. The pack pays for itself the first week and compounds from there.
Clear about the lane. No inflated promises.
- Six Claude skills built on the May 2026 Agent Skills open standard.
- Platform-specific where it matters — Amazon, Shopify, Walmart, Etsy, TikTok Shop.
- Variant-aware product copy that treats a 12-variant SKU as one variant family, not 12 listings.
- Composable — the description writer feeds the ad creative brief generator with shared brand-voice tokens.
- Agency-licensed — use across unlimited client stores under your engagement.
- Reviewed by ecommerce operators running real catalogs ($1M-$50M GMV).
- A copywriter replacement. The skills draft; you approve before publish.
- An auto-publishing tool. Every output is reviewed before listing.
- A platform integration. The skills produce copy; you paste it into Shopify/Amazon/etc.
- A subscription. One-time $89 with 12 months of template updates.
- Legal or platform-policy advice. Amazon's AI content policy is yours to read.
- Industry-locked. Defaults work for apparel, electronics, beauty, home, and CPG.
Six skills. Six revenue levers.
Each skill follows the June 2026 Anthropic Agent Skills structure: YAML frontmatter for trigger detection, body for the workflow, and reference files loaded on demand. All six compose with each other.
Platform-aware structure (Amazon 200-char title + 5 bullets; Shopify SEO-rich long-form; Walmart compliance-tight; Etsy narrative). Variant-family logic: shared copy stays shared, variant-specific copy varies, brand voice locked across the family. Built-in factuality check.
Sorts return reasons into actionable buckets (sizing, quality, expectation mismatch, fulfillment, fit-for-purpose). Surfaces the SKU-and-reason combinations driving most margin loss. Outputs a prioritized fix list with listing-copy edits, sizing-chart suggestions, and supplier escalation flags.
BATNA framing, leverage discovery, tone-calibrated outreach (firm but professional). Templates for price reductions, payment-term improvements, lead-time compression, defect-rate corrections, and exclusivity asks. Multi-turn negotiation tracker.
Pareto classification by revenue and units. A-class SKUs (top 80% of revenue), B-class (15%), C-class (5%) with reorder-priority logic. Identifies dead stock, fast-movers needing safety stock, and seasonal anomalies. Outputs a weekly action list — not a static spreadsheet.
Platform-specific briefs for Meta, TikTok, Google Shopping, and Amazon Sponsored Products. Pulls hook angles from product reviews and return-reason data when available. Outputs creator-ready specs: hook, beat sheet, on-screen text, CTA variants, and shot list.
Sorts reviews into respond/escalate/ignore buckets. Drafts platform-compliant responses for 1-5 star reviews with brand voice locked. Flags reviews that reveal listing-copy gaps, sizing issues, or product defects — feeding insights back to the description writer and supplier negotiation skills.
What a real ecom skill actually looks like.
Below is the abridged ecommerce-product-description-writer/SKILL.md file. June 2026 Anthropic format, deployable as shown. The full version in the pack adds the platform-conventions reference, the variant-family logic spec, the brand-voice extraction patterns, and 6 worked examples across apparel, electronics, beauty, and home.
---
name: ecommerce-product-description-writer
description: Use when an ecommerce operator needs to write or rewrite product descriptions for Shopify, Amazon, Walmart, Etsy, or TikTok Shop — especially for products with multiple variants (color, size, material, style). Triggers on phrases like “write product copy,” “Amazon listing,” “Shopify description,” “PDP copy,” “describe this variant,” “bulk descriptions for these SKUs.” Handles platform-specific length and bullet conventions, variant-family logic (shared vs variant-specific copy), brand-voice consistency, SEO/AEO keyword placement, and built-in factuality check.
---
When an operator asks for product description help:
1. Confirm platform. Amazon, Shopify, Walmart, Etsy, or TikTok Shop? Each has its own structure mode.
2. Detect variant family. If multiple SKUs share a parent product, treat as ONE variant family (not N separate listings).
3. Read references/platform-conventions.md for length, bullet structure, and required attributes.
4. Read references/variant-family-logic.md BEFORE writing. Variant rules are non-negotiable.
5. Anchor brand voice. Ask for 2-3 existing high-performing descriptions; extract tone + word patterns.
6. Draft shared copy first, then variant-specific differentiators. Never duplicate boilerplate across variants.
7. Run factuality check: every product claim must be supported by inputs or marked as unverified.
NEVER:
- Write 12 separate descriptions for a 12-variant SKU.
- Make claims about size, fit, materials, or compatibility without source data.
- Use hype words generic AI defaults to (“premium quality,” “best in class,” “perfect for everyone”).
- Exceed Amazon’s 200-char title limit or Shopify’s effective 2,500-char body.
ALWAYS:
- Lead variants with the SPECIFIC differentiator (color name, size, material) in the first 6 words.
- Place primary keyword in title, first bullet, and first 100 chars of body.
- Flag any inferred claim with “[verify]” in the draft so the operator catches it before publish.
- Output the variant family as a table (variant ID, title, bullets, body) — not a wall of text.The three-tier architecture, in 90 seconds.
You don’t need to memorize this — but knowing it makes customizing the pack faster.
YAML frontmatter (name + description). Claude reads this to decide whether to invoke your skill. The description is the trigger detector — specific wins, vague loses.
The workflow itself: numbered steps, NEVER/ALWAYS guardrails, and pointers to deeper detail in reference files.
Detailed knowledge loaded only when the skill body tells Claude to read it. Keeps Claude's working context lean while supporting deep domain expertise.
The Agent Skills format is now an open standard supported by Claude Code, Claude Pro projects, OpenAI Codex CLI, Cursor, Gemini CLI, and GitHub Copilot. Same SKILL.md files work everywhere; Claude-specific features (hooks, allowed-tools) are safely ignored by other agents.
The unlock: skills feed each other.
A generic AI tool is a one-shot. A skill is a node in a workflow graph. Three composition flows ship pre-configured; you can add your own.
Turn a new SKU drop into platform-ready descriptions and aligned ad creative briefs in one sitting.
Writes the variant family for the new SKU — platform-specific structures, brand voice locked
Auto-pulls brand voice tokens and product differentiators into Meta / TikTok / Google Shopping briefs
Stop letting returns eat margin invisibly. Turn return data into supplier leverage.
Sorts a month of returns into actionable buckets, flags SKU-and-reason combinations driving most loss
Drafts a defect-rate-correction email with the return data as leverage — specific SKUs, specific rates, specific asks
Reviews are a free source of customer truth. Most operators leave it on the floor.
Triages reviews and surfaces patterns — sizing complaints, copy gaps, feature requests, repeat defects
Updates affected PDPs with corrected sizing language, addresses inferred-then-disappointed claims, sharpens variant-specific copy
The integrity moat.
Exactly what you get for $89, and what you don’t.
- Six production Claude skills with full reference files.
- Three pre-built composition flows.
- Platform support: Amazon, Shopify, Walmart, Etsy, TikTok Shop.
- 12 worked examples across apparel, electronics, beauty, home, CPG.
- Agency license — unlimited client stores under your engagement.
- 12 months of template updates as platforms and the Skills standard evolve.
- Auto-publishing to your store. Every output is reviewed before listing.
- Direct API integrations to Shopify/Amazon — those are MCP-server territory.
- Legal or platform-policy interpretation. You're responsible for compliance.
- Image or video generation. Skills produce copy and structure, not creative.
- Replacement for an ecommerce manager. Skills structure work; you decide.
- Industry-specific skill packs for niche verticals — quotable separately.
Running a one-person company instead of a catalog? The Solo Founder Skills Pack ($79) is the equivalent set for investor updates, board memos, runway modeling, and weekly cadence. Same architecture, different audience. Most buyers need one or the other; multi-hat operators who need both can reply to the welcome email for a bundle discount.
The questions ecom operators actually ask before deploying their first skills.
Both, plus Walmart, Etsy, eBay, and TikTok Shop. The product description writer ships with platform-aware structure modes (Amazon's 200-character title plus 5-bullet pattern, Shopify's longer SEO-rich format, Walmart's compliance-tight format, Etsy's narrative voice). The ad creative brief generator covers Meta, TikTok, Google Shopping, and Amazon Sponsored Products. The other four skills are platform-agnostic. If you sell on a channel not listed, the customization guide walks through adding it in about 15 minutes.
Two differences that matter. (1) Variants. Generic AI writers (Copy.ai, Jasper, etc.) treat every variant as a separate description and produce 12 versions of the same paragraph for a t-shirt that comes in 12 colors. The product description writer in this pack understands which copy is shared, which varies by variant, and how to keep the SEO/AEO structure intact across the variant family. (2) Composition. The skills feed each other — the description writer can pass brand-voice tokens to the ad creative brief generator, and the return-reason classifier can feed insights to the supplier negotiation skill. Platform-built-in AI (Shopify Magic, Amazon's tools) is locked to its platform and doesn't compose across your stack.
Claude Pro or Claude Team is the minimum — the pack works in Claude Pro projects. Claude Code is the fastest auto-trigger experience if you live in the terminal. The Agent Skills format the pack uses is an open standard, so the same skills also work in OpenAI Codex CLI, Cursor, Gemini CLI, and GitHub Copilot — though Claude-specific advanced features may not transfer. The deployment guide covers all five environments.
Yes. The license covers unlimited stores under your control or agency engagement. The customization guide includes a per-client setup pattern that takes about 10 minutes per store: clone the skill folder, edit the brand-voice extraction file, set the platform mix, point at the client's KPI dictionary. Agencies running 10+ stores typically save 15-25 hours per week once the per-client setup is done across the book.
Each skill ships with a brand-voice anchor step. You feed it three things: 2-3 existing high-performing product descriptions, your brand do/don't word lists, and your tone targets (technical/casual/luxe/playful). The skill extracts the structural and lexical patterns and locks them across every output in the session. Most operators tell us the descriptions are indistinguishable from their existing best ones within the first 5-10 outputs as the anchor sharpens.
Amazon's published policy as of 2026 permits AI-assisted listing content provided the operator is responsible for the final accuracy of every claim, the content is original (not copied from another listing), and any product attributes match the actual product. The pack's skills are tools you operate; you remain responsible for review and accuracy. The product description writer includes a built-in factuality check step that flags any claim it can't verify from your inputs. We don't recommend wholesale auto-publishing — every output should be reviewed before listing.
Different audience, different skill set, same architectural pattern. The Solo Founder Pack ($79) is for the founder running the whole company through Claude — investor updates, board memos, runway modeling. The Ecommerce Skills Pack ($89) is for the operator running a catalog — descriptions, returns, suppliers, inventory, ads, reviews. Most buyers need one or the other. Some agencies and multi-hat operators buy both; if that's you, reply to the welcome email and we'll bundle a discount.
30-day no-questions refund. Install the pack, run one skill on a real catalog task (rewrite 5 product descriptions, classify a week of return reasons, draft a supplier negotiation email, whatever fits your week), and if it doesn't save you at least 2 hours in that first session, email RedHub AI support and we refund. Refunds across the catalog are countable on one hand to date — when you're the operator drowning in SKUs and returns, the time savings are self-evident the first run.
Stop drafting descriptions by hand.
Stop modeling inventory on Sundays.
Six Claude skills, twelve worked examples, three composition flows, deployment guides for five environments. Twenty minutes to install, twenty hours back in the first month. Operator-built. Composable. Yours forever.
Sold by RedHub AI LLC · Secured by Stripe · redhub.ai