The sale that ended, except on 31 sizes

One prompt to paste into Claude, ChatGPT or any AI agent that can build. It checks the middle square before anything else, builds the map square by square, carries the 5 rules above, and has the whole build: what starts it, every step, the data, the review, the alerts, what to do when it breaks, and what to count before you switch it on.

Help me build this in my business: Keeping every product on the brand's online store right, every day. That means five things. Putting each new product up: its photos, its description, every size, the price, the tags and collections it sits in, and the fields search engines read. Running each launch, price change and promotion on time, and taking it down on time. Keeping the product data the same in the brand's master list, on the store and on any marketplace. Checking the site every day for broken links, missing photos, sizes shown in stock that are not, and wrong prices. And a weekly report on what sold, what did not, and what shoppers searched for.

Before you build anything: this runs on one person's judgment, "What to feature, push or mark down, and how the brand sounds", held by the ecommerce merchandiser, with the co-founder and creative director. Ask me whether each of these is true in my business, one at a time:
1. One product master holds every style's sizes, prices and materials.
2. Every price change and promotion is approved by a named person before it runs.
3. Someone decides each week what the home page leads with.
If any is not true, stop and tell me which, and what I need in place first. Do not build around it.

Rules from others who built this and got it wrong. Keep every one:
- After every promotion ends, compare every size in it, including any added or edited during it, with the price list, and list any that did not come back.
- Check prices size by size, never product by product.
- Write a material, heel height or fit only from the product master; leave anything missing blank and flag it.
- Save every description as a draft for a person to publish, and check every draft against the brand's list of words it never uses.
- Save every size's price before any change and log every change with its old value, so any change can be put back.

The map of this system, a 3x3:
- The middle square, the judgment it all runs on: What to feature, push or mark down, and how the brand sounds (The ecommerce merchandiser, with the co-founder and creative director). It stays with a person.
- The work, in the order it moves: 1. Keep product data matching everywhere (runs on its own: The master, the store and any marketplace, every difference listed); 2. Draft each new product (runs on its own: From the master and photos, saved unpublished); 3. Check every price each morning (runs on its own: Every size against the price list, and every page and photo); 4. Run approved price changes (runs on its own: At the time set, then every size checked when it ends); 5. Approve launches and markdowns (a person does it: The merchandiser publishes, the co-founder signs off).
- Along the bottom, what it runs on: Automation (The five triggers, the saved prices and the off switch.); Measure (Wrong prices caught before a shopper, drafts rewritten, hours a week, sell-through.); Build (The product master, the price list, the brand's style notes and the words it never uses.).
Build the squares in that order, and keep the person's squares with a person.

The working set to build from:
- Trigger: A new or changed row in the product master; a price change or promotion approved with its start and end time; a daily check at 6:00 am; a check 5 minutes after any promotion ends; the weekly report at 8:00 am on Mondays.
- Steps: On a new master row: read the style, sizes, prices, materials and the photo folder; draft the product with every size, price, tag, collection, search title and description, and photo descriptions; write the description from the master and the brand's style notes only; leave any fact not in the master blank and flag it; save as a draft; send it to the merchandiser. On an approved price change: save every affected size's current price and "was" price; apply the change at the start time; at the end time put back the saved prices; 5 minutes later compare every size in the change, including any added during it, with the price list, and list any that differ. On the daily check: read every size's live price, "was" price, stock and published state; compare with the price list and the stock count; flag any "was" price below the price, any price that differs from the list, any size shown in stock with none or sold out with stock; open every product page and collection and flag missing photos and broken links; compare the master with the store and any marketplace and list differences; write one morning list, wrong prices first. On approval of a fix: set the price, check it again, log the old and new values and who approved. On Monday: sell-through by style and size, conversion, top sellers, slow sizes with stock, and searches with no results.
- Data: One record per size: style, colorway, size, price in the master, live price, "was" price, stock, published or draft, photos count, last change with its old value, new value, time and who approved, and what the check found that morning.
- Review step: The merchandiser reads the morning list and approves each fix back to the set price, and reads and publishes every draft. The co-founder approves every price change and promotion before it is scheduled, and reads the Monday report.
- Alerts: Any size whose live price differs from the price list; a "was" price below the price; a promotion with sizes not back 10 minutes after it ends; a published product with no photo or a broken link; a size shown in stock with none; the master and a marketplace disagreeing; the morning check not running; and the wrong-but-green count, the drafts and prices the merchandiser corrected after they passed every check.
- If it breaks: What you see: no morning list, a list of hundreds of sizes that were fine yesterday, or a draft with facts the master does not hold. Check first: the store connection is still allowed; the price list has not been replaced with an older copy; the master's columns have not moved. Safe fix: turn the step that applies changes off; the check keeps listing, and every fix is made by hand.
- Baseline to count before switch-on: One timed week by hand: minutes for each new product, each price change and each daily check; how many wrong prices the daily check found, and how many a shopper or a colleague found first; and how many hours a promotion's prices took to come back.
- Tools by name: The workflow runner n8n (Starter, 20 euros a month billed annually); the model Claude Sonnet 5.5 ($2 per million tokens in, $10 out); the Shopify Admin API for products, prices, stock and drafts; the product master wherever it is kept (not public); the marketplaces, not named in the posting. Inside the software: Shopify Flow (free on every plan) and Shopify's built-in assistant (one description at a time, for review). Standalone: Matrixify (Basic, $20 per 30 days). People: Crew27's Shopify product listing assistant (from $699 a month).

It may never, on its own: Publish a product; change a price nobody approved; write a material, heel height or fit not in the master; move the home page or a collection; or send an email.
Keep with a person: How the brand sounds, starting with the first lines of every description; what leads the home page and the order of each collection; every price, markdown and promotion; the go-ahead to publish each new product and each launch; which best sellers to reorder; and the email campaigns.
Not yet: Publishing a product on its own; changing any price no person approved; writing a material, a heel height or a fit that is not in the product master; rearranging the home page or a collection; and sending any email.

Ask me for the tools and numbers you need from my business before you start. Then build it one step at a time, and show me each step working before the next.