Shopify Sidekick Skills: Worked Recipes for Repeat Store Tasks

Save a Sidekick skill for any admin job you do at least weekly, with the same steps and the same output every time. Five are worth saving first: a weekly report, a product audit, a tag clean-up, a reorder list and a set of support macros. Each is below, with instructions you can paste.

A skill is a saved set of instructions that Sidekick follows when you select it or type its shortcut. Think of it as a prompt you stopped retyping. The value is consistency: the report has the same sections each Monday, the audit checks the same ten things, and a new team member gets the same result as you. What follows is the setup, then each recipe with sample output, the thing to check, and where it tends to go wrong.

Follow me on Instagram@sagnikteaches

How skills work, from Shopify's own help pages

Sidekick is Shopify's assistant inside the admin. Shopify's help pages say it works with your store data and shows you changes to review before applying them, and that it can plan and run multi-step tasks in the background, build Shopify Flow workflows, run code to compare or summarise store data (in some stores it cannot run code), edit orders and create content.

Connect on LinkedInSagnik Bhattacharya

To create a skill, click the plus icon in the Sidekick field, choose Skills, then New Skill. Sidekick asks questions about the task and saves the skill once it has enough detail. You can also type a slash and a shortcut in the chat and pick the option to create a skill from it, or fill in the Create a new skill dialog with a shortcut and instructions, then Save.

Subscribe on YouTube@sagnikteaches

To run one, open Skills from the plus icon, select it, add any details and send, or type the slash and shortcut. Sidekick can also apply a skill when your request matches it, though in some stores skills run only when you select them or type the shortcut. Plan on using the shortcut every time, so your routine does not depend on Sidekick guessing. The limit is 50 skills, or 25 in some stores.

The structure that makes a skill reliable

Whether you answer Sidekick's questions or type instructions into the dialog, aim for five parts:

1. Goal: one sentence on what the skill produces and who reads it.
2. Inputs: the date range, collection or tag it should use, or what to ask you for.
3. Steps: what to look at, in order.
4. Output format: headings, a table or a list, with a length limit.
5. Do not: the guard rails. Never change anything without listing the changes and asking first. Never guess a number that is missing; say it is missing.

The fifth part is what turns a clever prompt into a safe routine. Skills are persistent. Whatever you forget to forbid today will be repeated every week.

Which tasks deserve a skill

Not every prompt does. A simple test: would you be annoyed if a colleague asked you to do it differently each time? If yes, save it. Three more conditions help. You do it at least weekly. The inputs come from Shopify data Sidekick can see. And you will read and approve the result, not accept it blind.

Shortcut names are yours to choose. Pick short lowercase words you will remember, and keep a list. If Sidekick tells you a shortcut has changed (it can, when a skill from an earlier version is updated), update your list.

Recipe 1: the Monday report for a small manufacturer

A small manufacturer selling tools direct wants one page every Monday: what sold, what is low and what needs a decision. It takes 40 minutes by hand. As a skill it takes two.

Shortcut: weekly

Goal: Produce a one-page Monday report on last week (Monday to Sunday) for
the store owner.

Steps:
1. Total orders, net sales and average order value for last week, and the
   same figures for the week before, with the change in percent.
2. Top five products by units and by net sales.
3. Orders still unfulfilled after 3 days, with order numbers.
4. Refunds issued last week: count and total.
5. Products with fewer than 10 units available that sold last week.
6. Three observations a store owner should act on. Label each as
   "from the data" or "my guess".

Format: headings for each step, tables for 2 and 3, under 350 words.
Do not: change anything in the store. If a figure is not available, write
"not available" instead of estimating.

Illustrative output: "Last week: 96 orders, $5,880 net sales, average order $61.25. Previous week: 88 orders, $5,100 (+15.3%). Unfulfilled after 3 days: 4 orders (#2381, #2390, #2394, #2397). Refunds: 5 totalling $312. Low stock that sold: Mallet CT250 (6 left)." The sum checks: $5,880 / 96 is $61.25, and ($5,880 - $5,100) / $5,100 is 15.3%.

What to check: compare the net sales figure with the figure on Shopify's own analytics for the same dates, at least for the first month. And read the "from the data" and "my guess" labels, because they stop you mistaking a hunch for a finding. What the report cannot do is tell you profit; ad spend and costs are not in it, so for that use the approach in the profit report tutorial.

Recipe 2: a product page audit that runs the same ten checks

The audit is the skill most likely to save real time, because product pages decay quietly: a missing image here, a blank description there. Give the skill a fixed checklist, and ask it to work on one collection at a time so the output stays readable.

Shortcut: audit

Goal: Audit the products in the collection I name and list what needs fixing.

Inputs: Ask me which collection, and whether to include draft products.

For every product, check:
1. Title present and under 70 characters
2. Description at least 40 words
3. At least 3 images, each with alt text
4. SEO title and SEO description filled in
5. Cost per item filled in
6. Weight filled in
7. Product type and at least one tag
8. Price not lower than the compare-at price
9. Inventory tracked
10. Not published with 0 stock and "continue selling" off, unless I say so

Output: a table with one row per failing product and the numbers of the
checks it fails. Then a count per check. Under 400 words.
Do not change anything. Offer to prepare fixes only for checks 7 and 9,
and list the changes for my approval before applying any.

Illustrative output for a collection of 38 products: "11 products fail at least one check. Check 4 (SEO text): 9 products. Check 3 (alt text): 7. Check 5 (cost): 3." Hand the SEO rows to the bulk method in the SEO rewrite tutorial, and fix the missing costs and alt text by hand while you are in the admin.

What to check: open three of the flagged products and confirm the fault is real. Skills that rely on a model's reading of a page sometimes call a long description "short" or miss alt text that exists. If it is wrong more than once in ten, tighten the wording of that check, for example "count the words in the description field and state the number".

Recipe 3: tag clean-up for a catalogue that grew untidily

Tags drive collections, filters and automations, and they decay: "Mens", "men", "Men's" and "mens-wear" all exist at once. A tag clean-up skill finds the near-duplicates and proposes a single tag for each group, then waits.

Shortcut: tags

Goal: Find messy product tags and propose a clean set.

Steps:
1. List every tag in use with the number of products carrying it.
2. Group tags that look like the same thing (case, plurals, hyphens,
   apostrophes, obvious abbreviations).
3. For each group, propose one tag name and show which tags it replaces
   and how many products would change.
4. List tags used on only one product separately, as "possible typos".
Output: a table: proposed tag, tags it replaces, products affected.
Do not: rename, merge or delete any tag. Do not touch tags beginning
with "auto-" or "flow-" because automations use them. Wait for me to say
which rows to apply.

The protected prefix line matters. Tags are often read by Shopify Flow workflows and theme filters, so renaming one can silently switch an automation off. Before approving a merge, search your Flow workflows and collections rules for the old tag. Illustrative output: "Group 1: proposed 'cast-iron'; replaces 'Cast Iron', 'castiron', 'cast iron' (41 products). Group 2: proposed 'gift'; replaces 'Gift', 'gifts' (17 products)." Approve one group, check a collection that depends on it, then approve the rest.

Recipe 4: a reorder list for a wholesale distributor

Shopify's Stocky app stopped working on 31 August 2026, and purchase orders now sit in the Shopify admin, with Sidekick suggested for reorders. Shopify's notes say Sidekick can draft a purchase order or transfer. The same notes say you cannot email a purchase order from the Shopify admin, so you download it as a PDF and send it to the supplier yourself. A skill gives you the weekly list.

Shortcut: reorder

Goal: Tell me what to reorder this week and how many.

Steps:
1. For each product that is tracked, compare units available with the
   units sold in the last 28 days.
2. Work out days of cover: units available divided by (units sold in 28
   days divided by 28).
3. List products with under 21 days of cover. For each, suggest an order
   quantity that brings cover to 45 days, rounded up to the case size in
   the product's "case size" metafield if there is one.
4. Group the list by vendor.
Output: one table per vendor: product, available, daily rate, days of
cover, suggested quantity. Flag any product with fewer than 28 days of
sales history as "low history".
Do not: create purchase orders. I will review and create them.

A worked line, illustrative: a wholesale distributor sells 84 units of a workshop apron in 28 days, which is 3 a day. It has 40 available, so cover is 40 / 3 = 13.3 days, under the 21-day line. To reach 45 days it needs 135 units, so the order is 135 less 40 already held, which is 95. If the case size is 24, round up to 96. The supplier lead time is not in this sum, and it should be: if the vendor takes 30 days, you need to order for the lead time plus cover. Add a line to the skill stating lead time per vendor.

What to check: new products and seasonal ones fool a 28-day window, hence the low-history flag. Also check what is already on order, because a skill that ignores open purchase orders will tell you to reorder what is on a ship. Add "subtract units on open purchase orders" to the steps if your store tracks them. For the wider replacement options, see the Stocky alternatives tutorial.

Recipe 5: support macros, drafted once and reviewed quarterly

Support macros are standard replies for the questions that arrive every day: where is my order, how do I return this, can I change the address. A skill drafts them from your actual policies, so the wording matches. It never replaces your review.

Shortcut: macros

Goal: Draft reply templates for our ten most common customer questions,
using only our published policies.

Inputs: Ask me to paste the shipping policy, returns policy and any FAQ.
Steps:
1. Propose the ten questions I am most likely to get, based on the
   policies pasted.
2. For each, write a reply of under 90 words, plain and friendly, using
   [first name], [order number] and [your name] as placeholders.
3. Quote the policy sentence each reply relies on, so I can check it.
Do not: promise refunds, dates or exceptions that the policies do not
state. If the policies do not answer a question, write "policy gap" and
suggest what I should decide.

The "policy gap" instruction is the most useful line. An assistant asked to write ten replies from a thin policy will invent the missing rules. Told to flag gaps, it produces a to-do list: perhaps you never said who pays return postage on a faulty item. Fix the policy first, then the macro. If you use Shopify's customer-facing AI as well, test it on the same questions, as the policy test walk-through describes, so both give the same answer.

Letting Sidekick write the skill by interview

You do not have to start from a blank dialog. When you choose New Skill, Sidekick asks questions about the task and saves the skill once it has enough detail. That route suits the first one or two, because the questions expose what you had not thought through: which week counts as "last week", whether drafts are included, whether refunds are netted off. Answer as precisely as you can. After it saves, open the skill and read the instructions it wrote. Compare them with the five-part structure above, and add the Do-not section yourself if the interview skipped it, as it often does.

A good habit is to run a task by hand in an ordinary chat first, correct the assistant until the result is what you want, and only then ask it to save the exchange as a skill. The shortcut for this is to type a slash and a name in the chat, then choose the option to create a skill from it. The saved version then carries the corrections you made, not the first guess.

Proving a skill works with a week you already know

Before a skill goes into your routine, run it on a period where you know the answer. For the Monday report, pick a week you have already reported by hand. Suppose your own figures were 96 orders, $5,880 net sales and 5 refunds totalling $312. Run the skill for those dates and compare line by line.

Each difference is a missing instruction. Add "net sales exclude shipping" and "ignore orders tagged pre-order" to the steps and run it again. When it matches twice in a row, you can trust it enough to read it in two minutes instead of rebuilding it. Do the same for the audit with a collection you have already checked by hand, and for the reorder list with a product whose stock you know.

Skills, Flow workflows and plain checklists: which tool for which job

Skills are for work where you want a person in the loop. Shopify Flow is better for rules that should run on their own, such as tagging an order above a certain value. Sidekick can generate a Flow workflow from a plain description (on a desktop device; Shopify leaves generated workflows inactive until you review and turn them on), which suits a skill whose output you approve identically every time. If you find you press "approve" without reading for a month, that job has become a candidate for automation, and the right step is to build the rule, not to keep the skill.

• Keep as a skill: anything where you read the result and decide, such as the weekly report and the reorder list.
• Move to a Flow rule: a clear if-then, such as tagging orders from a given source. The tag clean-up, once your naming is settled, becomes a rule applied at the moment a product is created.
• Keep as a checklist: physical or off-system work, such as checking a batch before it ships. No skill sees that.

A Monday routine built from the five

With all five saved, a store owner's first hour might run like this, with illustrative timings: 5 minutes for the weekly report, 10 for the reorder list and the supplier emails that follow from it, 5 for the support macros check if any customer questions did not fit, and 20 minutes once a month on the audit and the tag clean-up. The first week will be slower, and the third will be faster because you will have fixed the instructions that caused rework. Aim to find five skills that together save an hour a week, rather than twenty skills that each save a minute.

Share the routine with anyone else who works in the admin. A one-page list of shortcuts and what each does is enough. If two people have different versions of "reorder", you will get two reorder lists and a muddle at the supplier. Keep one owner and one master copy, as the maintenance list below sets out, and note who last edited each skill.

When a skill should not be saved at all

A few jobs look repetitive and are poor fits. Anything that depends on judgement about a customer, such as deciding whether to refund a borderline return, should stay a human decision with Sidekick as a drafting aid. Anything with legal or financial consequences, such as price changes across the whole catalogue, deserves a human reading every line. And one-off clean-ups, like fixing a supplier's file that arrived badly formatted, are better done in a chat and forgotten. The test is the same as before: the work repeats, the inputs come from your store, and you read the output before anything changes.

A skill that did harm: the one with no guard rails

A small online reseller saved a skill called "tidy" with the instruction "clean up product titles". On its first run the assistant proposed removing brand names from forty titles to make them shorter, and because the owner approved the list without reading, product pages lost the very terms customers searched for. Nothing broke technically. The fault was in the instructions, which gave no format, no scope and no list of things never to change. The rewrite listed what to leave alone (brand, model codes, sizes), capped changes to 20 products per run and asked for a before-and-after table. The lesson: a vague skill is a vague prompt repeated weekly. (Illustrative.)

Test and maintain your skills

Treat the first month as a trial, using the known-week comparison above for each skill, and note the date and any fix in a small file, one line per change. Then, once a quarter, read each skill's instructions again. Your shipping policy may have changed, the case size field may have been renamed, or a vendor may have gone. A skill that quietly uses an old assumption is worse than none.

• Name the owner: if more than one person uses skills, one person edits them. Otherwise two versions will exist.
• Keep a master copy: store the text of each skill in a document outside Shopify. Skills can be edited or deleted, and you want a record of what worked.
• Cut what you do not use: with a limit of 25 or 50, a pile of half-used skills hides the five that matter.
• Review what it saved you: if a skill does not save at least ten minutes a week, edit it or delete it.

If you are still deciding whether Sidekick is the right assistant for each job, the Sidekick store tasks tutorial covers what it does well, and the comparison of Sidekick, ChatGPT and Claude helps you place the jobs that belong elsewhere. A rule I would follow: keep skills for work that sits inside Shopify, and use an outside assistant when you need to join data from other places, such as ad accounts.

Questions about saving Sidekick skills

How many Sidekick skills can I save?

Shopify's help page says you can create up to 50 skills, and that some stores have a limit of 25. Prompts saved as skills in earlier Sidekick versions do not count towards the limit. Five well-written skills usually cover a small store's weekly routine.

How do I run a skill?

Either open Skills from the plus icon in the Sidekick field, select the skill, add any details and send it, or type a slash followed by the skill's shortcut in the chat. Shopify notes that in some stores a skill runs only when you select it or enter its shortcut.

Will a skill change my store without asking?

Shopify says Sidekick shows changes for you to review before applying them. Still, write that into any skill that edits products, tags or orders: ask it to list the proposed changes first and wait for approval, so the habit holds if the behaviour changes.

Can I edit a skill later?

Yes. Open Skills from the plus icon, select the skill and use the edit icon to describe the changes, or edit the shortcut and instructions directly if an Edit Skill dialog opens. You can delete a skill from the same place after confirming.

Further reads

• AI for Ecommerce Analytics: Shopify Reports, GA4 or ChatGPT First? — Decide which report to open before the weekly one.
• Best AI for Ecommerce Operations: Orders, Stock, Returns, Suppliers — Wider view of AI for orders, stock and returns.
• AI Product Recommendations on Shopify: Built-In vs Paid Apps — Compare built-in and paid recommendation tools.
• How Much Does It Cost to Add AI to a Shopify Store? — Know what AI additions cost before you add them.
• Shopify Magic vs ChatGPT for Product Descriptions — Compare built-in and outside tools for copy.
• Shopify Pre-Order App: Which to Pick, Then Set Ship Dates With AI — Another repeat job: pre-order wording and delay emails.
• Purchase Order Tracking: Build a Supplier Chase Sheet With AI — Chase the supplier orders your reorder list creates.
• Ecommerce AI Chatbot Examples: 7 Store Conversations to Copy — Borrow tone and layout for the support macros.
• Best AI Ecommerce Store Builder: Shopify, WooCommerce or BigCommerce? — Shopify has Sidekick, BigCommerce has Companion, WooCommerce has plugins only. What each built-in AI does, and what the plans cost.
• Applications of AI in E-Commerce: 9 Uses Mapped to a Small Online Store — Nine uses of AI for a small store, ranked by weekly hours saved and matched to Sidekick, Shopify Inbox, Klaviyo and Gorgias, with plan costs.
• Does Shopify Have AI? What Is Built In and What Costs Extra — Which Shopify AI features come with Basic, Grow, Advanced and Plus, at what price, and where the extra costs begin.
• Shopify Sidekick Pulse: Which Recommendations to Act On — A scoring method for Pulse cards: effort, revenue, evidence in your reports, and the generic suggestions to skip.
• Shopify Sidekick Review: Is It Reliable for Store Decisions? — A test routine for Sidekick: known-answer questions, number checks against reports, and a review habit for changes before you apply them.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: Shopify Help Center (Sidekick overview, Sidekick skills, creating Flow workflows, transitioning from Stocky), checked November 2026.