Adapted from the DGF Agent Basics series — consumer-first guides for working with a personal AI agent. Pick only what you need.
> Pick only what you need. This skill is individual. This is one person's thoughts, offered as a pre-configured help, not a must. Take only the value you need from it, prune it to what fits you, and leave the rest. You do not have to use what the author thinks, and you do not have to take any other skill with it.
Summary: Shop for everyday items, new or used, with a sweep that goes around Amazon, cashback checked before checkout, and a cart handed over for you to review and pay yourself. Optional.
Use this skill when the user wants to shop for an everyday item: anything they are shopping for, new or used, for home, work, or personal use. It covers item shopping only. If the user asks for something outside item shopping, handle that as a separate task, without assuming another skill exists.
This skill stands alone. An agent that knows nothing else can follow it from the startup sequence below, then the item shopping method. This skill works on its own, with no other skill needed, and no knowledge of any other skill assumed.
Startup and how to use
Start by getting your own personal agent. Get your own personal agent and set it up. Bring your own accounts and logins. Use nothing that belongs to anyone else.
Tell your agent what you care about in plain words. Name your business, your work, your family, your hobbies, or any topic you want it to follow. Talking is the setup. You do not have to set up a system first.
Ask for the plan first, before any work runs. Say the task, then ask for the exact steps it will take, what it needs from you, and anything it cannot do. Wait for you to approve or correct the plan. Review the plan like a checklist and fix a wrong step or add a missing step before work starts.
Check that your agent understood you, before it runs. Ask it to show back what it thinks you mean and what output it will give you. Correct its repeat back right there when it does not match what you pictured.
Work in small single steps. Count the steps first. Give one step, check it, then give the next step. A ten step run is a huge run for an agent, so break it down and keep each step to exactly one thing.
Show it once, then save it by name. This is how you create a new task or a new skill. Take over the browser and do the task once yourself while your agent pauses. You drive alone and your agent does not watch your clicks. There is no record button to press. After you hand control back, tell your agent the steps in chat and tell it to save them by name, such as save this as Fun Things. Later, say pull up Fun Things and it picks up where you left off.
Correct it in small plain words when it does it wrong. Do not start over. Say what to change and say to remember that correction. That preference sticks and it gets better for you specifically over time.
Let it keep learning your interests every day. Your Feed looks like a newspaper with articles and topics related to what you told it you care about, updated daily. Your Ideas look like a lightbulb with suggestions for things you might like. It is not mind reading. It learns from what you actually talk about.
Use easy first and deep only when you want it. Never require a deep setup to use a basic task. Keep the wording in plain benefits and do not lead with virtual machine or VM. Say it has its own browser and workspace built in, you do not have to set anything up, it keeps your logins safe, and it saves and remembers your stuff.
Before any search
Read the request and list every fact that decides whether a result is correct. Do not search until those facts are known or the user has said to search without them.
Ask the user for the missing deciding facts in one plain message. Ask at most three short questions at a time.
Tell the user up front when a store cannot be searched. Amazon blocks automated browsing, so do not claim an Amazon price you did not verify. If Amazon matters, ask the user to check Amazon themselves or to paste the exact product link, then compare that link against the other results.
Never ask for a password or a card number in chat. Tell the user to load logins through the secure sign-in your agent shows them. If a login was not loaded securely, search public prices only and do not look inside that account.
Never spend the user's money. Finding a product, building a cart, and showing a total are not a purchase. The user reviews the cart and pays themselves.
Everyday goods: the search order
Collect the exact item first. Ask for model, variant, size, pack, condition, budget ceiling, delivery ZIP, and pickup or delivery. Do not search for a lookalike when the user asked for an exact item.
Ask the user to load in what they already hold before searching. Ask about value memberships such as Costco, Sam's Club, BJ's, and Walmart Plus. Ask about credit card loyalty and cashback by card name only. Ask about cashback tools, store loyalty accounts, and delivery preferences.
Search in this order for the exact item. Start with eBay for new and used, checking delivered total, seller rating, and returns. Then check Facebook Marketplace for local and used, checking price, condition, and safe pickup. Then check local stores for pickup or same day, such as Walmart, Target, Best Buy, and Home Depot, plus local shops that fit the item. Then check the brand direct store and general search. Then check budget and value stores with the membership loaded in, so a members only price is visible.
Check cashback and coupons before checkout, never after. Pick one cashback portal for that purchase. Click through that portal before checkout. Compare its live rate for that store, because rates change by store and by day.
Use these cashback and coupon tools by their plain function. Rakuten is a general portal with thousands of stores, paid quarterly by PayPal or check. TopCashback often shows a higher rate on the same store. BeFrugal covers thousands of stores and offers a price guarantee. Capital One Shopping is a browser extension that hunts coupons and compares prices, and its rewards come as gift cards. PayPal Honey hunts coupons at checkout and earns points for gift cards. Ibotta and Fetch Rewards work from receipts or linked loyalty accounts for groceries and in store shopping. Swagbucks and MyPoints pay points for shopping and small tasks, redeemable for gift cards, and are slower than cash back.
Warn the user when a coupon could cancel cash back. Compare the final total with cash back included, not the coupon alone. Do not stack two cashback portals on the same order.
Compare the top options by delivered total, returns, warranty, and seller trust. Treat a used local deal with no returns as a different product from a new retail deal with returns, even at the same price.
Categories that enhance item shopping
Condition: new, open box, refurbished, and used are different products, even at the same price. State condition, what is included, what is missing, and whether returns apply, for each option.
Compatibility and fit: check size, model, version, connector, and fitment before presenting an option, especially for electronics, auto parts, clothing, and home goods. A cheaper item that does not fit is not a saving.
Warranty and protection: check manufacturer warranty, seller warranty, extended protection, and what voids it. State warranty length and who honours it, for each option.
Returns and exchanges: check return window, restocking fees, return shipping cost, and whether a used or local item can be returned at all. State the return terms in plain words.
Price history and timing: check whether the price is a sale price, a regular price, or a clearance price, and whether a sale, a holiday, or a new model release is coming that would change it. Offer a price watch with a plain stop waiting date when the user can wait.
Reviews and seller trust: check seller rating, review count, recent reviews, and complaint patterns, not just a star average. A high average with recent complaints about the exact problem the user cares about is a warning.
Delivery and pickup: compare delivery time, delivery cost, pickup distance, and pickup hours, as part of the delivered total. A cheap item that arrives after it is needed, or far away, is not the best option.
Bundles and accessories: check what else the item needs to work, such as cables, batteries, mounts, or installation, and include those costs in the total when the user needs them. A bundle that adds unwanted extras is not a saving.
Product categories to ask about, so the search starts in the right place: electronics and appliances, clothing and shoes, home and garden, tools and hardware, sports and outdoors, toys and games, books and media, health and beauty, auto parts and accessories, office and school, and baby and kids. Ask which category the item belongs to, and search that category's stores first, then the general sweep.
Safety and recalls: check for recalls, safety warnings, and age or use limits when the item is for a child, for safety, or for a vehicle. A recalled item is not presented as an option, even at a low price.
Hand over the deliverable: cart set, user just pays
At the end, hand the user a finished cart, not just a summary. The user should be able to go from your message to paying, with no searching left to do.
Hand it over in one of two ways, whichever fits the seller. First, through the your agent browser: prepare the browser task with the chosen items in the cart, quantities, variants, and delivery or pickup choice set, cart page loaded, and hand control to the user in browser takeover, so they review and pay themselves. Second, with the exact URL: give the exact cart or checkout link, verified to open to that cart with those items and quantities, not a generic product page.
Before handing it over, check the cart against the decisions made. Check item, variant, size, condition, quantity, price, delivery or pickup, and delivered total, line by line. If anything differs from what the user decided, fix it before handing it over, and say what changed.
State plainly what is set and what is not set. State when the cart is set and ready to pay, when a link is a product page that still needs the user to add it to cart, and when a seller has no cart link and the user must add items themselves, with exact steps and exact product links. Never claim a cart is set when it is not.
Your agent never pays. The user reviews the cart, enters payment themselves, and places the order themselves. Handing over a ready cart is not placing an order.
Diverse, not concrete: let the user prune
Do not make this skill concrete, narrow, or one size fits all. Present a diverse set of options first, across price, style, risk, format, and approach, so different people can see themselves in it.
Let the user prune. Ask what to cut, what to keep, and what to change, and then narrow to what they kept. Pruning is the user's job, not your agent's guess. Do not prune for them before they have seen the options.
Do not lock a single choice early. Offer at least two or three meaningfully different options at each decision point, with the trade for each, unless the user already locked that choice themselves.
Keep options that do not fit one person, because they may fit another person using the same skill. A skill that only works for one kind of person is too concrete to reuse.
When the user prunes, save the pruned version by name when they want to repeat it, so their version starts from what they kept, not from the full diverse set again.
Present the result
Lead with the recommendation or the trade, in plain words. Keep the reply short and name only the decisive difference between the top choices.
State every price as a delivered total when a total decides the choice. Include shipping, tax, and fees when they are known, and say when a fee is unknown instead of guessing it.
State what you could not verify. State when a price is stale, a store was blocked, a membership was not loaded, or a fitment is provisional.
Stop after a denial. Do not re-pitch a rejected option in new words.
Available connectors
A product catalog search, if your agent has one, for fast product search with prices, images, variants, and checkout eligibility.
[Facebook](/opt/hatch/connectors/facebook_cli/CONNECTOR.md) for Marketplace listings.
[Instagram](/opt/hatch/connectors/instagram/CONNECTOR.md) for products shown in Instagram posts.
Use your agent's connector and skill search, if it has one, to find other connectors and related skills.
If nothing here fits, look on the web for a maintained API or MCP server and set it up with your agent's custom connector setup, if it has one, or build a custom connector.
Websites
[Walmart](https://www.walmart.com/) for general goods and grocery pickup in the US. Carry the exact item, ZIP, and pickup or delivery choice into the search and verify availability on its own page.
[Target](https://www.target.com/) for general goods and local pickup in the US. Verify availability and pickup terms on its own page.
Alternatives for better results
Offer a larger pack, a store brand, a used or refurbished unit, or a price watch when the user is flexible, and explain the trade in plain words.
Offer pickup when delivery fees, minimum orders, or tips outweigh the convenience of delivery.