agentleFS
taivoptaivop/marketplace/skills/estonian-store-search/SKILL.md

estonian-store-search

Skill17 starsChanged 56 days agoMITView on GitHub

Search Estonian building supply stores (Bauhof, Ehituse ABC, Decora, K-Rauta, Bauhaus, Depo), grocery stores (Prisma, Rimi, Selver), home furnishing (IKEA), and second-hand marketplaces (Vinted, Yaga) for products, prices, and stock. USE WHEN user asks to find products at Estonian stores, compare prices across them, or check availability.

  • Sends data out

The file

We split this skill into the rules it gives an agent

7 rules. About 0 are common, 2 are distinctive, and 5 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 5

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Run curl searches across stores in parallel.

    line 13
  2. Distinctive:

    Search APIs are candidate discovery, not final proof that a product page or its availability is current.

    line 22
  3. Distinctive:

    Never invent a public URL from a slug or replace an API/staging hostname without applying the store-specific rule below and verifying the …

    line 32
  4. Very specific:

    Search availability hints: in_stock, tags, and delivery_methods

    line 97
  5. Very specific:

    Do not treat tags, including "in-shops", or in_stock as final stock truth.

    line 99
  6. Very specific:

    Search is store-specific, so pick a storeId first:

    line 151
  7. Very specific:

    No JSON-LD; ingredients are under the "Koostisosad" heading and the marketing copy under the description section.

    line 210

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Run curl searches across stores in parallel. By default, filter out out-of-stock items.

Building supply stores: Bauhof, Ehituse ABC, Decora, K-Rauta, Bauhaus, Depo. Grocery stores: Prisma, Rimi, Selver. Home furnishing: IKEA. Second-hand marketplaces: Vinted, Yaga.

Verify every product link before returning it

Search APIs are candidate discovery, not final proof that a product page or its availability is current. Before returning a product URL:

  1. Open the final public URL, follow redirects, and verify that the rendered product name or stable identifier (SKU, EAN, item number) matches the search record. Also check the requested variant, dimensions, and pack size.
  2. Reject 404 pages, generic storefront shells that never hydrate, product pages that say the item is unavailable, and pages that only show substitute items.
  3. Let the detail page override a stale search index when price or availability conflicts. Record the date checked for price/stock comparisons.
  4. Never invent a public URL from a slug or replace an API/staging hostname without applying the store-specific rule below and verifying the result.
  5. Some sites block command-line clients. If curl reaches an anti-bot page or a JavaScript shell, use an available browser-capable tool and inspect the rendered title/name and availability. If no browser is available, omit the link or label it explicitly as unverified; do not call it verified.

For dynamic pages, an initial generic <title> is not sufficient. Wait for the product view to render, then verify the visible product identity.

Bauhof — Magento GraphQL

curl -s -X POST "https://www.bauhof.ee/api/magento/customQuery" \
  -H "Content-Type: application/json" \
  -d '{"query":"query{products(search:\"SEARCH_TERM\",pageSize:10){items{sku name stock_status price_range{minimum_price{final_price{value currency}regular_price{value currency}}}url_key}total_count}}","queryVariables":{}}'

Product URL: https://www.bauhof.ee/et/p/{sku}/{url_key}. The bare https://www.bauhof.ee/{url_key} form redirects to the English site and can produce a 404. Verify the rendered heading and Tootekood against name and sku. Stock: stock_status (IN_STOCK / OUT_OF_STOCK); this is aggregate availability, so use the detail page's purchase controls or store-availability view when location-specific stock matters. Price: final_price.value.

Ehituse ABC — Klevu Search

curl -s -X POST "https://eucs32v2.ksearchnet.com/cs/v2/search" \
  -H "Content-Type: application/json" \
  -d '{"context":{"apiKeys":["klevu-168180264665813326"]},"recordQueries":[{"id":"productSearch","typeOfRequest":"SEARCH","settings":{"query":{"term":"SEARCH_TERM"},"typeOfRecords":["KLEVU_PRODUCT"],"limit":10,"sort":"RELEVANCE"}}]}'

Product URL: url. Stock: inStock (yes/no). Price: salePrice. Results in queryResults[0].records. Open the returned URL and verify its product name or SKU before including it; Klevu records can outlive the storefront page.

Decora — Klevu Search

curl -s -X POST "https://decoracsv2.ksearchnet.com/cs/v2/search" \
  -H "Content-Type: application/json" \
  -d '{"context":{"apiKeys":["klevu-159479682665411675"]},"recordQueries":[{"id":"productSearch","typeOfRequest":"SEARCH","settings":{"query":{"term":"SEARCH_TERM"},"typeOfRecords":["KLEVU_PRODUCT"],"limit":10,"sort":"RELEVANCE"}}]}'

Same response format as Ehituse ABC. Extra fields: stock_availability (store locations), volume_l, shortDesc, usage_indoor_outdoor. Verify the returned URL's rendered product name/SKU and do not rely on a successful HTTP status alone, because some storefronts serve branded not-found pages with status 200.

K-Rauta — LupaSearch

curl -s -X POST "https://api.lupasearch.com/v1/query/j9gky3z0nx3z" \
  -H "Content-Type: application/json" \
  -d '{"searchText":"SEARCH_TERM","limit":10}'

Results are in .items[] (the top level is an object with items, facets, filters, total, and other search metadata). Useful fields include:

  • Product name: title.et
  • Relative product URL: url.et (join with https://www.k-rauta.ee)
  • Product code: product_code
  • Prices: price_groups.default.price and price_groups.loyalty.price
  • Search availability hints: in_stock, tags, and delivery_methods

Do not treat tags, including "in-shops", or in_stock as final stock truth. Search records can remain positive while the rendered product page says Vabandame, antud toode pole hetkel saadaval and shows substitutes. Open the detail URL and exclude the item if that message appears or no purchase option is available. Keep regular and loyalty prices clearly distinguished.

Bauhaus — Magento REST (2-step)

Step 1 — search (returns product IDs):

curl -s "https://secure.qs-m2web.bauhaus.ee/rest/V1/search?searchCriteria[pageSize]=10&searchCriteria[requestName]=quick_search_container&searchCriteria[filterGroups][0][filters][0][field]=search_term&searchCriteria[filterGroups][0][filters][0][value]=SEARCH_TERM"

Step 2 — get details (entity IDs from step 1):

curl -s "https://secure.qs-m2web.bauhaus.ee/rest/V1/products-render-info?searchCriteria[filterGroups][0][filters][0][field]=entity_id&searchCriteria[filterGroups][0][filters][0][value]=ID1,ID2,ID3&searchCriteria[filterGroups][0][filters][0][condition_type]=in&storeId=13&currencyCode=EUR"

Results are in .items[]. Each item currently includes name, price_info.final_price, price_info.regular_price, is_salable, images, and url. The returned url may use the non-public https://headless-staging.bauhaus.ee hostname. Build the public candidate by keeping its path and using https://www.bauhaus.ee, then verify the rendered page before returning it. Do not expose the staging URL.

is_salable == "1" means the item can be sold; it is not a quantity. The public detail page may expose stronger signals such as Laos (N), delivery status, and Click & Collect availability. Command-line requests may receive a Vercel security response, so use a browser-capable tool for verification when needed.

Depo — GraphQL

curl -s 'https://online.depo.ee/graphql' \
  -H 'Content-Type: application/json' \
  -d '{"query":"{ products(searchString: \"SEARCH_TERM\", start: 0, rows: 10) { edges { node { id name primaryBarcode specificationBrand prices { yellow { priceWithVat unit } orange { priceWithVat priceQuantity unit } } stockItems { locationAddress quantity } } } } }"}'

Name, brand, barcode, and stock counters associated with Tallinn Veskiposti. Two price tiers: yellow (retail), orange (bulk price with priceQuantity threshold). Prices include 24% VAT.

Product URL: https://online.depo.ee/product/{id}. The page is client-rendered, so wait for hydration and verify the visible name/barcode. DEPO's product page explicitly says its availability information applies to online orders and does not guarantee the corresponding quantity in the physical store. Report stockItems.quantity as online availability associated with the location, not as confirmed shelf stock. The detail page's Saadaval/purchase state is the final online-availability check.

Prisma (grocery) — GraphQL

Search is store-specific, so pick a storeId first:

curl -s -X POST "https://graphql-api.prismamarket.ee" \
  -H "Content-Type: application/json" -H "x-client-name: skaupat-web" \
  -d '{"query":"{ stores{ id name } }"}'

Stores (as of 2026-06): Mustamäe 542855267, Annelinna 697951770, Sikupilli 613763242, Lasnamäe 660914326, Rocca Al Mare 634972293, Sõbra (Tartu) 685847808, Kristiine 542860184. Default to Mustamäe if unspecified.

Then search:

curl -s -X POST "https://graphql-api.prismamarket.ee" \
  -H "Content-Type: application/json" -H "x-client-name: skaupat-web" \
  -d '{"variables":{"storeId":"542855267","queryString":"SEARCH_TERM","limit":10,"from":0},"query":"query($storeId: ID!, $queryString: String!, $limit: Int, $from: Int){ store(id:$storeId){ products(queryString:$queryString, limit:$limit, from:$from){ total items{ ean name brandName price priceUnit comparisonPrice comparisonUnit slug } } } }"}'

Results in data.store.products.items. price (per priceUnit, e.g. KPL/KG), comparisonPrice per comparisonUnit (e.g. €/L). Product URL: https://www.prismamarket.ee/et/tuote/{slug}/{ean}. Prices include VAT. total is the full hit count for pagination via from.

Single product details — root field product(id, storeId) where id is the EAN:

curl -s -X POST "https://graphql-api.prismamarket.ee" \
  -H "Content-Type: application/json" -H "x-client-name: skaupat-web" \
  -d '{"variables":{"storeId":"542855267","id":"EAN"},"query":"query($storeId: ID!, $id: ID!){ product(id:$id, storeId:$storeId){ ean name brandName price priceUnit comparisonPrice comparisonUnit slug productType countryName{ et } description ingredientStatement } }"}'

Adds description, ingredientStatement (ingredients), countryName{ et } (country of origin) on top of the search fields.

Rimi (grocery) — server-rendered search page

curl -s "https://www.rimi.ee/epood/ee/otsing?query=SEARCH_TERM&pageSize=80"

Returns an HTML page (locale ee, not et). Each product card embeds JSON: data-gtm-eec-product='{"id":"200280","name":"...","price":1.39,"currency":"EUR"}'. Extract those with grep/regex. Product URL: the href of the sibling <a class="card__url" ...> (pattern /epood/ee/tooted/.../p/{id}). Per-unit price (e.g. 0,93 €/l) is in the card text. Out-of-stock cards carry an out-of-stock class. pageSize controls result count (80 works); page through with &page=N.

For lightweight autocomplete (max 5 results) instead:

curl -s "https://www.rimi.ee/epood/api/v1/search/suggestions/products?query=SEARCH_TERM" \
  -H "X-Requested-With: XMLHttpRequest"

Single product details — fetch the product page and parse its JSON-LD:

curl -s "https://www.rimi.ee/epood/ee/tooted/.../p/PRODUCT_ID"

The <script type="application/ld+json"> block gives name, description, sku, image, offers.price, offers.priceCurrency, offers.availability (InStock/OutOfStock). Ingredients live in the HTML under the "Koostisosad" heading (no clean JSON — parse the section if needed).

Selver (grocery) — Klevu Search

curl -s -X POST "https://eucs3v2.ksearchnet.com/cs/v2/search" \
  -H "Content-Type: application/json" \
  -d '{"context":{"apiKeys":["klevu-14410928010151845"]},"recordQueries":[{"id":"productSearch","typeOfRequest":"SEARCH","settings":{"query":{"term":"SEARCH_TERM"},"typeOfRecords":["KLEVU_PRODUCT"],"limit":10,"sort":"RELEVANCE"}}]}'

Same Klevu response format as Ehituse ABC / Decora. Results in queryResults[0].records; totalResultsFound for paging (via offset in settings). Per record: name, sku, price/salePrice, inStock (yes/no), url (full product URL), image, klevu_category.

Single product details — fetch the product page (url from the search record) and parse the server-rendered HTML:

curl -s "PRODUCT_URL"

No JSON-LD; ingredients are under the "Koostisosad" heading and the marketing copy under the description section. Price/stock/name are already in the Klevu record, so only hit the page when you need ingredients or description.

IKEA (home furnishing) — Search API

curl -s "https://sik.search.blue.cdtapps.com/ee/et/search-result-page?types=PRODUCT&q=SEARCH_TERM&size=10&c=sr&v=20240110" \
  -H "User-Agent: Mozilla/5.0"

Results in searchResultPage.products.main.items[].product. Per product: name, typeName (e.g. "Põrandalamp"), itemMeasureReferenceText (size), id/itemNo (article number), pipUrl (product URL), mainImageUrl. Price: salesPrice.numeral + salesPrice.currencyCode (EUR, incl. VAT). The search response's availability[] is coarse (delivery flags only); for real stock use the availability API below. Total hit count is searchResultPage.products.badge.

IKEA — in-store + delivery stock

For actual stock, query the availability API with the itemNo from search. The Tallinn store (the only IKEA in Estonia) is store id 648.

curl -s "https://api.salesitem.ingka.com/availabilities/ru/ee?itemNos=ITEM_NO&expand=StoresList,Restocks,SalesLocations" \
  -H "X-Client-ID: ef382663-a2a5-40d4-8afe-f0634821c0ed" \
  -H "Accept: application/json;version=2" \
  -H "User-Agent: Mozilla/5.0"

Results in availabilities[], one entry per store (classUnitKey.classUnitType == STO, classUnitCode == store id; 648 = Tallinn). Per store:

  • availableForCashCarry — whether it can be bought in-store now.
  • buyingOption.cashCarry.availability.quantity — units physically in the Tallinn store.
  • buyingOption.cashCarry.availability.probability.thisDay.messageType — HIGH_IN_STOCK / LOW_IN_STOCK / OUT_OF_STOCK (with a colour token).
  • buyingOption.homeDelivery.range.inRange — whether home delivery is offered.

Pass multiple itemNos comma-separated. X-Client-ID is the public web key from the IKEA storefront; the Accept: application/json;version=2 header is required.

Vinted (second-hand marketplace) — catalog API (2-step)

The API needs an anonymous session cookie. Fetch the homepage once to get cookies, then call the catalog endpoint with them.

Step 1 — get a session cookie:

curl -s -c /tmp/vinted_cookies.txt -o /dev/null \
  "https://www.vinted.ee/" \
  -H "User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/124.0 Safari/537.36"

Step 2 — search the catalog:

curl -s -b /tmp/vinted_cookies.txt \
  "https://www.vinted.ee/api/v2/catalog/items?search_text=SEARCH_TERM&per_page=10&order=newest_first" \
  -H "User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/124.0 Safari/537.36" \
  -H "Accept: application/json"

Results in items[]. Per item: title, brand_title, size_title, status (condition in Estonian, e.g. "Hea"/"Rahuldav"), price.amount + price.currency_code (EUR), path (append to https://www.vinted.ee for the product URL), photo.url, user.login. Total hit count is pagination.total_entries (pagination.total_pages for paging via &page=N). order accepts newest_first, price_low_to_high, price_high_to_low, relevance. A real User-Agent is required or the request is blocked. Everything is second-hand/used; there is no "stock" — an item is either listed or sold.

Yaga (second-hand marketplace) — search API

curl -s "https://www.yaga.ee/api/product/search/?query=SEARCH_TERM&offset=0&limit=60" \
  -H "User-Agent: Mozilla/5.0" -H "Accept: application/json"

Note the trailing slash on /search/ and the param name query (not q). No cookie or auth needed. Do NOT use /api/product — that is a generic popularity feed that ignores the query and always returns the same listings.

Results in data.list[]. data.total is the real match count; data.searchEngine names the backend (e.g. legacy-postgres). Per item: brand, price (EUR, incl. VAT), size, slug, shop.activeSlug, likeCount, images[].gallery (image URL, also ?s=600/?s=300 size variants). Product URL: https://www.yaga.ee/{shop.activeSlug}/product/{slug}. Page with offset (multiples of limit).

Listings have no free-text title/description field — they are brand + size + attributes + photos — so matching is on brand/category, not a title string. Gibberish queries correctly return total: 0. The search item's own brand field is often null/unreliable; treat the query term as the brand.

Filtering the search (avoid the womenswear/off-palette trap)

The search payload has no category, colour, or gender per item, so a bare brand query mixes in women's and off-palette listings. Two complementary tools:

  1. Server-side filters — append as URL-encoded JSON int-arrays, format key=[1,2] (a wrong format returns a VALIDATION_ERROR naming the param + expected shape, which is how to discover more). Working keys: colors, materials, brands, conditions. Example (navy+grey+brown wool): &colors=[9,11,14]&materials=[16]. The categories key exists but uses a bespoke nested union that resisted reverse-engineering — don't filter by category server-side; post-filter via the detail API instead.
  2. Detail API — https://www.yaga.ee/api/product/{slug}/ (or /{id}/) returns the rich fields the search omits: description (the only human text — has brand/cut/measurements/colour), categoryId, categoriesMap (root→leaf path, e.g. [2,471]), colors, materials, condition.namesMultilang.en. Gender = categoriesMap[0]: men's 2, women's 1, kids 3. Pipeline that works: brand query + colors=[palette] server-side → fetch each hit's detail (parallelise) → keep categoriesMap[0]==2 and the leaf categories you want → read description to confirm.

Taxonomy lookups (id ⇄ name)

  • https://www.yaga.ee/api/category/ — full tree; top level is flat, children nest under each node's categories key. Men (2) leaves incl.: knits 298, shirts 167, tees 168, jeans 300, trousers 301, dress-trousers 372, blazers 306, coats 305, suits 471, loafers/shoes 312, sneakers 318, belts 323.
  • https://www.yaga.ee/api/color/ — colour ids. Palette-relevant: black 1, blue 3, navy 9, grey 11, white 13, brown 14, light-brown 15, cream 16, khaki 19.
  • https://www.yaga.ee/api/brand/ — brand id→name (needed for the brands filter).

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.