SuperAPI AI Directory

linkfox-wb-temu-data

Data & analytics

View repository

Updated 10-28 23:57

About this skill

[!IMPORTANT] This Skill must rely on the MCP "LinkFox". If this MCP is not installed, please explicitly notify the user: Please install the LinkFox connector first:

  • In WorkBuddy, find "Experts · Skills · Connectors" on the left → "Connectors";
  • Search for the keyword "LinkFox" and install the connector "LinkFox" (note: it is linkfox, not linkfox-product-selection), and complete login authorization following the instructions;
  • Return to WorkBuddy and retry.
LinkFox Temu Product Research

A unified data query capability for Temu product selection and market research: it helps the Agent locate market opportunities from the three dimensions of products, stores and categories, and returns structured lists, counts, sorting results and reusable category identifiers according to explainable filtering conditions. It is suitable for product filtering, store performance research, category positioning, fully managed/semi-managed comparison, and subsequent conditional cross-queries. This capability covers 3 sub-capabilities and 4 operations: product query (linkfox-temu-product-query, 1), store query (linkfox-temu-store-query, 1), and category search and sync prerequisite (linkfox-temu-category-search, 2).

Capability boundaries

✅ Scope of capabilities

  • Temu product filtering and ranking: Use linkfox-temu-product-query to filter via ehunt/temu/productQuery by keyword, product ID/store ID, frontend or backend category, price, rating, review count, total/weekly/daily sales, listing date, listing status, fully managed/semi-managed, semi-managed region and tags, and sort by price, sales or rating. Suitable for product pool discovery, initial screening of bestsellers, price band comparison and product selection candidate organization.
  • Temu store filtering and performance research: Use linkfox-temu-store-query to filter via ehunt/temu/storeQuery by store name/ID, country site, backend category, management mode, total/weekly/monthly sales and revenue, rating, reviews, followers, product count and store opening date, and support store rankings.
  • Temu category ID positioning: Use linkfox-temu-category-search to search via ehunt/temu/temuCategorySearch in the synced local category library by Chinese name, English name or category ID substring, returning the hierarchy, parent category, whether it has child categories, and the id/categoryId that can be used for product/store queries.
  • Category data sync prerequisite check: Sync the category tree to the local library via ehunt/temu/syncTemuCategory (no parameters); when the category search is empty and the user does need to find a category, first explain the sync prerequisite and call this operation, then search again.
  • Cross-dimension cross-query: First use category search to obtain the category ID, then put the ID into the product query's categoryHome/categoryBackend or the store query's category; first use store results to obtain the storeId, then return to the product query's searchKey to locate the store's product pool. The returned count and billing for each step are recorded separately.
  • Structured results and traceability: The query endpoints return products, stores or categories lists along with total, source and rendering metadata; retain the parameter summary, result identifiers and points fields for traceability.

❌ Boundaries and limitations

  • Data format: This capability provides filtered/search results and aggregate fields; it is not equivalent to a product detail page, store detail page, or the original review text. It cannot be used to fill in SKU, variant, logistics, profit, supplier, advertising, or order information that was not returned.
  • Category prerequisites: Category search relies on an already-synced local category library; when syncing is incomplete, a category has been deleted, or keywords do not match, empty results may be returned. Empty results must not be inferred to mean that Temu does not have that category.
  • Filter semantics: Range fields for products and stores should be passed as paired Begin/End or Min/Max; multi-value formats for site ID, category ID, tags, and region are comma-separated. Natural-language conditions must not be arbitrarily converted into undefined fields.
  • Pagination and real-time behavior: page starts at 1, and the per-page limit varies by endpoint; each page turn is a new request. Data is subject to what the data service returns, and no guarantee can be made that it is real-time, complete, or permanently consistent with the platform backend.
  • Authentication and billing: WorkBuddy automatically injects the MCP Authorization; product and store queries dynamically consume credits based on the number of results returned on this page, while local category search does not consume credits, but syncing and MCP availability are still subject to environment limits. Before repeating, paginating, changing parameters, or retrying, first warn that consumption may increase.
  • Inapplicable scenarios: Does not handle Temu ordering, orders/inventory/logistics, ad placement, store operations, listing copy optimization, review sentiment analysis, supply chain procurement, image-based same-item search, trademark/patent/copyright judgments, or other non-Temu data query tasks.

Tool selection guide

Need / user says Default recommendation When to switch to another
Find Temu products, filter products by keyword, view price/sales/rating, or do product selection linkfox-temu-product-query When you need to study a seller's overall business scale, followers, or store product count, switch to linkfox-temu-store-query
Find Temu stores, filter stores by site/category/sales/sales revenue linkfox-temu-store-query When you need to view a specific product pool or filter by product tags or daily sales, switch to linkfox-temu-product-query
Find the category ID needed for product or store queries linkfox-temu-category-search If there are no local results, first use ehunt/temu/syncTemuCategory from the same sub-capability, then search again
Only know part of a category's Chinese name/English name/ID linkfox-temu-category-search Do not directly use the name as the category ID for product/store APIs; first locate the id/categoryId
Filter products by frontend category categoryHome of linkfox-temu-product-query When the backend category is known or store category filtering is needed, use categoryBackend or the store query's category respectively
Filter stores by backend category category of linkfox-temu-store-query When filtering by product frontend category is needed, switch to categoryHome of product query
Study fully managed and semi-managed products isLocal of product query, and add region for semi-managed When studying store operating model or store weekly/monthly sales, switch to isLocal of store query
Need large-scale results First set clear conditions, and preferably keep pageSize no greater than 50 Do not blindly paginate for "full volume"; first confirm the cost and whether the user needs more pages

Selection approach: first determine whether the entity the user wants to observe is a product, a store, or a category; when a category ID is needed, look up the category first. Both products and stores can express "high sales/high rating", so choose based on whether the user wants to evaluate single-product opportunities or the seller's overall operations. Prefer the path whose filter conditions fit better and which requests fewer pages; when equally satisfying, use the default recommendation in the table above. When paging, syncing, or repeated queries are involved, explain the data scope and credit impact first, then continue.

Business Requirement Routing Quick Reference

Sub-capability Business capability reference Exact toolName Local script
linkfox-temu-product-query Temu product multi-dimensional filtering, sorting, and pagination references/linkfox-temu-product-query.md ehunt/temu/productQuery None
linkfox-temu-store-query Temu store multi-dimensional filtering, sorting, and pagination references/linkfox-temu-store-query.md ehunt/temu/storeQuery None
linkfox-temu-category-search Temu category keyword/ID retrieval in the synced local database references/linkfox-temu-category-search.md ehunt/temu/temuCategorySearch None
linkfox-temu-category-search Temu category tree sync prerequisite operation references/linkfox-temu-category-search.md ehunt/temu/syncTemuCategory None

MCP invocation method

  • MCP server:LinkFox
  • URL:https://mcp-tool-gateway.linkfox.com/mcp/any-tool
  • Unique tool:proxy_call
  • Invocation structure:{"toolName":"<allowlist 中的精确值>","arguments":{...}}. Use the original request object directly as arguments, without adding body、params、data or another layer of arguments; parameterless sync must also pass {}。
  • allowlist:ehunt/temu/productQuery、ehunt/temu/storeQuery、ehunt/temu/temuCategorySearch、ehunt/temu/syncTemuCategory. Exact match and case-sensitive; do not substitute display names.
  • Authentication: WorkBuddy automatically injects Authorization. Configuration, documentation, and arguments must not include headers, API keys, tokens, JWTs, phone numbers, or other credentials.
  • Direct connection prohibited: do not use curl,urllib、requests、httpx , or other methods to connect directly to the tool gateway; if MCP or proxy_call is missing, stop; no HTTP fallback.
  • Response handling: first check the MCP connection, schema, and isError, then read text/structured content; when text is a JSON string, parse it only once, and finally check the downstream errcode、errorCode、code、errmsg or msg。
  • Output and deduplication: the new pipeline does not provide the old script's 24-hour cache, response persistence to disk, or stdout summary. Do not automatically retry on failure, empty results, or unsatisfactory returns; the same toolName + arguments must not be called repeatedly; retain parameter summaries, result identifiers, and credit field traces.
  • Data reading recommendations: after confirming total/productNum/storeNum or the categories count, read by field; do not put an entire large response into context just to view a single field. When displaying, preserve original values, units, dates, and the meaning of null values.
  • Full parameter location: for the parameters, responses, pagination, sorting, error behavior, and examples of product, store, and category retrieval and sync, see the corresponding reference; for authentication and credit guidance, see references/onboarding.md. Fields not defined in the reference must not be fabricated.

Usage examples

Product selection filtering

Use product queries to filter by price, rating, weekly sales, and fulfillment mode, suitable for building a candidate product pool:

{"searchKey":"kitchen organizer","categoryBackend":"123456","priceBegin":5,"priceEnd":30,"ratingBegin":4,"ratingEnd":5,"salesWeeklyBegin":100,"soldOut":0,"isLocal":0,"sortBy":"order_week-0","page":1,"pageSize":20}

Key output parameters: read productNum to learn the total number of matching items, and from products[] view productId, productName/productNameCn, logoUrl, price, orderTotal/orderWeek/orderDay/orderMonth, rating, reviewNum, publishTime, soldOut, isLocal, localRegion, storeId, tags, customTags, and category fields. Product queries are billed dynamically based on the number of items returned on this page (1.8 credits per product); if a next page is needed, you must first warn that it will consume credits again. categoryBackend should use a confirmed category ID; semi-managed filtering can add region.

Store performance filtering

Use store query to filter stores by US site, semi-managed, weekly sales, and rating:

{"siteId":"211","isLocal":"1","orderWeekMin":100,"orderWeekMax":5000,"ratingMin":4,"ratingMax":5,"productNumMin":10,"sortBy":"order_week_count-0","page":1,"pageSize":20}

Key output parameters: read storeNum to learn the total number of matching stores, and from stores[] view storeId, siteId, storeName, logoUrl, orderTotal/orderWeek/orderMonth, totalRevenue/weekRevenue/monthRevenue, rating, listedTime, reviewNum, followerNum, productNum, categoriesCn/categories, and isLocal. Store queries are billed dynamically based on the number of stores returned on this page (1.8 credits per store); monetary fields are in USD per the API convention, and pagination or changing conditions will generate new requests.

Category ID lookup and sync

First search the locally synced category library:

{"keyword":"kitchen","page":1,"pageSize":50}

Key output parameters: in categories[], id, categoryId, parentId, level, categoryName, categoryNameEn, isDeleted, hasChildren; use the confirmed id/categoryId for subsequent products' categoryHome/categoryBackend or stores' category. Local category search does not consume credits, and pageSize is at most 200. If the result is empty and the user needs to keep looking for categories, first call the sync operation with no input parameters:

{}

A successful sync usually returns result=success and totalRows; syncing changes the local category data state, so you should re-run the search after syncing rather than treating the empty result as a final conclusion. Whether the sync operation itself is available is subject to the current WorkBuddy tool metadata.

MCP invocation

Business calls use only proxy_call; for complete parameter examples, see the corresponding reference. The query costs for products and stores, category library state, authentication, and error handling follow the rules above.

Display rules

  1. First give the result type and scope: clearly state whether a product, store, or category list is being displayed, label the query conditions, page number/pageSize, number of returned items total, and productNum/storeNum (if provided in the response); do not write the number of items returned on this page as the total across the whole database.
  2. Preserve source metadata: if title, sourceType, sourceTool, or type is returned, preserve it or explain its meaning; Temu data is usually marked sourceType=temu, sourceTool=ehunt, and type may be tableListWorkbenches. Do not rewrite these fields as a third-party source that was not returned.
  3. Explain product fields: product results should preferentially display productId, Chinese/English name, main image, price, total/weekly/daily/monthly sales, rating, review count, listing time, listing status, managed mode, region, store ID, tags, and category; soldOut=0 means listed, soldOut=1 means delisted; isLocal=0 means fully managed, isLocal=1 means semi-managed.
  4. Explain store fields: store results should preferentially display storeId, site, store name, Logo, total/weekly/monthly sales, total/weekly/monthly revenue, rating, opening time, reviews, followers, product count, category, and managed mode; display revenue in USD per the returned field convention, and do not perform currency conversion yourself or round to an amount that was not declared.
  5. Explain category fields: Category results show id/categoryId, parentId, level, Chinese and English names, isDeleted, hasChildren; when isDeleted=1 mark as deleted and do not recommend direct use for subsequent filtering; when hasChildren=true indicate that further breakdown by category tree is still possible.
  6. Handle empty values and sync status: Empty arrays, missing fields, not synced, no match in local database, and no upstream data must be clearly marked; for empty category results, check sync status first. Do not fill empty values with estimates, common sense, or data from other platforms.
  7. Handle errors and cost: If the response has root-level code, errcode, errmsg, msg, display only what is actually returned and give next-step correction suggestions; do not confuse 401/402/403. When displaying product/shop results, also note the per-page billing rules and that continuing to page/retry may increase credits.
  8. Ensure traceability: Preserve toolName, parameter summary, result identifiers and credit fields; the new MCP link does not promise caching or response persistence to disk; when the user requests specific fields, read them from the current response and do not speculate based on the summary.

Resolve authentication and credit issues

When an exception occurs, read references/onboarding.md, and determine at each layer the MCP connection, proxy layer, and downstream business response. WorkBuddy automatically injects Authorization; do not require users to manually enter credentials or place credentials in arguments。

  • MCP or proxy_call missing: stop business calls, provide the MCP configuration and prompt to restart or reload WorkBuddy; the Skill cannot install or hot-load MCP by itself.
  • MCP connection 401: prompt to log in again and reload the connection; if it still fails, contact the WorkBuddy administrator.
  • Proxy succeeds but downstream JSON is 401: report automatic authorization or downstream business authentication failure; do not request a Key.
  • Downstream JSON is 402 or insufficient credits: guide the user to visit https://agent.linkfox.com/, and do not automatically select a plan, place an order, pay, or poll.
  • 403 and other errors: handle as-is according to the business reference; do not misclassify as 401/402, and do not automatically retry.

Credit consumption notes

Users pay fees due to credit consumption. Please fully assess: when this skill needs to be called frequently, or the user's expectation of credit consumption is insufficient, be sure to remind the user and let the user decide whether to continue.