Skills

Turns one or more product photos into Seedance 2.0 video prompts structured as three layers of shots. Prompts only; the video is made in Seedance.

Turns product facts, certificates, target platform and audience into storyboards for main images, detail pages and A+, with on-image copy, image prompts and a rules check.

Teaches the agent to receive Shopify webhooks (order, product, customer events) and verify the HMAC-SHA256 signature, with tested Express, Next.js and FastAPI examples.

Searches 1688 wholesale offers for tiered price, MOQ, reorder rate and volume, separates factories from traders, and stores snapshots locally to track price changes and compare regions. Read-only. Runs through the author's 1688-pp-cli tool.

Average eBay sold prices over 90 days with outliers trimmed, and auctions filtered by bid count and ending time. Runs through the author's ebay-pp-cli tool.

Covers creating the developer app, OAuth, category and required attributes, uploading your own photos, publishing and editing, all through the official API. Captchas, authorisation, final confirmation and paid tiers stay with you.

Answers Google Merchant API questions and helps migrate old Content API code. Looks up Google's official docs and code samples before answering.

Fetches public sources such as Amazon release notes, Helium 10, AMZ123, WeAreSellers, podcasts and YouTube, and has the agent write a Chinese or bilingual digest with a link on every item. Can be scheduled to Telegram, email or Feishu.

Shopify's own developer skill: Shopify CLI, Admin API, Storefront GraphQL, ShopifyQL, themes and app extensions. For developers.

Reads pixels to tell whether an image will be rejected on Amazon, TikTok Shop, Temu or Shopee: white purity, coverage, resolution, ratio, alpha, borders and file size, with a one-step fix to a compliant JPEG.

Opens your store in a fresh browser session, adds a product, clicks Checkout and records where the shopper lands and which third-party scripts load, to catch injected code sending buyers to a phishing page. Stops at checkout and fills in nothing.

For a few keywords, collects organic ASINs, titles, prices, ratings, review counts and main images from the first three Amazon result pages, deduplicates across keywords and writes one Excel file.

First shows the text an AI crawler actually gets from a product page, then reads the detail images with the agent's vision to find facts trapped in pictures. The verdict is ready, needs_fixes or analyzing.

Ask plain questions of CSV / XLSX exports from Shopify, Stripe or ad platforms. Profiles each sheet first (total rows, duplicates, currency, time zone), computes in code and states the definition used.

Locates the leaking funnel step and segment from data, scores desktop and mobile screenshots against a yes/no checklist, and combines them with customer feedback into hypotheses with evidence and confidence, sorted into fix, test or research.

The model tags each review (eight complaint causes, praised selling points); a script counts, checks totals and ranks fixable rate and selling points, then you get ready-to-paste Portuguese fixes for the product page.

15 / 30-second Portuguese scripts, shot lists, voice-over and subtitles for Shopee Brazil and Mercado Livre, with a checker for timing and line length.

From landed cost and platform fees, works out net margin, break-even and target prices, price ladders and coupon headroom, with fee presets for Shopee Brazil, Mercado Livre, Amazon Brazil and TikTok Shop Brazil, or your own rates.

From a product description or competitor names, collects real complaints from Reddit, Hacker News, GitHub, G2 and Google Trends, scores and clusters them into a positioning frame. Quotes must come from collected text.

Fetches the Amazon product page and search results as a shopper would, checking availability, price, Buy Box seller, cart buttons, delivery, rank and negative reviews, and whether your keywords surface the ASIN. 10 marketplaces.

Reads SP / SB / SD search term reports; scripts aggregate 7 / 14 / 30-day windows and cluster terms by root, the agent labels each root, then each term gets a negate, cut, keep-testing or scale call, with a negative list and a filterable HTML console.

Before launch, checks the live product page and lists unrelated or mismatched shopping intents as a five-column negative keyword sheet, then a script checks no core product term was negated. Can also review an existing search term report.

Scans titles, bullets, descriptions, image copy and backend keywords across four areas: IP, advertising claims, platform policy and destination-market rules, with a risk level, fix and compliant rewrite per item plus an HTML report.

Turns a snapshot or supplier photo into a pure-white main image: locks the product's look, writes the prompt or calls an image API, then a script checks white purity, coverage and size and can fix them.

Identifies items from photos, checks sold comps, decides single vs lot, drafts and publishes through the eBay Sell API, then tracks orders and adds tracking. Every publish waits for your approval by default.

Re-encodes one master video to the recommended ratio and codec for TikTok Shop, Amazon, Shopee and Chinese platforms, with blurred fill or crop, a spec report and a UI safe-zone preview.

Translates product specs into feature, advantage and benefit copy for titles, bullets, description, A+ and search terms; lists questions when facts are missing and runs a script for length limits and banned wording.

Builds a keyword pool from competitor or own ASINs (or your own list), judges standard vs non-standard product, sorts terms into six root types and marks each as attack, intercept or negate. Outputs Excel, Markdown and JSON.

Main, scene and set images with consistency control and platform size checks. Generates through an OpenAI-compatible endpoint you configure, or exports prompts only.

Tags a batch of product reviews (CSV first, SellerSprite optional) on 22 dimensions; the agent writes a 15-chapter insight report while scripts handle stats and five anomaly checks. Optional HTML dashboard or Feishu sync.

Normalises headers (Chinese and English aliases), dates and amounts across platform, ERP, ad and 3PL bill exports, dedupes, checks required fields and quarantines bad rows instead of dropping them.

Computes ROAS, ACOS, TACOS, CTR, CVR, gross and net profit from ad and sales exports, grouped by campaign, SKU or marketplace with period-over-period change, written up as a weekly review.

Line-by-line net profit per product, marketplace and channel after commission, payment, shipping, duty and ad costs, with break-even and target prices, flagging loss-making rows and missing inputs.

Turns product photos, specs and old listings into sourced product facts, then writes Amazon / eBay / Temu copy (English, German, Japanese) and briefs for 7+ images. Inferred specs are marked unconfirmed and kept out of copy.

Generates a full set of main, benefit, scene and model images, with size rules for Chinese platforms, Amazon and independent stores. Works with several image providers.

From size, weight, price and cost, works out the size tier, FBA fulfilment and storage fees, referral fee and margins. Runs locally, no network.

Finds unauthorised sellers and likely counterfeits, watches MAP pricing, flags trademark misuse in titles, with VeRO complaint templates and steps. No API key needed.

Collects sold listings on 8 eBay sites by keyword and filters (category, price, condition, location): sold price, date, best-offer-accepted, bid count, shipping and seller, as structured records.

Crawls a Shopify store's public pages and checks identity, policies, price consistency and structured data against Google Merchant Center's misrepresentation policy, producing an HTML report with evidence and a manual checklist.

Scans any public Shopify store and identifies its theme and installed apps from page code, script URLs and globals, each with evidence and a confidence level, in an HTML report.

From one product card, makes white-background, showcase, benefit, hero, detail and lifestyle images. Uses the agent's built-in image tool first and asks before falling back to a paid API.

From one store URL, crawls home, product, collection and blog pages, scores AI search visibility, keyword distribution and how well pages answer buyer questions on a 100-point card, and gives prioritised fixes and a content plan.

From a market and category, collects Amazon competitors and negative reviews, 1688 prices, TikTok, Reddit and Google Trends data into a sourced Excel workbook, then a 16-chapter HTML report. Data it can't collect is left out.

Finds live products missing FAQ, perfect-for or story metafields: colour variants copy the base product, new products get agent drafts that are written back only after your review. Policy claims and links must come from store pages read in that run.

Connects to your own logged-in Chrome to collect one ASIN's listing copy, specs, image screenshots and as many reviews as it can reach, with per-review provenance and an audit file stating counts and stop reasons.

Pulls every order from Klaviyo with a private API key, scores each customer on recency, frequency and monetary value into eight segments, measures the gap between the top 10% and the average, and writes a single-file HTML report.

Plans 6-10 detail-page modules from product photos and a short brief: separates supplied, visible and unconfirmed facts, has the agent make text-free bases, sets exact type with Pillow, slices to each platform's size and validates.

Scans a merchant site for a /.well-known/ucp profile, platform, product structured data, payment providers and a public product API, scores it out of 100 and lists the gaps.

Runs up to 39 analyses on an orders CSV from Shopify, WooCommerce or Tiendanube, including RFM, repeat rate, basket pairs, lifetime value, churn and seasonality, with JSON and HTML reports.

A four-phase audit of Klaviyo flows, segments, campaigns, deliverability and attribution that finds missing or weak flows. Built for Klaviyo's official MCP in read-only mode, with read-only local scripts as a fallback.

Pulls Seller Central reports from a plain request: orders, returns, removals, inventory ledger, sales and traffic, settlements and other GET_* types. Resolves phrases like "last month", saves CSV or tab-delimited files and logs each pull.