E-commerce Visual Copywriting skill
ecommerce-visual-copywriting
A skill that turns product data into a plan for main images, the product page and A+ content, with on-image copy, generation prompts and a claims check
- 980 ★ on the repository
- Last commit
- MIT
Minimal access
Why this level
- Instructions only: the agent reads the product materials you give it and writes plans and copy, and the skill runs nothing
- To check platform rules the agent may visit official sites if it has network access
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/ecommerce-visual-copywritingDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewMinimal accessThe tool runs nothing and makes no network calls.
- Instructions only: the agent reads the product materials you give it and writes plans and copy, and the skill runs nothing
- To check platform rules the agent may visit official sites if it has network access
This is third-party code. Review the repository files before installing.
What it does
The skill walks the agent from product materials to ready briefs for a designer or an image model. First it determines the task type: main images, a product page, an Amazon listing and A+ content, an ad creative, on-image copy, a storyboard, a review of an existing page or localization for another market. It asks no more than three questions and takes the rest as clearly marked assumptions. Next the agent builds an evidence ledger: every claim gets a status of confirmed, supported, needs proof or disallowed, and only the first two reach the final text. Then it picks the purchase driver (look, pain and convenience, specs, trust, emotion), turns the top three reasons to buy into a chain of feature, advantage, benefit and evidence, and locks one style for the series: product identity, composition, color, light, type, background and what must not change relative to reference images. The output is a storyboard whose frame count depends on the platform, and a card for each frame: task, scene description, on-image copy, a note for the designer, a generation prompt, a negative prompt and evidence notes. Before delivery a five-axis check runs with hard blockers: invented certificates, a changed package look or unchecked platform rules stop delivery.
Who it is for. For e-commerce sellers and marketers who prepare images and product pages for Taobao, Amazon, Shopify, TikTok Shop and other platforms and work with Claude Code, Codex or Cursor.
Good fit when
- You need designer briefs for main images and a product page based on packaging, specs and competitor links
- You are preparing an Amazon listing and A+ content and want the agent to produce on-image copy and generation prompts
- You want an existing page reviewed with minimal fixes instead of a rebuild from scratch
- A Chinese product page must be rewritten for Shopify or Amazon for another market instead of translated word for word
- Image generation must not distort the packaging, logo and specs shown in a reference photo
Not a fit when
- You need finished pictures: the skill produces briefs and prompts and does not draw images
- You sell on Wildberries or Ozon: they are not among the skill's platforms, so it does not know their rules
- You need a legal opinion on claims for dietary supplements or medical devices: the skill gives a conservative frame, not a final conclusion
Example request
Here are the product data and a photo of the packaging. I need an Amazon US listing: seven images and A+ content with English on-image copy, prompts and negative prompts, no invented certificatesLimitations
- The main `SKILL.md` is written in Chinese and the agent answers in Chinese by default. The English mirror is in `SKILL.en.md`, and English or bilingual output is turned on by the user's request.
- The skill does not check platform rules itself: it tells the agent to verify sizes, banned words and category requirements against official sources, and marks an item as unchecked when that is impossible.
- The named platforms are Taobao, Tmall, JD, Pinduoduo, Douyin, Amazon, Shopify, TikTok Shop, Temu, Shopee and Lazada, and Russian marketplaces are not on the list.
- The skill does not invent certificates, sales, reviews or rankings: where materials are missing, items go into assumptions or into needs proof.
- The sample results in the repository illustrate the format and do not guarantee that a platform will accept the page.
How to disable. Delete the skill folder from the agent's skills directory, for example `~/.codex/skills/ecommerce-visual-copywriting`. If you installed it with `npx skills add`, remove the skill with the same tool.
FAQ
How does the skill differ from asking for product page copy?
It first builds an evidence ledger and picks the purchase driver, and then hands out a task, scene, copy, prompts and prohibitions per frame. Each frame answers one buyer question instead of repeating the same strengths.
What is the evidence ledger?
Each claim gets one of four statuses: confirmed, supported, needs proof or disallowed. Only the first two go into the final page, and the rest is replaced with neutral wording or marked as waiting for proof.
How many images does the skill plan?
Five frames are a starting point, not a rule. The number depends on the platform and the amount of material, and frames are not padded just to reach a count.
How does the skill keep the packaging looking right?
It locks one style for the series and a list of what must not change relative to the reference photo: logo, color, structure and specs. The negative prompt forbids packaging distortion, logo errors and garbled text in the picture ahead of time.
Can I use the skill on an existing page?
Yes, there is a review and revision mode: the agent finds problems in the selling logic, visual hierarchy, evidence and style and makes the minimum necessary changes instead of rebuilding.
Can the skill make an English page from a Chinese one?
Yes, localization is a separate scenario: it keeps the boundaries of fact but rewrites the reasons to buy, units, scenes and information density for the market.
Discussion
No questions or reviews yet.
Related
SkillsSkill with SaaS product-design principles: conversion, onboarding, retention, pricing and behavioral tactics
Marketing Skills for AI agents
Marketing Skills for AI Agents
About fifty skills for CRO, copywriting, SEO, analytics, ads and launches, built for coding agents
A skill of the author's marketing principles and playbooks that the agent reads before writing copy, emails and funnels
Yandex Direct combinatorial creatives skill
ya-direct-combinatorial-creatives
A Codex skill prepares and audits image systems for Yandex Direct combinatorial ads: crops, manifest and compatibility checks
SEO article writer skill with images
SEO Article Writer
A 10-stage Claude Code pipeline: Wordstat semantics, top-5 competitor analysis, AI cover and illustrations, publishing to WordPress, Dzen and Prisma Blog
Infostyle
infostyle
A Russian copy-editing skill based on Ilyakhov's information style: cuts filler and cliches, adds facts, shows before and after
README in short
The repository holds the main skill file in Chinese and its English mirror, three reference files (evidence and compliance rules, platform playbooks, output formats), a ChatGPT interface description in `agents/openai.yaml`, request examples, a `.claude-plugin/marketplace.json` manifest, the structure check script `tools/verify-skill.py` and pictures with sample finished frames. The main page shows the workflow, the modes, the platform list, the safety boundaries and three ways to install.
README badge
Original SKILL.md
--- name: ecommerce-visual-copywriting description: 将商品资料、包装与资质、目标平台、目标人群和视觉参考转成可直接执行的电商视觉方案，包括转化策略、Campaign Style Lock、主图/详情页/Listing/A+ Storyboard、图内文案、设计说明、生图 Prompt、证据与合规审查。适用于淘宝、天猫、京东、拼多多、抖音小店，以及 Amazon、Shopify、TikTok Shop、Temu、Shopee、Lazada 等电商场景；当用户提到主图、详情页、Listing、A+、商品卖点、视觉脚本、图片规划、图内文案、生图 Prompt、点击率/转化率、平台合规、广告法审查、设计师执行稿或跨境商品本地化时使用。不得编造资质、数据、销量、评价、认证或功效；涉及当前平台规则或高风险合规结论时优先核验官方最新来源。 --- # 电商视觉策划与执行 把“原始商品资料”转成“可直接交给设计师或生图模型执行”的视觉方案。默认中文；用户要求英文或双语时切换语言，但不改变证据边界和合规边界。 ## 运行原则 1. **证据优先**：用户提供的包装、参数、资质、检测、品牌资料和明确事实 > 已核验的官方最新平台规则 > 本 Skill 的参考资料 > 合理假设。 2. **按任务路由**：只交付用户需要的层级，不强制走完整流程；不要为了“流程完整”重复追问已知信息。 3. **不把假设写成事实**：缺资料时标记“待确认”，不要补造销量、评价、证书、检测、专利、资质、功效、排名或平台规则。 4. **先统一视觉，再写细节**：同一套图必须先锁定产品身份、构图、色彩、光影、字体、背景和禁止漂移项。 5. **平台规则有时效性**：涉及当前尺寸、禁限词、类目政策、主图要求、资质或审核规则时，优先查官方最新来源；无法核验时明确标记“未核验”，不要凭旧规则下结论。 6. **高风险合规不做最终法律结论**：食品、保健品、医疗器械、健康护理、儿童用品等场景，按保守口径处理，并说明需要以目标市场法律、监管和平台最新规则为准。 ## 执行模式 根据用户意图自动选择，不要机械暂停。 ### A. 标准模式 适用于方向不清、视觉成本高、用户明确要求“先策划/先分镜/确认后再做”。 1. 先输出《视觉策略与证据账本》。 2. 再输出 Storyboard。 3. 在真正进入大量最终文案或生图 Prompt 前等待确认。 ### B. 一次性交付模式 用户明确说“直接出完整方案”“一次性给我”“按你判断做完”，或资料已经足够明确时，直接完成策略、Storyboard、执行稿和质量门，不做不必要的中途确认。 ### C. 审查/改稿模式 用户已经有主图、详情页、文案、Prompt 或设计稿时，不从零重做。先找出问题，再给最小必要修改： - 转化逻辑问题 - 视觉层级问题 - 证据或合规问题 - 平台适配问题 - 文案与画面不一致 - 风格漂移