Ecommerce Visual Copywriting Skill

Overview

Ecommerce Visual Copywriting Skill (by feichanggege, ~475 stars, MIT) is an open-source AI Agent skill that turns a raw product brief into a designer-ready e-commerce visual copywriting deliverable. It's a gated SOP: collect product info and qualification boundaries → refine verifiable "must-buy reasons" for user confirmation → then generate 5 main images + detail-page modules. Every image ships exactly three things — scene description, in-image copy, and design notes — plus a compliance self-review score that must reach 80+ before output. It's built and hardened against a hundred-plus live product listings across categories, and works on Claude Code, Codex, Cursor, Workbuddy and other SKILL.md-based agents.

Key Features

• 6-step gated workflow: collect information → refine compliant selling reasons (pause for user confirmation) → main-image storyboard (pause) → generate 5 main images → detail-page modules → five-part self-review → final execution plan.
• Main images (click-through oriented): 5 KVs — hero selling point (product fills 60–80%), pain point/problem, difference comparison, scenario immersion, action guidance — each with scene, in-image copy (max 5 lines, ≤6 chars/line) and design notes.
• Detail-page narrative (conversion oriented): modules from M1 first-screen hook → core benefits → material/craft → scenarios → trust evidence → comparison → FAQ → CTA/disclaimer, each capped at 6 lines.
• Category-based compliance rule library (references/compliance-rules.md): layered red lines — generic (no absolute claims, data must have evidence, no competitor bashing), 蓝帽子 health foods (only claim approved functions), sports equipment (de-medicalized), and strictest for ordinary food (no function/health/treatment implications at all).
• Five-dimensional independent self-review: compliance, benefit translation (Feature→Benefit), visual focal point, tactile/medium expression, narrative coherence — each scored 0–100; any dimension below 80 must be rewritten (no score averaging to pass).
• Campaign Style Lock: locks ≤3 colors, product proportion, lighting system, typography and no-drift rules across every image.
• No fabricated evidence: missing qualifications/reports/patent numbers are marked "missing" and never stated as fact; unverified selling points become "explorable directions" only.
• Zero-dependency verification script (tools/verify-skill.py): checks frontmatter, triggers, pause gates, self-review scoring, README quality and secret leaks.

How to Use

1. Use a Skill-compatible agent such as Claude Code, Cursor, or Codex.
2. Install from the terminal (Node.js required):

npx skills add https://github.com/feichanggege/ecommerce-visual-copywriting-skill

Or paste the install prompt from this tool page into your agent and ask it to install for you. 3. Restart / reload the agent so the new Skill is loaded. 4. Invoke the Skill with its slash commands or by asking the agent to run the workflow described in Overview / Key Features. 5. Review outputs before publishing — treat model-generated copy, images, and data as drafts.

Pricing

• Free & open source: MIT-licensed on GitHub; install via npx skills add feichanggege/ecommerce-visual-copywriting-skill or git clone. It's a pure workflow/prompt skill with no API dependencies — outputs copy and design briefs for your designer (or your own image tooling) to execute.

Best For

E-commerce sellers, marketers and design beginners (on Taobao, Tmall, JD, Pinduoduo, Douyin Shop, or Amazon/Shopify) who need compliance-safe, designer-executable main-image and detail-page copy quickly — and want a repeatable SOP that passes platform/advertising-law review instead of raw ad copy.

FAQ

What does the Skill output?

For each image: scene description, in-image copy (with line limits), design notes (visual focal point, eye flow, color psychology, disclaimer placement) and, in the full workflow, image-generation prompts. Plus a five-part compliance self-review score.

Why is there a "pause for confirmation" step?

Step 2 (must-buy reasons) and Step 3 (storyboard) are gates — direction is confirmed with the user before expensive execution, so wrong directions are caught before final artwork.

How strict is the compliance check?

Very. Ordinary food can't imply any health or function benefit, health foods can only claim the exact functions on their approval, and any claim needs evidence. Any dimension scoring below 80 is rewritten; fabricated reports/patents/sales/reviews are forbidden.

Does it generate images?

Not directly — it produces scene descriptions, in-image copy, design notes and prompts. Rendering happens with your own image-generation tooling or a designer.

Which platforms does it cover?

Taobao, Tmall, JD, Pinduoduo, Douyin Shop, live-room product images, plus cross-border channels such as Amazon and Shopify PDP.