feichanggege/ecommerce-visual-copywriting-skill

电商视觉文案设计SOP Skill - 让电商设计小白也能一键做出5年资深设计经验的商品图片.兼容 Claude / Codex / Cursor / Workbuddy 等 SKILL.md Agent. AI-powered e-commerce visual strategy skill for main images, PDP/A+, copywriting, localization, evidence-aware compliance and image prompts.

📦 What the project is

ecommerce‑visual‑copywriting‑skill is a Skill (a plug‑in) for AI‑agent platforms (e.g., Codex, OpenAI‑based agents). It lets a user feed product data, packaging images, and competitor references and receive a complete, AI‑generated visual plan for an e‑commerce listing. The output includes:

A storyboard that explains the buying‑decision logic for each image.
Per‑image design cards with layout description, on‑image copy, design notes, a text‑to‑image prompt, and a negative prompt that blocks prohibited changes (logo distortion, false claims, etc.).
An “evidence ledger” that tracks which selling points are verified, need proof, or are disallowed.
Platform‑specific compliance checks for Amazon, Shopify, TikTok Shop, etc.

The skill can work in four modes: standard (strategy → storyboard → confirm), one‑shot delivery, audit/re‑write of existing pages, and hand‑off to a designer.

🚀 How you use it

Install with the Codex‑style command:

npx skills add feichanggege/ecommerce-visual-copywriting-skill

Ask the agent in natural language, e.g.:

这是我的产品资料、包装图和竞品参考。面向 Amazon US，直接给我 7 张商品图 + A+ 视觉规划：每张包含画面、英文图内文案、设计说明、生图 Prompt 和 Negative Prompt。不要中途确认，缺失信息请标注假设，不要编造。

The agent runs the skill’s workflow (evidence ledger → driving‑force analysis → style lock → storyboard → per‑image cards) and returns a structured JSON/YAML that can be handed directly to a designer or a text‑to‑image model.

🛠 Core technical ideas (from the README)

Feature	What it does
Evidence Ledger	Tags each claim as confirmed, needs proof, prohibited, etc., preventing false advertising.
Reference Fidelity	Locks immutable elements (logo, packaging shape, specs) while allowing creative background changes.
Negative Prompt	Generates a list of things the image model must avoid (e.g., logo distortion, wrong dosage).
Dynamic image count	No forced “5‑image” rule; the number of visuals adapts to platform and product.
Multi‑platform playbooks	Built‑in style and compliance rules for Chinese marketplaces, Amazon, Shopify, TikTok Shop, Temu, Shopee, Lazada, etc.
Cross‑border localisation	Rewrites copy for target markets instead of literal translation, adjusting units, scenarios, and tone.
Quality gates	Pass/Fail checks replace vague scoring, ensuring every card meets a five‑dimensional quality checklist before delivery.

📂 Repository layout (quick map)

SKILL.md – entry point that the agent runtime reads.
agents/openai.yaml – UI metadata for the OpenAI‑based chat UI.
references/ – compliance rules, platform playbooks, and output contract templates.
examples/ – sample inputs and expected outputs.
assets/ – screenshots showing the visual cards the skill produces.
tools/verify-skill.py – a helper script that validates the skill’s structure and privacy before publishing.

🔒 Safety & compliance

The skill explicitly does not fabricate certifications, sales numbers, or user testimonials. It refuses to turn “pending evidence” into a factual claim and will not breach medical, food, or platform‑specific advertising rules. All compliance logic lives in references/compliance-rules.md and is consulted during generation.

🎯 Who would benefit?

E‑commerce sellers who need a fast, AI‑assisted way to create Amazon or Shopify listing images without hiring a full‑time designer.
Design agencies that want a structured hand‑off from AI to human designers.
Compliance teams looking for a guard‑rail that flags unsupported claims before images go live.

📜 License

MIT – you can use, modify, and redistribute the skill freely.