---
name: ecommerce-detail-images
description: 'Generate ecommerce detail images from product photos and confirmed product notes. Use when the user asks to create Taobao, Tmall, 1688, Amazon, Shopee, Temu, or marketplace-ready close-up detail graphics showing texture, structure, hardware, seams, components, craftsmanship, packaging details, or product feature closeups with GPT Image 2.'
metadata:
  icon: "\U0001F50D"
  title: Ecommerce Detail Images
  companionSkill: gpt-image-2
  packageName: '@zjxuyunshi/ecommerce-detail-images'
  version: 0.0.1
---

# Ecommerce Detail Images

## When to Use

Use this skill when the user wants product detail graphics from real product photos. Typical requests include close-up ecommerce images for texture, stitching, hardware, seams, edges, buttons, zippers, handles, labels, packaging details, components, internal structure, material appearance, or craftsmanship.

Use this for detail-page modules and main-image detail variants. Use `ecommerce-size-images` for measurements, `ecommerce-main-images` for covers, and `ecommerce-selling-point-images` for broad benefit callouts.

## Inputs

- Required: at least one original product image. Preserve product identity, color, shape, logo placement, material appearance, packaging, and distinctive details.
- Optional: requested detail focus, such as `五金`, `走线`, `面料纹理`, `拉链`, `包底`, `内里`, `接口`, `瓶口`, `按键`, or `包装细节`. If absent, choose visible low-risk detail points.
- Optional: confirmed product notes. Use only supplied facts for material, process, waterproofing, load-bearing, durability, safety, or certification claims.
- Optional: reference images. Use only for close-up composition, callout layout, crop style, lighting, background, or typography.
- Optional: platform, target language, output size, format, quality, and variation count. Default to Chinese, PNG, high quality, and square or vertical depending on the requested module.

Before prompting, separate `confirmed`, `visible`, `missing`, and `do not claim`. Do not invent material names, process names, certifications, performance claims, or test data.

## Execution

Use `$gpt-image-2` image editing mode through `fusion-api.openai_image_edit_async_submit`. This skill supplies the ecommerce detail-image strategy and prompt; `$gpt-image-2` supplies execution. Do not run new oo capability discovery during normal use.

Upload local images with `oo file upload "<path>" --json`, then pass returned `downloadUrl` values as:

```json
{"image_url":"<downloadUrl>"}
```

Put the product image first, then optional references.

Prompt structure:

```text
Create a marketplace-ready ecommerce product detail image.
Primary product: preserve image 1 accurately, including color, shape, material appearance, logos, labels, and distinctive physical details.
Detail focus: <requested or visible detail points>.
Confirmed facts to display: <only user-supplied facts>.
Visible facts: <safe visual observations>.
Reference usage: use images 2+ only for close-up layout, lighting, background, typography, or callout style.
Composition: close-up detail crop plus product context, mobile-readable labels, clean callout lines, professional ecommerce hierarchy.
Avoid: invented material/process/certification/performance claims, fake test data, QR codes, phone numbers, platform badges, watermarks, distorted product details, clutter, unreadable text.
```

Run with `$gpt-image-2`:

```bash
oo connector run "fusion-api" --action "openai_image_edit_async_submit" --data @payload.json --json
```

Use exact fields: `model: "gpt-image-2"`, `prompt`, `images`, `output_format`, `quality`, `size`, and optional `n`.

## Result Handling

Read result URLs from `.data.data[].url`. Download each image with `oo file download "<url>" "<output-dir>" --name "<name>" --ext "<ext>"`.

Use `/Users/yunshi/Downloads/ecommerce-detail-images/<short-product-name>-<timestamp>/` and names like `detail-image-01.png`. Preview the image when practical and mention the detail focus plus confirmed facts used.

## Failure Handling

- Missing product image: ask for one.
- Missing proof for material/process claims: omit the claim or ask for confirmation.
- Requested detail is not visible: ask for a clearer detail photo or create a conservative layout based on visible areas.
- Too many references: select the most relevant layout/style references.
- Connector handle, timeout, auth, billing, upload, or schema failure: follow `$gpt-image-2` guidance and report the exact blocker.