E-commerce visuals by agent, not by agency

What it does

ecom-details-image is an agent skill for Claude Code, Codex, and similar tools that generates e-commerce visuals via GPT-Image-2. Feed it a product photo and a brief; it returns either a structured creative brief with ready-to-use prompts or the actual images if an API key is configured. The skill maps requests to 25 scene templates—white-background heroes, lifestyle flat lays, livestream backdrops, exploded views, and more.

The interesting bit

The “Campaign Style Lock” is the clever part: on multi-image jobs, the skill automatically pins down a shared color palette, lighting scheme, and layout rules so the output looks like a coordinated shoot rather than a random prompt grab bag. It also claims to diagnose whether a product sells on visual appeal, pain-point relief, or emotional value, then tailors the image sequence accordingly—more creative strategy than you typically get from an API wrapper.

Key highlights

• 25 built-in scene templates covering Amazon PDP layouts, UGC-style social posts, virtual try-ons, and seasonal campaigns.
• Pure Python standard library for the generation script; no pip install required.
• Dual-mode operation: output prompts for manual use, or call an OpenAI-compatible image API directly to render images.
• Reference-image support to keep generated products faithful to the actual item.
• A separate web UI repository exists if you prefer a browser to agent chat.

Caveats

• The skill still requires an OpenAI-compatible image API for generate mode and an agent environment (Claude Code, Codex, etc.) to orchestrate the workflow.
• The “conversion driver diagnosis” is mentioned as a feature, but the README never explains how it classifies products or whether it uses hardcoded rules versus LLM reasoning.
• This is fundamentally a curated prompt engine and thin API wrapper; the real work is still done by whatever image model you plug in.

Verdict

Useful if you run a small e-commerce shop or build agent workflows for sellers and need consistent product imagery without a creative team. Skip it if you want a standalone design tool—this is an agent skill, not a self-contained app, and it bills your API key for every render.