
from gpt-image2-ecommerce126
Generate commercial e‑commerce and promotional images using GPT-Image-2 via Codex CLI with scene-specific templates for hero, lifestyle, flat-lay, UGC, and more
This skill turns a user's e-commerce visual brief into concise, structured JSON prompts for GPT-Image-2 and runs generation via the Codex CLI. It matches request intent to one of 25 scene templates (hero, lifestyle, flat lay, detail macro, poster, social media, UGC, model shots, packaging, storefront, etc.), fills template variables from product info and style preferences, and outputs a simplified prompt ready for image generation.
Use this skill whenever you need product photography or marketing assets generated quickly: main product images, lifestyle scenes, social media banners, UGC-style shots, packaging mockups, exploded views, multi-angle grids, or seasonal campaign imagery. It is especially helpful when you have product details and optionally a reference image to preserve visual identity.
Designed for Claude Code / Codex CLI driven image-generation workflows and agents that can call bash or the Codex runtime.
E-commerce image generation skill using GPT-Image-2 via Codex CLI with 25 scene-specific JSON templates. The bash script is well-structured with proper error handling, argument parsing, and supports CLI and HTTP modes. Script failed only because it requires a --prompt argument which isn't supplied in dry-run testing — expected behavior. No security concerns beyond minor shell interpolation in prompt text.
Well-documented skill with clear progressive disclosure (only load matched template). Anti-AI tips for UGC/livestream scenes are a nice touch. Shell script uses set -euo pipefail and jq for safe JSON construction in HTTP mode. The rm -rf cleanup step in SKILL.md is scoped to a specific session directory, not destructive.