ch-imagegen

name

ch-imagegen

description

Generate or edit raster images through a running local codex-helper or OpenAI-compatible relay, including text-to-image, image-to-image, multi-image edits, masks/inpainting, streaming image responses, batch generation, 2K/4K gpt-image-2 sizing, and sub2api/NewAPI-compatible JSON or multipart OpenAI Images API requests. Use when the user asks for ch-imagegen, codex-helper image generation, local proxy image generation, OpenAI Images API generation or edits, gpt-image-2 images, image-to-image, or generated bitmap assets through the user's relay/provider chain.

CH Imagegen

Use this skill for local image generation and image editing through a running codex-helper proxy or another OpenAI-compatible relay. It calls /v1/images/generations or /v1/images/edits, saves returned image bytes, and treats the script exit code plus stdout JSON as the source of truth.

Rules

• Do not use the system .system/imagegen workflow when the user explicitly asks for ch-imagegen, codex-helper image generation, or local proxy image generation.
• Require a running proxy or relay that exposes OpenAI-compatible Images API endpoints.
• Do not ask the user to paste provider API keys. Upstream credentials belong in codex-helper config, the relay, or local environment variables.
• Default model: gpt-image-2.
• Default resolution: 4k; default aspect ratio: 16:9; default output format: png; default quality: high.
• Save final outputs under output/imagegen/ unless the user specifies another directory.
• Never scan old output files to infer success. Validate only the files written by the current script invocation.

Quick Commands

Text-to-image:

python .agents/skills/ch-imagegen/scripts/image_gen.py generate \
  --prompt  \
  --aspect  \
  --resolution