name: image-agent-plus description: "Use image-agent-plus CLI to generate images, prioritizing image2 for high-quality text and prompt-following, with Codex/Gemini fallback." version: 2.0.5 license: Apache-2.0 homepage: https://github.com/webkubor/image-agent-plus store: https://clawhub.ai/webkubor/image-agent-plus author: webkubor compatibility: platforms: - codex - openclaw - hermes metadata: openclaw: emoji: "🎨" requires: bins: ["node"] install: - id: node-brew kind: brew formula: node bins: ["node"] label: "Install Node.js (brew)"

image-agent-plus

Direct CLI invocation. No HTTP server required.

Showcase

The image above was generated locally from this prompt:

AI 時代，我們應該沉澱下來的是什麼，中文社論海報

It is included as a ClawHub store preview so users can see the intended output style before installing the skill.

What Went Wrong During Setup

We first tried to treat Codex image generation like a normal nested CLI provider. That was wrong.

The working Codex flow is:

小蔥技能站7w4.net每天更新，海量AI技能等你發現。

1. Generate the image inside the active Codex session.
2. Codex saves the result under ~/.codex/generated_images/<session>/ig_*.png.
3. Run image-agent-plus collect-codex to copy the latest generated image into your chosen output path.

Do not rely on nested codex exec for image generation. It can hang or fail because the interactive Codex image tool event is not the same as a normal non-interactive shell command.

Runtime Rule

Before asking for an API key, run:

image-agent-plus check

If an OpenAI API key (IMAGE_AGENT_OPENAI_API_KEY or OPENAI_API_KEY) is set, the tool defaults to image2. If Codex CLI or Gemini CLI is installed (no OpenAI key), use those without API keys. API keys are only for direct provider API mode or for machines without local runtimes.

Quick Start

image-agent-plus generate \
  --prompt "一隻橘貓坐在雨天窗臺上，電影感光線" \
  --aspect-ratio "16:9"

Common Commands

image-agent-plus check

# Default: image2 via OpenAI (when API key available)
image-agent-plus generate \
  --prompt "AI 時代，我們應該沉澱下來的是什麼，中文社論海報" \
  --aspect-ratio "16:9"

# Fallback: Codex CLI (no API key needed)
image-agent-plus generate \
  --prompt "AI 時代，我們應該沉澱下來的是什麼，中文社論海報" \
  --provider codex \
  --aspect-ratio "16:9"

For Codex image generation, use the active Codex image tool first. After Codex reports Generated Image and saves under ~/.codex/generated_images, run:

image-agent-plus collect-codex

Options

Codex Collection

image-agent-plus collect-codex --filename ~/Desktop/output.png

If --filename is omitted, the copied image goes to ~/Desktop/image-agent-plus-output/.

Environment

Prefer local runtime. For image2 direct API mode, configure the OpenAI key in your normal shell or agent secret manager before running the CLI.

Legacy NANOBANANA_* variables are accepted as fallback reads by the CLI.

這是一款功能完整的影像生成工具，支援多種 AI 模型（OpenAI、Codex、Gemini）生成圖片。文件寫得非常詳細，中文示例清晰，還專門說明了常見使用誤區，對新手友好。配置規範，支援多個平臺。不足之處是實際程式碼未包含在包中，需依賴外部工具。總體質量良好，適合需要本地影像生成能力的使用者。