AgentSkillsCN

imagegen-gpt

当用户要求生成新图像（横幅、帖子、原型图、插画），或希望将文字简报转化为可投入生产的图像提示，并通过 OpenAI gpt-image-1 进行渲染时，可使用此功能。切勿用于视频生成。也切勿用于单纯的纯文本头脑风暴。

SKILL.md
--- frontmatter
name: imagegen-gpt
description: Use when the user asks to generate a new image (banner, post, mockup, illustration) or to turn a textual brief into a production-ready image prompt and render it via OpenAI gpt-image-1. Do not use for video. Do not use for simple text-only brainstorming.

Goal

Generate a single high-quality image using OpenAI gpt-image-1 and save it into ./assets/generated/.

Inputs to collect (ask only if missing)

  • •Purpose: (e.g., Instagram feed 1:1, story 9:16, banner 16:9, icon 1:1)
  • •Style: (photorealistic, flat illustration, 3D, minimal, etc.)
  • •Text on image? If yes: exact text and language (warn about small text legibility)
  • •Brand constraints (colors, logos, layout rules)
  • •Output filename (default: image.png)

Output

  • •Generate the image file in ./assets/generated/<filename>.png
  • •Print the saved path and a short summary of the prompt used.

Execution steps

  1. •Draft a final image prompt (concise, unambiguous).
  2. •Run python3 .agents/skills/imagegen-gpt/scripts/generate_image.py --prompt "<PROMPT>" --out "./assets/generated/<filename>.png".
  3. •If the command fails due to missing API key, instruct the user to set OPENAI_API_KEY and re-run.
  4. •If the user requests multiple variants, run multiple times with different --seed values and distinct filenames.