For AI Agents
Agent Access
An entry point for AI agents, automation scripts, and prompt research tools. Use it to discover the site guide, full prompt dataset, and category pages.
Agent Index
/agent/index.json
Search all case titles, categories, and page numbers in a compact JSON file, then fetch only the matching full-prompt page.
Legacy Full Dataset
/agent.json
Retained for existing integrations. New agents should use the compact index and paged prompt files.
LLMs Guide
/llms.txt
A lightweight text guide for agents to learn the site purpose, available endpoints, and crawling guidance first.
Full Gallery
/en/gallery
Useful for agents that need visual browsing, filtering by category/style/scene, and opening case detail pages.
Category Index
/en/categories
Understand the library by use case, from UI, posters, portraits, product shots, anime, and more.
Copy For Your Agent
Let Codex / OpenClaw / Claude Code use this prompt library
Send the block below to your agent tool. It will read the guide and JSON dataset first, then retrieve complete prompts for your request.
Use "黑哥 AI" as a GPT Image-2 prompt source: https://www.heigeai.com
Your task:
1. Read https://www.heigeai.com/llms.txt first to understand the site structure, data endpoints, and crawling guidance.
2. Read https://www.heigeai.com/agent/index.json and search titleZh/titleEn, category, and page in the catalogue.
3. Read only the matching https://www.heigeai.com/agent/cases/{page}.json file for complete prompt records.
4. Search category, styles, scenes, promptZh, promptEn, image, and caseUrlZh/caseUrlEn according to my request.
5. For Chinese-community output, prefer promptZh; when I ask for English or international usage, return promptEn.
6. Do not invent prompts and do not return summaries only. Output the complete copy-ready prompt for GPT Image-2 / the latest ChatGPT image model, plus the matching case link and image link.
7. If visual context is needed, open the matching caseUrlZh/caseUrlEn page and inspect the image field.
Confirm that you can read these endpoints, then recommend or organize prompts based on my specific request.Recommended Flow
- 1Read /llms.txt first to understand the site structure and recommended crawling flow.
- 2Search titles, categories, and page numbers in /agent/index.json.
- 3Read only the matching /agent/cases/{page}.json file for complete prompts.
- 4When visual context is needed, open the caseUrlEn or image field.
- 5For prompt study, compare promptZh, promptEn, category, styles, and scenes together.