Find Image Prompt Skill

An open, portable skill for AI agents that turn creative ideas or public reference images into production-ready image prompts and image descriptions.

Website: Find Image Prompt — AI Image Prompt Generator

Find Image Prompt is built around a standard SKILL.md entrypoint and an HTTP API, so it is not tied to one agent, model, or framework. Use it with any agent host that can load a skill, make HTTP requests, or call tools—for example, coding agents, custom assistants, workflow automations, and creative applications.

What the skill does

• Turn a text idea into a polished image-generation prompt.
• Analyze a public reference image and turn its visual direction into a reusable prompt.
• Describe visible subjects, objects, style, or text in a public image.
• Format prompt output for ChatGPT Images, Gemini, Midjourney, Flux, and Stable Diffusion.
• Give agents clear operation selection, authentication, safety, and error-handling guidance.

Use the core skill with any agent

1. Copy this repository or the SKILL.md file into your agent host's skills directory.

2. Create a Find Image Prompt API key in Find Image Prompt Developers.

3. Keep the key in the agent's server or local environment:

   export FIND_IMAGE_PROMPT_API_KEY="your_api_key"
   

4. Invoke find-image-prompt whenever the user needs an image prompt, image-to-prompt conversion, or image description.

The skill contains the API contract and operational guidance. The API itself is model- and agent-neutral: any HTTP-capable runtime can call it.

Direct API example

curl https://findimageprompt.com/api/v1/prompts \
  -H "Authorization: Bearer $FIND_IMAGE_PROMPT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"operation":"text_to_prompt","text":"Rainy Kyoto bookshop at blue hour","target_model":"Midjourney"}'

The Find Image Prompt API requires a paid Starter or Pro account. Read the API documentation for supported operations, parameters, and response formats.

Bring your own model provider

The core skill uses the Find Image Prompt API. The repository also includes an optional local adapter for teams that prefer their own model-provider billing account:

• OPENROUTER_API_KEY
• OPENAI_API_KEY
• GEMINI_API_KEY

The adapter generates comparable prompt text directly with the selected provider. Those keys are never sent to Find Image Prompt, and they do not authenticate requests to the managed Find Image Prompt API.

Optional DeepSeek integration

DeepSeek is one optional function-calling integration, not a requirement for this skill. The bundled Node.js adapter exposes a single create_image_prompt tool and lets DeepSeek select it during a conversation.

export DEEPSEEK_API_KEY="your_deepseek_api_key"
export OPENROUTER_API_KEY="your_openrouter_api_key"
node plugin/deepseek/src/index.mjs \
  --provider openrouter \
  --message "Write a Midjourney prompt for a rainy Kyoto bookshop at blue hour"

See the DeepSeek adapter guide for commands, provider selection, and troubleshooting. It can also serve as the tool layer for a DeepSeek Harness (DSH) workflow, but this repository does not currently include a native, installable DSH apply() module.

Compatibility

Integration style	How to use this repository
Skill-capable AI agent	Load SKILL.md and set FIND_IMAGE_PROMPT_API_KEY.
Custom agent or product backend	Call POST /api/v1/prompts using the included API contract.
Tool/function-calling agent	Expose the operations in SKILL.md as a tool, or adapt the included provider client.
DeepSeek	Use the optional adapter in plugin/deepseek/.
DeepSeek Harness (DSH)	Reuse the adapter's create_image_prompt tool contract in a native DSH plugin.

Provider selection for the optional adapter

Use --provider find-image-prompt, openrouter, openai, gemini, or auto. In auto mode, the adapter selects the first configured key in this order: Find Image Prompt, OpenRouter, OpenAI, Gemini.

Optional model overrides are OPENROUTER_MODEL, OPENAI_MODEL, and GEMINI_MODEL. Keep every key private: do not commit it, put it in URLs, or paste it into chats.

Learning resources

The bundled blog reference collects practical Find Image Prompt articles on writing, evaluating, and adapting AI image prompts. The current online collection is available on the Find Image Prompt blog.

Links

• Find Image Prompt homepage
• Developer portal and API keys
• API documentation
• OpenAPI schema
• DeepSeek adapter documentation

License

Distributed under the MIT License.