Teach an agent a visual style and generate matching art

A designer used Hermes Agent to learn a whiteboard-like look from references, package it as a skill, and then create images in that style.

Done with Hermes Agent· Advanced

What they did
They worked in a separate designer profile with its own SOUL.md file describing a hand-drawn whiteboard aesthetic. After pasting in sample illustrations, they asked the agent to break down the palette, line thickness, and layout. Using skill_manage, the agent wrote a style skill that included a Python script for Nano Banana through OpenRouter, then ran a test on a sample prompt before drafting style-aware prompts and saving the generated images.
How it went
The designer saved the skill in that profile’s skills folder and confirmed the script could run. The example output is shown, but no scoring, counts, or other quality metrics were given.
Tips
The script depends on OPENROUTER_API_KEY. If you create the profile with --clone, it copies your default setup; otherwise, set it with hermes config set before generating.

Try it yourself with Hermes Agent

Look at the reference images in [folder] and identify the shared visual style in plain terms: color choices, line quality, composition, spacing, and overall mood. Write that as a reusable skill for this profile, then use it to draft prompts for [number] new illustrations of [subjects] with [image generator]. Before you generate or save anything irreversible, show me the plan and wait for my approval. After approval, save the finished images to [folder].

Fill in the [brackets]. Agents can make mistakes, so review anything before it is sent, paid or deleted.

Read the original on blog.dailydoseofds.com ↗

Get 3 agent uses like this every morning

With the day's top AI news. Free, and one click to unsubscribe.