Generate illustrations in a learned visual style
A designer Hermes profile learns a style from reference designs, writes a skill for it, then crafts prompts and generates and saves images via Nano Banana.
What they did
A separate designer profile has its own SOUL.md describing a hand-drawn whiteboard style. The user pastes reference illustrations and asks it to study palette, line weight and composition. Using skill_manage, it writes a style skill with a Python script that calls Nano Banana (Gemini 2.5 Flash Image) via OpenRouter, then tests the script on a sample prompt. After that, it writes style-informed prompts and saves images.
How it went
The source says the designer writes the skill, saves it under the profile's skills folder and verifies the script runs. It shows example images but reports no quality measures or numbers.
Worth knowing
The script needs OPENROUTER_API_KEY. Creating the profile with --clone copies it from your default setup. Otherwise set it with hermes config set.
Labels
then crafts prompts and generates and saves images via Nano Banana.
A designer Hermes profile learns a style from reference designs, writes a skill for it, then crafts prompts and generates and saves images via Nano Banana.
Using skill_manage, it writes a style skill with a Python script that calls Nano Banana (Gemini 2.5 Flash Image) via OpenRouter, then tests the script on a sample prompt.
The “Material” label failed its quality check and is not used for selection.
Details
- Agents
- Hermes Agent
- Source
- other
- Source date
- undated
- Accounts
- 1
- In registry since
- Oct 12, 2026