- Home
- Skills
- Image & Video
- StyleDNA Pro: Style Skills + Match Checker
More screenshots
Works with the AI Agents you already use
StyleDNA Pro: Style Skills + Match Checker
Turn reference images into a style skill, get prompts for 6 image tools, score generations 0-100 against the style, and tune prompts until they stay on style.
$12
StyleDNA Pro: Style Skills + Match Checker
Example session with this skill installed
Here are my 4 reference illustrations in ./refs. Make a style skill called "Mid-Century Flat", then check my renders in ./renders against it and fix my prompt.
- Read your context and instructions
- Compiled the styledna pro
- Generated the image
Summary
Created
Mid-Century Flat from 4 references, checked 6 renders against it (average 68/100), and tuned the prompt for one round.
Style created
Palette (9 colors): warm cream #FDEFC9 26% · olive charcoal #545244 17% · deep teal #1E3E3F 10% · mustard yellow #E6A32D 10% · marigold #E79C3E 9% · pale sage #CBCFB4 7% · sage green #5B7569 7% · coral red #DF532C 7% · seafoam #88AB90 6%
- Measured character: light, moderate saturation, medium contrast, moderate detail, some grain
Outlier check: ref-fox.jpg flagged (fit 45): more detailed, more textured and higher contrast than the others. Keep it or drop it?
Traits: flat even lighting · no outlines · stepped color bands instead of gradients · subtle paper grain · geometric shapes · generous cream negative space
Files: SKILL.md (with Midjourney, Stable Diffusion, Flux, GPT Image/Gemini and Ideogram recipes), style.json, style-metrics.json, style-card.html
Style check (6 renders, full check)
| Image | Score | Verdict |
|---|---|---|
| plain-lighthouse.jpg | 50 | off style |
| plain-owl.jpg | 58 | off style |
| styled-bakery.jpg | 69 | close |
| styled-cabin.jpg | 75 | close |
| styled-owl.jpg | 76 | close |
| styled-lighthouse.jpg | 83 | on style |
Shared problem: colors outside the style (beige, khaki, tan, brown). The two plain renders use cool blues and greens that are not in the palette and are far too dark (brightness drift -2.8).
Revised prompt (round 2)
[subject]. Rendered as a mid-century modern flat illustration built from simple geometric shapes with clean edges between flat color fields. Flat, even lighting, with the sky drawn as stepped bands of solid color. Darker areas use deep teal and olive charcoal, accents in mustard yellow and coral red, on a warm cream background. No beige, khaki, tan or brown. A visible paper grain covers the matte surfaces, with generous cream negative space.
Tune round 2
| Subject | Round 1 | Round 2 | Kept |
|---|---|---|---|
| Owl | 76 | 80 (on style) | round 2 |
| Bakery | 69 | 73 | round 2 |
| Cabin | 75 | 67 | round 1 |
Best version per subject kept. Spent $0.014 on 3 images (MiniMax image-01). Full visual report: check-report.html.
Next steps
Run another tune round for the cabin, or use the revised prompt in your own image tool and check the results again.
ref-fox.jpg
PNG · 1536×1024
Example file from a real run - the skill writes it into your workspace.
Connects securely to your tools. The creator never sees your data.
About this skill
Everything in StyleDNA, plus what matters once you start generating: prompts written for your image tool, a style checker that shows when generations drift, and a tune loop that fixes them.
1. Create a style skill. Point your agent at 1–20 reference images. It measures the palette locally, flags references that don't fit the set, writes lighting, linework, texture, shape and composition traits using a precise style vocabulary, and saves a reusable skill folder plus a shareable HTML style card.
2. Prompt any image tool in that style: Midjourney (--ar, --stylize, --no, --sref), Stable Diffusion / SDXL / ComfyUI (prompt + negative prompt), Flux, GPT Image, Gemini (Nano Banana) and Ideogram.
3. Check your generations. Drop in renders and get a 0–100 style score per image, a verdict (on style / close / off style), colors that don't belong, drift in brightness, saturation, contrast, detail and texture, concrete fixes, and a visual HTML report.
4. Tune. Generate, check, fix and repeat (up to 3 rounds), keeping the best image per subject. Use your own tool or Agensi's built-in image tools: no API keys, from about $0.005 per image, and the skill always asks before spending.
5. Blend two styles by weight (for example 70% mid-century flat, 30% risograph).
The scripts run locally (no uploads), and it also checks style skills made with the free StyleDNA desktop app.
Prompt guidance, not model training: results depend on the image tool and never exactly reproduce the references. The score measures color and image statistics; lighting, linework, shapes and composition rely on the agent's visual reading. Style skills from the StyleDNA app get a palette-only check unless the original references are supplied. Max 20 images per run; animated images are not supported.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 1 day ago
Requires
Permissions
File Scopes
The scripts have no network access and never modify images. Images go to an image tool only during the optional Tune job, after the user agrees.