Simple AIWhat it does: Reads your past Higgsfield images and their prompts, finds the looks you use, and writes one style guide and a base prompt so new pictures match.
Use it when: Your AI pictures look different every time and nobody knows which look is the brand.
You get: Your looks grouped with counts, the one to keep, a base prompt, a never list, and a saved skill.
You are my brand designer, a senior art director with 15 years of writing style guides that keep a brand's pictures looking the same. You can read my Higgsfield history. Goal: find the looks in my past Higgsfield pictures and write one style guide and one base prompt, so I end with new pictures that match the brand every time. Context: I make pictures for my business with AI. My customers want to know a post is from me at a glance, because when every picture looks different, nobody knows which look is the brand. First, check that you can open my Higgsfield history. Tell me how many images and videos you found. If you can't open it, stop and follow the Higgsfield steps at simpleai.community/tools/higgsfield/, or ask me to paste my last prompts instead. Ask me these, then run. Skip any I already answered. 1. How many to read? (Default: my last 30 images and videos.) 2. Is this for one brand or more? (Default: one brand. If two looks seem to be for different brands, write [check: same brand?].) Do this: 1. Read my last 30 images and videos: the prompt, the model, the size and the date. Tell me how many you read. 2. Group them into looks (for example: photo style, flat drawing, lo-fi art, app icons). For each look: how many, the dates, the colors named in the prompts, and one short example prompt line. Check the counts add up to the number you read. 3. Tell me which look I use most lately, and where looks clash (two color sets for the same brand). 4. Pick the one to keep: the look I use most in the newest half of what you read. Give the reason in one line. It's my call in the end. 5. Write a style guide for that look: light, colors (only hex codes found in my prompts), camera or drawing style, mood, and what's always in it (for example: "no text, no logos"). 6. Write one base prompt under 50 words I can start every new picture from. Every new picture starts from these same words, so it's made the same way each time. 7. Write a never list with every avoid line in my prompts (no text, no logos, and so on), and say which look each came from. 8. If you can't tell whether two looks are for the same brand, write [check: same brand?]. Give me: the looks as a table (look, count, dates, colors, example line), the one to keep, the style guide in under 10 lines, the base prompt, then the never list. Before you show me, check: - The look counts add up to the number of images and videos you read. - Every hex code and every never line is copied from one of my prompts. - The base prompt is under 50 words and uses only words from the style guide. Rules: - Only what's in my history. Don't guess colors or rules that aren't in my prompts. - Don't make or change any images. Drafts only. - Put a line under "always in it" only if every prompt of that look has it. If not, say how many of them have it. - Use only what is in my data and what I tell you. Never make up a name, number, date, price, link or promise. - If something you need is missing, write [check: what is missing] in that spot and keep going. Last, save these steps as a skill called 'Picture Style Guide' so you do it this way each time I make a new picture. To run it again, just say: "Run my Picture Style Guide skill."
You need
Higgsfield
Any AITested on sample data
Connect first: your files. The steps for each AI are on one page. Then the prompt reads it for you. It ends by saving itself as a skill.
Built on The E-Myth Revisited's rule to do it the same way every time.
Copilot version
There is no separate Copilot version of this one yet. Paste the prompt above into Microsoft 365 Copilot Chat.
See a real example
Tested on sample data Oct 29 2026, after the rewrite: a Claude agent ran it on made-up sample input (tool results simulated). A separate Claude agent judged the run and the card.
Watch: no video for this one yet.
AuthorCorey Leon
