Guide · Agent skill
How to use the Reference-Inspired Brand Worlds agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The generate-reference-inspired-brand-worlds skill teaches an AI agent to create several new brand campaign images from one visual example, such as a poster, moodboard or editorial campaign you supply. It takes the general style (mood, palette, texture, composition logic) and applies it to new brand names, meanings, scenes and people, while avoiding protected signature elements. This guide explains what the skill does, how to install it, how to trigger it and how to get useful results.
Reference-Inspired Brand WorldsMengTo/Skills · MIT Download the skill Quick start
- 1Download the ZIP. Download the skill ZIP from the page where you found it, using the same source you would use for any agent skill.
- 2Unzip into the skills folder. Unzip it into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder to limit it to that project.
- 3Or upload it in the Claude apps. In the Claude apps, go to Settings → Capabilities → Skills and upload the skill.
- 4Confirm an image tool is available. Make sure the agent has an image-generation tool, such as the built-in imagegen skill, since the skill depends on one.
- 5Supply your reference image. Attach the poster, moodboard or brand identity image you have rights to use, and say what you want: alternatives, a set of brands or a closer V2.
- 6Pick a similarity level. Name a target such as 'not too different' (70%) or 'distant inspiration' (30%). The percentages guide intent only.
- 7Review and version the outputs. Check each image for spelling, extra text and family resemblance, then keep V1 and save new versions as sibling files.
What an agent skill is and what this one does
An agent skill is a set of instructions the AI agent can load when a request matches the skill's description. This one is defined in a file called SKILL.md, which the agent reads before it starts prompting an image tool.
Given a reference image, the skill has the agent first label each input, then write two short lists: reusable visual grammar (density, palette relationships, surface, lighting, type placement, amount of human presence) and protected signature elements (exact people, poses, objects, layouts, logos, slogans and distinctive letterforms) that must not be reproduced. The agent then generates one image per brand, checks each output, and saves the files.
- Creates multiple distinct brands that share one visual family
- Sets a similarity target from 30% to 85%, expressed as intent rather than a measured score
- Writes originality-safe image prompts, one call per brand
- Makes closer-to-reference V2 versions without overwriting V1
- Checks exact wordmark spelling, extra text and aspect ratio before handing off
Install the skill in Claude (ZIP or project folder)
Download the ZIP of the skill from the page where you found it. Unzip it into your personal skills folder, ~/.claude/skills/, so it is available across your projects. To limit it to one project, unzip it into that project's .claude/skills/ folder instead.
If you use the Claude apps, you can instead upload the skill under Settings → Capabilities → Skills.
- Personal install: ~/.claude/skills/
- Project install: .claude/skills/ inside your project
- Claude apps: Settings → Capabilities → Skills, then upload
Check what the skill needs before you start
The skill does not run scripts and does not use the network itself. It does need an image-generation tool that the agent can call. The skill mentions the built-in imagegen skill; when that is available, the agent loads it and follows its rules for saving, referencing, iterating and validating images.
You also need to supply the reference images. The skill is written for images you have the rights to use.
- An image-generation tool available to the agent
- One or more reference images you provide (poster, moodboard, brand identity image, editorial campaign or generated concept)
- Runs scripts: no
- Uses network: no
Trigger the skill with natural requests
You do not need a special command. Describe what you want in plain language and attach your reference. The skill is meant to activate when you ask for inspired alternatives, several distinct brands in one look, a closer V2, exact wordmarks inside images, or originality-safe prompts.
Name the similarity level if you have one in mind. The skill maps your wording to a target. If you say 'not too different from the inspiration', it uses 70%. If you say 'inspired, not copying', it keeps originality even in the 70–85% range.
- 'Here is our poster. Make five new brands in this same visual family, each with its own meaning.'
- 'Make a closer V2 of this concept at about 70% similarity, keeping the same name.'
- 'Create an image prompt for a brand called MOSSWEFT, spelled exactly, inspired by this moodboard.'
Understand the similarity dial
The percentages are intent guides, not measurable similarity scores. Use them to tell the agent how far to move from the reference.
- 30% distant inspiration: only abstract principles such as mood, density or contrast are kept
- 50% recognizable influence: shares medium and emotional tone; palette, composition, subjects and type system change
- 70% adjacent brand family: shares composition logic, texture, palette relationships, human presence and wordmark scale; subjects, actions, arrangement and letterforms are new
- 85% close art-direction neighbor: keeps most high-level grammar but changes at least four signature elements; no exact commercial identity is recreated
Review the outputs and iterate
The skill asks the agent to inspect every image for exact wordmark spelling with one occurrence, no extra copy or marks, visible brand meaning, family resemblance across the set, and a plausible similarity level. If one check fails, the agent should make one targeted change and re-inspect rather than rewriting the whole prompt.
For a closer version, the skill uses the original reference as Image 1 and the current concept as Image 2, keeps the name and meaning, and saves a sibling file such as brand-name-v2.png. V1 is not overwritten by default.
- Confirm spelling of the wordmark on every file
- Ask for one targeted change at a time
- Keep V1 and V2 side by side for comparison
What people use it for
- Art directors exploring one campaign look across several fictional or new brands
- Marketers building a family of related campaign visuals from a single inspiration image
- Designers writing originality-safe image prompts with a clear similarity target
- Producing a closer-to-reference V2 of a concept while keeping its name and meaning
Frequently asked questions
How do I install the generate-reference-inspired-brand-worlds skill?
Download the ZIP from the page where you found the skill and unzip it into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder. In the Claude apps, you can instead upload it under Settings → Capabilities → Skills.
Is the generate-reference-inspired-brand-worlds skill free to use?
The skill is published under the MIT licence. The source does not mention pricing for the image-generation tool it depends on, so check the terms of whichever image tool your agent uses.
What do I need to run this skill?
You need an agent that can call an image-generation tool, such as the built-in imagegen skill, and reference images you supply. The skill runs no scripts and does not use the network itself.
Can it copy an existing brand's look?
No. The skill is designed to avoid reproducing exact people, poses, objects, layouts, logos, slogans or distinctive letterforms. It is not meant for cloning an existing brand's identity, which may raise trademark or copyright issues.
Do I need to own the reference image?
The skill says reference images should not be kept in reusable assets unless you own them and are authorised to redistribute them. Only use references you have rights to, especially if you plan to share outputs.
How accurate is the similarity percentage?
It is not. The 30%, 50%, 70% and 85% settings express intent for how close the new work should feel. There is no measurable similarity score.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use this skill if you are an art director, marketer or designer who wants a family of new brand visuals from one reference you have rights to, and you are comfortable reviewing outputs by eye. Pick a different approach if you need to reproduce or closely clone an existing brand's identity, or if you need a measurable similarity score, since the skill explicitly avoids the first and does not provide the second.
Download the skill