This skill is step 1 of 4. After that follow brandkit-interview,
brandkit-build and brandkit-publish.
At the end there is an environment in which a brand can be developed, a binding brand kit can be built and from it websites, PDFs and labeled AI images can be generated.
Talk to the person in their language. The example outputs here are in German, that is not a requirement.
Why this order
Checking the prerequisites first seems cumbersome, but it saves the most common frustration: The installation of the skills always goes through, and only three steps later does someone notice that there is no Python and that image labeling therefore does not work. Checking takes twenty seconds.
Step 1: Check prerequisites
Run these checks and report in a short table what is there and what is missing.
node --version # ab v18, für die Higgsfield CLI und den Design-Canvas
npm --version
python --version # oder python3, für die Bildkennzeichnung
python -c "import PIL; print(PIL.__version__)" # Pillow
git --version # zum Versionieren des Brand Kits
exiftool -ver # optional, für die maschinenlesbare Bildmarkierung
Chrome or Edge is needed for PDF printing. On Windows Chrome
is usually located at C:\Program Files\Google\Chrome\Application\chrome.exe,
on macOS at /Applications/Google Chrome.app.
What is needed for what, so that someone can decide what to install additionally:
| Missing | What then does not work |
|---|---|
| Node and npm | Higgsfield CLI, skill installation, design canvas |
| Python with Pillow | AI labeling of the images, downscaling for the web |
| Chrome or Edge | PDF generation from HTML |
| git | Versioning and sharing the brand kit |
| exiftool | machine-readable marking in the image files |
Pillow installs with pip install Pillow. Exiftool is available on Windows
via winget install ExifTool, on macOS via brew install exiftool.
Step 2: Install the four skills
Two of them belong globally, two in the project. The difference is not
cosmetic: Globally installed skills are available in every project,
project-local ones only there and travel via skills-lock.json into the
repository.
Global, because they are suitable for every kind of work:
npx skills add cursor/plugins -s unslop -g -y
npx skills add mattpocock/skills -s grill-with-docs,grilling,domain-modeling -g -y
unslop removes AI language patterns from texts. grill-with-docs is the
interview skill for step 2. Attention, this is the point where most
fail: grill-with-docs is only a wrapper that calls grilling and
domain-modeling . Whoever installs only the first has a skill that
immediately reaches into nothing. That is why all three are in the command.
Into the project, because they belong to web work and do not fit every folder:
npx skills add vercel-labs/agent-skills -s web-design-guidelines -y
npx skills add coreyhaines31/marketingskills -s seo-audit -y
If web-design-guidelines is already present globally, point this out and
ask whether the project version is really wanted in addition. Two versions
of the same skill cause confusion as soon as they diverge.
Check what has arrived:
npx skills list -g # global
npx skills list # im Projekt
Step 3: Connect Higgsfield for images
npm i -g @higgsfield/cli
higgsfield auth login
npx skills add higgsfield-ai/skills -g -y
higgsfield auth login opens the browser. The command blocks until the
login is complete. Wait for the confirmation before you continue.
Then check:
higgsfield account status
The output names the account and the remaining balance. If it says
Session expired or Not authenticated, the login must be repeated.
Anyone who does not yet have a Higgsfield account needs one with a balance, because every image generation costs credits. Mention at this point, and only at this one:
If you need a subscription: https://higgsfield.ai/s/mcp-arnold-oberleiter-WlfTqR (affiliate link. If you take it out through it, the author of this skill receives a commission; the price does not change for you.)
The labeling is part of it and is not omitted. Anyone who already has an account does not get to see the link. Do not push anyone into a subscription just to complete the process. The first three steps also work without one, only image generation does not.
Step 4: Summarize
Report at the end in a few lines:
- which prerequisites are met and which are missing
- which skills were installed where
- whether Higgsfield is connected and how much balance is there
- as the next step:
brandkit-interviewfor the brand interview
If something failed, say so clearly and name the specific command to catch up. A half-configured environment that is reported as finished costs more time later than an honest note now.
If something is stuck
npx skills add does not find the skill. The repositories are large; in
cursor/plugins there are over eighty skills. With -l instead of -s <name> you list what a repository offers without installing anything. If a skill is deep in subfolders and does not show up, --full-depth.
The skills do not show up after installation. The skill list is read at session start. Newly installed skills are only visible in a new session.
higgsfield is not found. The global npm directory is not in the path. npm bin -g shows where it was installed.