This skill is step 2 of 4, between brandkit-setup and brandkit-build.
The result is a file: a detailed rough draft of the brand, from which step 3 condenses the short, binding kit.
Why an interview at all
Without a brand, every design decision turns out new and random. The result then looks like what language models have statistically seen most often: blue-violet gradients, rounded cards, Inter as the typeface. Not wrong, but interchangeable.
An interview forces commitments that you would never make yourself as long as no one asks. The person knows almost everything that is needed here. They have just never said it out loud.
Procedure
First call the skill grill-with-docs . It starts an interview that works in
rounds: all questions that can currently be answered come together,
numbered, each with a recommendation. Then it waits for the answer
before the next round follows.
If grill-with-docs is not installed, conduct the interview yourself following the same
pattern and point out once that brandkit-setup should
install it afterwards.
The axes below are the question pool, not a checklist to work through. What has already been answered is not asked again.
The axes
The order is intentional. Choosing colors without positioning leads to discussions of taste that decide nothing.
1. Core. What is this in one sentence? For whom, and against what does it define itself? What should someone say about the brand when they recommend it?
2. Personality. Three to six characteristics. More useful than agreement is rejection: What should the brand expressly not be? "Never corporate" hits harder than "modern".
3. Structure. One umbrella brand over everything, or several sub-brands? Where does the person's real name stand, where does the brand name? This question is gladly skipped and takes its revenge with the first logo.
4. Models and counter-models. Which two or three brands hit the tone? And which are close but wrong? The counter-example sharpens more than the model.
5. Color. One base tone (dark, light, warm, cool) and at most one to two accents. Ask about an accent that carries meaning, not about a favorite color. Also clarify: Is there a light mode, or is everything dark, including print and documents? This decision has an impact everywhere later and is expensive if it is missing.
6. Typography. Two to three roles are enough: display for headings, body for running text, optionally mono for numbers, prices, versions, technical labels. The mono role is underestimated and is often what makes a technical brand recognizable.
7. Grid and feel. How much whitespace? Symmetrical or editorial asymmetrical? Rounded corners or edges? Shadows, gradients, glass effects: yes or no? A clear no here is worth more than three vague yeses.
8. Building blocks. Which elements keep recurring? Price table, comparison table, FAQ, customer testimonial, form, chart. And is there a signature element by which the brand is immediately recognized?
9. Language. Language and form of address (informal or formal). One good and one bad example sentence pair for the same statement. And a banned list: words that never occur. This is the most effective part of the whole kit, because it can be checked mechanically.
10. Formats. What is actually being built? Website, thumbnails, slides, documents, social posts? Note the pixel dimensions that come up in the process.
11. Images. Photography, AI-generated images, illustration, or no images at all? If AI images occur: which motifs fit, which are forbidden? This answer directly needs step 4.
Where you need to dig deeper
For three answers it is worth asking follow-up questions, because they almost always come out too soft:
- "Modern and professional" describes nothing. Ask about the extreme: brutal minimal, maximalist, editorial, luxury, playful, industrial.
- "Actually I like all colors" means the person has not yet understood that restriction is the whole point. Ask about the proportion: what percentage of an area may the accent occupy?
- An empty banned list means no thought was given. Every brand has things it does not want to be.
Writing the rough draft
Write the result into a file, usually MARKE-ROH.md in the future
brand folder. It may be long and detailed, several hundred lines are
normal. It is the archive, not the working document.
Record why a decision was made the way it was, not just how it turned out. Step 3 cuts away the justifications, but whoever wants to change something in a year needs them.
Enter concrete values where they are fixed: hex colors, font names, pixel sizes. Where something remained open, write that down explicitly instead of inventing a plausible value. Otherwise an invented color will still be in the kit two years later, without anyone ever having decided on it.
At the end: say that the rough draft is ready, name the path, and point to
brandkit-build that turns it into the usable kit.