About brand-identity
brand-identity is an interview-driven skill for creating a product’s visual identity, aimed at people without design or marketing backgrounds. It develops three distinct directions for side-by-side browser comparison, then refines the selected option into a self-contained HTML identity document.
Key features
- 01Five-question discovery conversation
- 02Optional moodboard and inspiration review
- 03Three distinct visual directions
- 04Side-by-side HTML comparison page
- 05Final product identity HTML document
- 06Computed WCAG contrast ratios
Use cases
- 01Creating a visual identity for a new product
- 02Choosing colors and typography for an app
- 03Comparing identity directions before selection
- 04Documenting a selected product identity
Install
Copy the command and run it in a terminal.
npx skills add https:/ / github.com/ dhernz/ brand-identity --skill brand-identityAfter installing, type/brand-identityin Claude Code to confirm it loaded.
Bundled filesdesign-mapping.mdoutput-spec.md
Details
From the project READMESkill version: 1.0.0
Brand Identity — Show 3, Pick 1
Overview
Interview-driven brand creation for people who are not designers, marketers, or developers. You run a short conversation (five questions plus an optional moodboard), propose three genuinely distinct visual directions, render them side by side in one browser page so the user can see the difference, let them pick one, then refine that pick into a single self-contained product identity HTML document (logo/wordmark, color, typography, components, spacing, voice & tone, in-use examples).
Core principle: the user makes taste decisions by comparing finished options; you make the design decisions that produce them. Be a consultant, not a form. Never ask design-jargon questions ("what hue? what border-radius?"). Ask about their product and what they want remembered, then translate — every recommendation carries a rationale.
This is product identity only. A full brand strategy (purpose, mission, archetype, personas, messaging) is a future extension — do NOT expand into it unless the user asks.
When to Use
- User wants a brand / brand kit / design system / visual identity and has little or nothing yet
- User asks "what colors/fonts should I use?" for a product, app, or company
- NOT for refreshing an existing mature brand system (audit that instead), and NOT for logo illustration work (the wordmark is typographic only)
Process
Run these five phases in order. Phase 2 (concept confirmation) and Phase 4 (the pick) are gates — do not render or finalize past them without the user.
1. Discovery — 5 questions, ONE message at a time
Ask in plain language, conversationally (NOT AskUserQuestion — these are open-ended). Wait for each answer before asking the next. Tell the user up front this is a conversation, not a rigid form, and they can drop into chat anytime.
- What do you do? (also capture the brand/product name — ask if not given)
- Who is it for?
- Why should they care? (what makes it different)
- Who are you competing with? (names or "nobody direct" — used later to push the directions away from their look)
- The one memorable thing — "What's the one thing you want someone to remember after they see this for the first time?" One sentence: a feeling, a visual, a claim, or a posture. Write it down. Every later decision must serve this. Design that tries to be memorable for everything is memorable for nothing.
If an answer is thin, ask ONE follow-up max, then move on. Non-designers stall when over-interviewed.
No perception sliders. The three directions explore the personality space visually instead of pinning it with a questionnaire.
2. Inspo & moodboard — invite, then read it back
Ask the user for any of: a moodboard (several images dragged into chat), brands / websites they admire, or colors they like (names or hex). Optional but strongly encouraged — a moodboard makes the directions far sharper.
If they give images, read them and extract a vibe profile (see
design-mapping.md → "Moodboard → vibe profile"): dominant palette + temperature,
saturation, contrast, texture/materiality, shape language, typographic feel, and
3–5 mood adjectives. Show the user your read and let them correct it before
generating concepts: "Here's what I'm getting from your moodboard: warm,
editorial, a little vintage, ink-on-cream with one hot accent. Right?"
If the moodboard images conflict, name the tension rather than averaging them into mud: "Two of these are warm and hand-made, one is cold and technical — I'll treat warm-and-made as the through-line unless you'd rather chase the cold one."
If no inspo/moodboard is given, proceed from discovery alone and derive the directions from the answers.
3. Three concepts → render side by side
3a. Name the three concepts in words first. Derive three genuinely distinct
directions from discovery + vibe profile (see design-mapping.md →
"Deriving three distinct concepts"). Present as a lettered list, one line each:
A) "Quiet Ledger" — precise, restrained, monospace accents; numerals as hero
B) "Warm Signal" — editorial warmth, a real serif, an ember accent
C) "Field Notes" — soft, human, organic; grain and rounded forms
Anti-convergence is a hard requirement: the three MUST use different font families, different color temperatures, and different shape/layout rhythms. If someone could swap the wordmark between two of them without noticing, one failed — regenerate it in a deliberately different direction. They should feel like they came from three different design teams.
Get a quick confirm ("build those" / "swap C for something bolder"). Max 2 rounds of concept revision — this is the cheap place to iterate, before rendering.
3b. Build the comparison page. Generate ONE self-contained
brand-directions.html with all three directions rendered full-fidelity, per
output-spec.md → "Comparison page". Each direction panel shows: wordmark, color
swatches (name + hex), a type specimen, a few real components (buttons, a card,
an input) using the user's own domain content, one in-use example, and a
plain-language Safe / Risk note.
Self-gate before showing — hard gate. For each direction ask: "Would a designer be embarrassed to put their name on this?" If yes, discard and regenerate. Triggers: purple/violet gradient hero, 3-column icon-in-circle grid, centered-everything uniform spacing, Inter or Space Grotesk body text, bubble radius on everything, gradient CTA button, generic stock-photo vibe, system-ui as the display/body font. Any one = reject and redo that direction.
Open it in the user's browser (SendUserFile with display: "render", or open).
Tell them: "Three directions, side by side. Look them over and tell me which one
clicks — you can also mix ('B, but the accent from A')."
4. The pick — chat is the chooser
The user states their pick in chat, with any remix notes. There are no interactive buttons and no server — the page is for looking, the chat is for deciding. Confirm you understood: "Going with B — Warm Signal — and pulling A's brighter accent into it. Right?" before proceeding.
5. Refine & deliver the final identity doc
Refine the chosen direction conversationally (apply remix notes, adjust anything).
Then build the final product-identity.html (name it after the product, e.g.
pact.html) following output-spec.md → "Final identity document". Then:
- Compute contrast ratios with a script — never guess. Every text/background pair shown must be real WCAG ratios; fix any body-text pair under 4.5:1 before delivering.
- Fill component examples and voice do/don'ts with content from THEIR domain, not lorem ipsum.
- Render or screenshot to verify nothing is broken, then send it (SendUserFile, render).
- Close by telling them what they got and the natural next step (logo design, full brand strategy) without starting it.
Quick Reference
| Input | How collected | Feeds |
|---|---|---|
| Discovery ×5 | open questions, one at a time | concepts, copy, voice, component content |
| Memorable thing | question 5 | the anchor every decision serves |
| Inspo / moodboard | images / brands / hexes | vibe profile → concepts, palette, type |
| The pick | chat, after the comparison page | which direction becomes the final doc |
References: design-mapping.md (moodboard→vibe, deriving 3 concepts, palette
recipe, font guidance, blacklists, anti-slop, safe/risk) · output-spec.md
(comparison page + final identity document structure)
Common Mistakes
- Asking all discovery questions in one message → users answer one and ignore the rest. One at a time.
- Three directions that look like siblings → the convergence trap. Different fonts, temperatures, and rhythms, or it isn't three options.
- Skipping the concept-confirm gate → rendering three full directions the user could have redirected in one text line.
- Averaging a conflicted moodboard → muddy, characterless result. Name the tension and pick a through-line.
- Defaulting to Inter or Space Grotesk → every AI tool converges there. Never use as primary unless the user names it. See the overused list.
- Guessing contrast ratios → published numbers that are wrong. Compute them.
- Generic filler content ("Lorem ipsum", "Acme Corp") → the doc only clicks when the button says their action and the card shows their data.
- Expanding into full brand strategy (mission, archetype, personas) → product identity only; offer the rest as a follow-up.