Круассан Студио — Identity & Design System

Single source of truth for everything that defines Круассан Студио as a brand and as a product team: brand voice, visual identity, design system, code conventions, SEO playbooks, and agent guidelines.

This repo is meant to be read — by humans starting a new project, and by AI agents (Claude, Cursor, Copilot, custom subagents) that need durable context about how we work.

Live reference: ~/Dev/croissan-landing — the marketing site. When this repo and the code disagree, the code is right; update this repo to match.

---

Use this in another workspace (3 ways)

1. Just paste a URL (zero install)

The single-file context summary lives at:

https://raw.githubusercontent.com/CroissanStudioDev/studio-identity/main/SUMMARY.md

In Cursor, Claude.ai, ChatGPT, Windsurf, Continue — paste that URL into the chat (or attach it as context). SUMMARY.md is ~250 lines and self-contained — it's the 20% of the brand that delivers 80% of the value (color tokens, type scale, voice principles, top do/don'ts, anti-slop pre-flight). No clone, no install.

2. Install as a Claude Code skill (one line)

curl -sL https://raw.githubusercontent.com/CroissanStudioDev/studio-identity/main/bin/install | bash

This:

• Clones (or updates) the repo at ~/.croissan/identity
• Symlinks the skill to ~/.claude/skills/croissan-identity
• Auto-detects open-design at common paths and links the integration in too

After install, every project on the machine has access. Claude Code activates the skill when the user mentions Croissan, the studio, or asks for KP / commercial-proposal output.

To update later: re-run the same one-liner. To re-sync just the open-design integration after editing the brand: bash ~/.croissan/identity/bin/sync-opendesign. To uninstall: bash ~/.croissan/identity/bin/uninstall (add --purge to also remove the cloned repo).

3. Clone the full repo

For editing the brand itself, or for projects that need the actual binary assets (logo SVGs, font files, deck-template HTML):

git clone https://github.com/CroissanStudioDev/studio-identity ~/Dev/studio-identity

Then point your project's CLAUDE.md / .cursorrules at the cloned path.

---

How to use this repo

Starting a new project

1. Read SUMMARY.md and brand/voice-and-tone.md to set the mood.
2. Copy design-system/design-tokens.css into your globals.css.
3. Copy logo assets from assets/logos/ and fonts from assets/fonts/.
4. Drop agents/CLAUDE.md into the new project root (or extend the existing one).
5. Default to the stack in tech/stack.md unless there's a real reason not to.

Working inside an existing studio project

1. Make sure the project's CLAUDE.md references back to this repo.
2. When you're unsure about a decision (color, copy, naming, structure), grep here first.
3. If you ship something new and good, back-port the pattern to the relevant doc here.

As an AI agent

Read agents/README.md first. It tells you which files to load up-front, which rules are hard, and how to reconcile this repo with the project you're working in.

---

Repo map

SUMMARY.md         — single-file self-contained brand context (paste-into-any-chat)
brand/             — voice, tone, messaging, facts, sub-brands, proposal structure, reference brands
identity/          — logo, colors, typography, iconography (incl. canonical Lucide allowlist)
design-system/     — tokens, components, sections, decks, animations, a11y, OG cards
tech/              — default stack, Next.js patterns, code standards (incl. perf budget), SEO/GEO
agents/            — rules, checklists, anti-slop, creative-latitude principle
assets/            — logos, client logos, fonts, deck-template, identity-snapshot (the README header)
skills/            — Claude Code skill (croissan-identity) — symlinked into ~/.claude/skills/
integrations/      — plug-in artifacts for external tools (open-design skill registry)
bin/               — install / uninstall scripts for the curl one-liner

How to think about it

There are three concentric rings:

• Locked — brand colors, typefaces, logo, voice principles, factual metrics, accessibility. See agents/design-rules.md and agents/content-rules.md.
• Negotiable — spacing, motion durations, novel pill variants, slight type-scale tweaks. Adjust with cause; document the deviation.
• Free — section layouts, copy, slide compositions, novel section types not in the docs. Invent freely inside the token system.

This is documented top-to-bottom in agents/creative-latitude.md. Read that first if you're an AI agent. It's the lens for everything else here.

Also read agents/anti-ai-slop.md before generating — concrete LLM defaults the studio explicitly rejects (Tailwind indigo, two-stop trust gradients, gradient text on H1, emoji as icons, fake testimonials, invented metrics, Latin quotes in Russian copy, etc.). All P0 entries are non-negotiable; fix them before declaring done. The companion "good tells" section at the bottom is what to grade toward.

---

What we are

• Круассан Студио — full-cycle AI product studio based in Innopolis, Russia.
• Founded 2022. Around 10 people, 5 founders, 15+ shipped projects.
• We embed AI into existing products and ship new ones from scratch in 2–4 weeks.
• Clients include МТС, Ростелеком, Яндекс, T1, Positive Technologies.
• Core message: «Зовите нас, когда нужен продукт с ИИ».

Full company facts live in brand/facts.md.

---

Maintaining this repo

• This is a context repo, not a published library. No CI, no versioning beyond git.
• When you change a token or brand decision, update both this repo and any downstream project.
• Don't add aspirational content — only document what we actually do. If a rule isn't followed in croissan-landing, it isn't a rule.
• Keep docs short. A scannable table beats a paragraph.