Immutable. This exact content is served forever at /api/v1/blob/a3e5508cf55aecb6.
# CLAUDE.md — Development Guardrails 🛡️ ## 💎 Project Philosophy This is the **Design Skill OS**, an elite cognitive layer for AI agents. We don't build "assets"; we build **reasoning systems**. Every code change must improve the AI's ability to think like a master designer. ## 🛠 Command Center - **Sync Dependencies**: `npm install` - **Global Dev Install**: `npm install -g .` - **Local Test Sync**: `design-skill init --ai claude --offline` - **Maintenance**: `design-skill versions` | `design-skill update` ## 🧩 Architecture Snapshot - **`cli/`**: The deployment engine. Uses `commander` and `fs-extra` to map the Skill OS to various assistants. - **`src/design-skill/`**: The Core Intelligence (The "Brain"). - `SKILL.md`: The 900+ line master knowledge base. - `data/`: Industry-specific reasoning rules (161+ rules). - `scripts/`: Python-based logic engines. - `templates/`: High-fidelity UI pattern library. ## 🧪 Development Standards 1. **The Brain is Sacred**: Changes to `SKILL.md` must be rooted in proven design principles (Glaser, Lupton, etc.). 2. **Adversarial First**: Every new feature must be compatible with the **Red Team Protocol** (Section 23). 3. **Multi-Assistant Native**: Always verify path resolution across `claude`, `cursor`, `windsurf`, `antigravity`, and `copilot`. 4. **Pro-Max Output**: CLI feedback should be informative and use `chalk` for high-fidelity terminal styling. ## 📂 Folder Map - `/cli`: Node.js CLI source. - `/src/design-skill`: Core design logic. - `/docs`: Strategy and historical context. - `/.claude`: Development/Test environment for Claude Code. --- **Build with intent. Design for the future.**