Add shared Supabase pattern and specialized agents to MEMORY.md

New sections added:
- Rule 2.1: Shared Supabase pattern for all Vercel web apps
  - Same Supabase project (qnatchrjlpehiijwtreh)
  - Table naming: <projectName>_tableName
  - Copy .env.local from existing projects

- Specialized Agent Profiles section
  - List of all available agents
  - When to use each
  - How to spawn them
  - Key benefit: auto-loaded skills

Now documented in both agent configs and curated memory

MEMORY.md

@@ -60,6 +60,42 @@
 
 ---
 
+### 2.1. SHARED SUPABASE PATTERN FOR ALL VERCEL WEB APPS
+❌ **FORBIDDEN:** Creating separate Supabase projects for each web app  
+✅ **REQUIRED:** All Vercel apps share the same Supabase project with prefixed tables
+
+**The Pattern:**
+- **Shared Supabase Project:** `qnatchrjlpehiijwtreh` (qnatchrjlpehiijwtreh.supabase.co)
+- **Table Naming:** `<projectName>_tableName` format (e.g., `blog_messages`, `mission-control_tasks`)
+- **Credentials:** Copy `.env.local` from existing project (blog-backup or gantt-board)
+- **Auth:** Shared across all apps (same users table)
+
+**Why This Pattern:**
+- Prevents table name collisions (gantt-board uses unprefixed tables - legacy)
+- Single auth source (users can access all apps with one login)
+- Easier management (one Supabase dashboard)
+- Cost efficient (free tier covers all apps)
+
+**Example from blog-backup (CORRECT):**
+```typescript
+// Table name: blog_messages (prefixed)
+supabase.from("blog_messages").select("*")
+```
+
+**Example from gantt-board (LEGACY - created before pattern):**
+```typescript
+// Table names: tasks, projects, sprints (NOT prefixed - problematic)
+supabase.from("tasks").select("*")
+```
+
+**New Project Setup:**
+1. Copy `.env.local` from blog-backup (same Supabase credentials)
+2. Create tables with prefix: `<projectName>_tablename`
+3. Set up RLS policies for the prefixed tables
+4. Document table names in project README
+
+---
+
 ### 2.5. CLI MUST STAY IN SYNC WITH WEB UI — ALWAYS
 ❌ **FORBIDDEN:** Web UI has features the CLI can't access  
 ✅ **REQUIRED:** Every web API endpoint has a matching CLI command
@@ -748,6 +784,70 @@ When creating documents for gantt board tasks:
 
 ---
 
+## Specialized Agent Profiles
+
+**Created:** February 22, 2026  
+**Location:** `~/.openclaw/agents/<agentId>/agent/agent.json`  
+**Documentation:** `/Users/mattbruce/.openclaw/workspace/SPECIALIZED_AGENTS.md`
+
+### Available Agents
+
+| Agent ID | Name | Emoji | Purpose | Auto-loaded Skills |
+|----------|------|-------|---------|-------------------|
+| `main` | Main (default) | 🎉 | General purpose | — |
+| `ios-dev` | iOS Developer | 📱 | iOS/Swift/SwiftUI | ios-26-role, swiftui-expert-skill, swift-clean-architecture |
+| `web-dev` | Web Developer | 🌐 | Next.js/React/Supabase | nextjs-expert, frontend-design, shadcn-ui, ui-ux-pro-max |
+| `research` | Research Assistant | 🔍 | Research & analysis | browser-automation, session-logs |
+
+### When to Use Each
+
+**ios-dev:**
+- Any iOS app development
+- SwiftUI view creation
+- Xcode project work
+- App Store preparation
+
+**web-dev:**
+- Next.js apps
+- React components
+- Supabase integration
+- Vercel deployment
+- **Follows shared Supabase pattern automatically**
+
+**research:**
+- Information gathering
+- Documentation compilation
+- Best practices research
+- Competitive analysis
+
+### How to Spawn
+
+```json
+// iOS task
+{
+  "agentId": "ios-dev",
+  "task": "Build a habit tracker app..."
+}
+
+// Web task
+{
+  "agentId": "web-dev",
+  "task": "Create a Next.js dashboard..."
+}
+
+// Research task
+{
+  "agentId": "research",
+  "task": "Find OpenClaw best practices..."
+}
+```
+
+### Key Benefit
+
+Each agent **automatically loads relevant skills** and follows **domain-specific patterns**. No more generic solutions — iOS work uses iOS patterns, web work uses web patterns.
+
+---
+
 ## Memory Maintenance — Critical Rule
 
 **Update memory files CONTINUOUSLY throughout the day, not at the end.**