feat: Implement Phase 2 dashboard for K8s agent system

Lightweight Go-based dashboard for Raspberry Pi cluster:

Backend:
- chi router with REST API
- Embedded static file serving
- JSON file-based state storage
- Health checks and CORS support

Frontend:
- Responsive dark theme UI
- Status view with nodes, alerts, ArgoCD apps
- Pending actions with approve/reject
- Action history and audit trail
- Workflow listing and manual triggers

Deployment:
- Multi-stage Dockerfile (small Alpine image)
- Kubernetes manifests with Pi 3 tolerations
- Resource limits: 32-64Mi memory, 10-100m CPU
- ArgoCD application manifest
- Kustomize configuration

API endpoints:
- GET /api/status - Cluster status
- GET/POST /api/pending - Action management
- GET /api/history - Action audit trail
- GET/POST /api/workflows - Workflow management

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

dashboard/README.md

@@ -0,0 +1,97 @@
+# K8s Agent Dashboard
+
+Lightweight web dashboard for the K8s Agent Orchestrator system. Designed to run on Raspberry Pi clusters with minimal resource usage.
+
+## Features
+
+- **Cluster Status** - View node health, resource usage, and conditions
+- **Pending Actions** - Approve or reject actions requiring confirmation
+- **Action History** - Audit trail of all agent actions
+- **Workflows** - View and trigger defined workflows
+
+## Resource Requirements
+
+Optimized for Raspberry Pi 3B+ (1GB RAM):
+- Memory: 32-64Mi
+- CPU: 10-100m
+
+## Development
+
+### Prerequisites
+
+- Go 1.21+
+- Docker (for building images)
+
+### Local Development
+
+```bash
+# Run locally
+go run ./cmd/server --port 8080 --data ./data
+
+# Build binary
+go build -o server ./cmd/server
+
+# Build Docker image
+docker build -t k8s-agent-dashboard:dev .
+```
+
+### Build for ARM64
+
+```bash
+# Build multi-arch image
+docker buildx build --platform linux/arm64 -t ghcr.io/user/k8s-agent-dashboard:latest .
+```
+
+## Deployment
+
+### Using Kustomize
+
+```bash
+kubectl apply -k deploy/
+```
+
+### Using ArgoCD
+
+1. Copy `deploy/` contents to your GitOps repo under `apps/k8s-agent-dashboard/`
+2. Apply the ArgoCD application:
+   ```bash
+   kubectl apply -f deploy/argocd-application.yaml
+   ```
+
+## Configuration
+
+### Environment Variables
+
+| Variable | Default | Description |
+|----------|---------|-------------|
+| PORT | 8080 | Server port |
+
+### Command Line Flags
+
+| Flag | Default | Description |
+|------|---------|-------------|
+| --port | 8080 | Server port |
+| --data | /data | Data directory for persistent state |
+
+## API Endpoints
+
+| Method | Path | Description |
+|--------|------|-------------|
+| GET | /api/health | Health check |
+| GET | /api/status | Cluster status |
+| GET | /api/pending | Pending actions |
+| POST | /api/pending/{id}/approve | Approve action |
+| POST | /api/pending/{id}/reject | Reject action |
+| GET | /api/history | Action history |
+| GET | /api/workflows | List workflows |
+| POST | /api/workflows/{name}/run | Trigger workflow |
+
+## Integration with Claude Code
+
+The dashboard reads/writes state files that can be shared with Claude Code agents:
+
+- `status.json` - Cluster status (written by agents)
+- `pending.json` - Pending actions (read/write)
+- `history.json` - Action history (append by agents)
+
+To share state, mount the same PVC or directory in both the dashboard and Claude Code's data directory.