network-platform-resources

What problem does it solve?

Inventory and manage network-accessible compute platforms, connection methods, credentials, and specs. Load this skill when a task could benefit from remote execution, offloading, or cross-platform deployment.

Core Features & Use Cases

• Fleet topology is auto-detected at runtime via fleet-config.js: IPs resolved from .env overrides or tailscale status --json; device metadata from inventory/devices.json; run node scripts/global/fleet-config.js profile for current state
• Platform inventory is defined in inventory/devices.json with roles: Primary Dev Machine (local), SML Node (remote inference tier), Inference Host (remote compute)
• Connection Method: All cross-machine networking uses Tailscale VPN mesh; SSH via ProxyCommand: sudo tailscale nc %h %p; Auth: Ed25519 key pair (passwordless); SSH config references device alias from inventory/devices.json
• Quick Connect: ssh <device-id>; ssh <device-id> "<remote-command>"
• Prerequisites: Tailscale daemon running; Tailscale on remote system service; OpenSSH on remote
• When to Offload: Memory pressure → offload build/test to fleet; Long-running daemon → always on inference host; Heavy npm install / build → faster on fleet node
• When to Keep Local: IDE + Copilot sessions; Quick edits / git ops; Cloudflare deploys
• Adding New Platforms: Install Tailscale and authenticate; Install OpenSSH server; Copy SSH pubkey to remote authorized_keys; Add Host entry in ~/.ssh/config; Add device to inventory/devices.json; Test: ssh <new-alias> echo "OK"

Quick Start

Run the fleet discovery script to load the current platform inventory.