custom-skills

Securely deploy OpenClaw (autonomous AI agent) on a Hetzner VPS with full hardening — Tailscale VPN, firewall, non-root user, loopback-only gateway, Telegram integration, and end-to-end security verification. Use this skill whenever the user mentions OpenClaw, ClawdBot, setting up an AI agent on a VPS, deploying OpenClaw on Hetzner, securing a VPS for AI bots, or wants to run an autonomous AI agent 24/7 on their own server. Also trigger when the user asks about hardening a Linux server for OpenClaw, Tailscale VPN setup for bot hosting, or securing Telegram bot infrastructure. This skill acts as a friendly step-by-step guide that adapts to the user's experience level and walks them through every decision.

Reference for Claude Agent SDK (@anthropic-ai/claude-agent-sdk). MUST use when working with Claude Agent SDK, claude-agent-sdk package, or code that imports from @anthropic-ai/claude-agent-sdk. Covers the SDK's query API, tool configuration, hooks, subagents, MCP integration, and structured output. Do NOT use for generic agent development or other agent frameworks.

Run Codex CLI /review via tmux to review uncommitted changes. Launches Codex in isolated tmux session, sends /review command, selects option 2, captures output. Use when you want a second opinion on uncommitted code changes.