steviebuchicago

Getting Started With Openclaw — Security skill for Claude Code

Security community

Getting started with OpenClaw, the self-hosted personal AI assistant.

How to install Getting Started With Openclaw

This entry records only its repository, not the path inside it, so there is no exact command to give. Open steviebuchicago/getting-started-with-openclaw and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Getting Started With Openclaw does

Getting started with OpenClaw, the self-hosted personal AI assistant. Install, architecture, skills, security hardening, and honest guidance, verified against a real install.

Alternatives in Security

Harden — Run the harden skill to perform a systematic project hardening audit 1k ★
MCP Security Checklist (SlowMist) — Comprehensive checklist: input validation, rate limiting, RBAC, credential management, container hardening 819 ★
Pocketpaw — Your AI agent in 30 seconds 775 ★

README

Getting Started with OpenClaw

**A real personal AI assistant, running on your machine, living in your chat apps — and everything they don't tell you before you run the install script.**

[OpenClaw](https://github.com/openclaw/openclaw) went from side project to a quarter-million GitHub stars in a few months, and it earned it: this is the fastest route that exists today from "I want a personal agent" to *texting your own assistant on WhatsApp before dinner*. This repo is the guide I wanted when I set it up — written for students and anyone standing at the front door, and honest about the part the viral demos skip: **you are now the security team.**

Everything here was verified against a real install of **OpenClaw 2026.6.34** — the CLI output you'll see in these pages is captured from an actual machine, not transcribed from marketing.

The mental model

OpenClaw has four ideas. Everything else is configuration.

**The Gateway is the product.** One self-hosted process on a machine you own. It binds to loopback (`ws://127.0.0.1:18789`), serves a browser Control UI on the same port, and owns everything: channel connections, sessions, skills, cron jobs, credentials. There is no cloud backend. When people say "my OpenClaw," they mean their Gateway.

**Channels are how you reach it.** WhatsApp, Telegram, Discord, Slack, Signal, iMessage, Matrix, Teams — about 29 in all. The assistant doesn't get a new app; it moves into the apps you already check. This is the design decision that made OpenClaw explode, and it is also the decision that makes access control matter (strangers can message those apps too — see the security chapter).

**Agents do the work.** An agent is a model plus tools plus a workspace — a real directory (`~/.openclaw/workspace`) with files that define who your assistant