Feb 12
Loading video player...
OpenClaw is the fastest-growing open-source AI project in recent memory - 185,000 GitHub stars already! A legitimately impressive personal AI assistant that can manage your life and you can talk to it anywhere. But it has serious security issues - the docs literally say "there is no perfectly secure setup." So I took the core genius ideas from OpenClaw - the memory system, the proactive heartbeat, the multi-platform adapters, the extensibility through skills - and I built my own version using just Claude Code. It took me two days. The result is simpler, more secure, and tailored exactly to what I need. I'll introduce you to how I did this now and more content on this coming soon! ~~~~~~~~~~~~~~~~~~~~~~~~~~ - The Dynamous Agentic Coding Course is now FULLY released - learn how to build reliable and repeatable systems for AI coding: https://dynamous.ai/agentic-coding-course - The Excalidraw diagram I cover in this video: https://drive.google.com/file/d/1s97vx40pWb2kSH-rfdbzoqIESbX9lKg5/view - Second Brain Skills Collection (open source): https://github.com/coleam00/second-brain-skills - OpenClaw Repository: https://github.com/openclaw/openclaw ~~~~~~~~~~~~~~~~~~~~~~~~~~ Sources for the OpenClaw security issues: - CVE-2026-25253 - One-click RCE via WebSocket origin bypass (CVSS 8.8, discovered by Mav Levin at depthfirst): https://thehackernews.com/2026/02/openclaw-bug-enables-one-click-remote.html - depthfirst original writeup of the RCE chain: https://depthfirst.com/post/1-click-rce-to-steal-your-moltbot-data-and-keys - CVE-2026-25157 - OS command injection in macOS SSH handling: https://www.securityweek.com/vulnerability-allows-hackers-to-hijack-openclaw-ai-assistant/ - 230-414+ malicious skills published to ClawHub registry (76 confirmed malware payloads, 91% used hybrid attacks): https://www.bleepingcomputer.com/news/security/malicious-moltbot-skills-used-to-push-password-stealing-malware/ - 42,665 publicly exposed OpenClaw instances (Censys scan: 21,000+): https://www.darkreading.com/application-security/openclaw-ai-runs-wild-business-environments - Cisco AI Defense: "Absolute nightmare" + Skill Scanner released: https://blogs.cisco.com/ai/personal-ai-agents-like-openclaw-are-a-security-nightmare - Researcher hijacked OpenClaw in 1 hour 40 minutes: https://thenewstack.io/openclaw-moltbot-security-concerns/ ~~~~~~~~~~~~~~~~~~~~~~~~~~ 0:00 OpenClaw - The Magical AI Assistant 2:10 The Major Security Risks 5:16 4 Core Components of OpenClaw 8:21 Replicating the Memory System 10:31 The Heartbeat & Proactive Tasks 11:22 Channel Adapters & Interfaces 12:17 Building Secure Custom Skills 13:38 Step-by-Step Build Process (Simple, High Level) ~~~~~~~~~~~~~~~~~~~~~~~~~~ Join me as I push the limits of what is possible with AI. I'll be uploading videos weekly - at least every Wednesday at 7:00 PM CDT!