Openclaw Security Hardening

If you're running high-privilege AI agents with terminal access, this deploys the OpenClaw Security Practice Guide's three-tier defense: behavioral blacklists for destructive commands, permission narrowing during execution, and automated nightly audits tracking 13 security metrics. It's designed to be agent-executable, meaning you can hand the guide directly to your AI agent for self-deployment. The red/yellow line system blocks things like "rm -rf /" outright while pausing on risky patterns like piping curl to bash. Includes cron job setup for 2 AM audits and Git-based disaster recovery. Built by the SlowMist team, clearly written for paranoid operators who've thought through prompt injection and supply chain attacks.

Install to Claude Code

npx -y skills add aradotso/security-skills --skill openclaw-security-hardening --agent claude-code