ClawSec: Hardening OpenClaw Agents from the Inside Out

This report introduces ClawSec, an open-source "skill-of-skills" that hardens OpenClaw agents by verifying skill integrity, monitoring for configuration drift, scanning for prompt-injection risks, automating audits, and enforcing zero-trust outbound communications. Operating via modular skills and a transparent, community-driven advisory feed (leveraging NVD and GitHub) without centralized servers, ClawSec aims to provide a privacy-first, auditable security baseline for agent ecosystems—motivated by recent findings of malicious OpenClaw skills harvesting credentials.