ClawSec

ClawSec is an open-source security skill suite for AI agent platforms, developed by Prompt Security (part of SentinelOne). It's designed to protect the "cognitive architecture" of AI agents against threats like prompt injection, configuration drift, and malicious instructions. The suite supports several agent platforms—OpenClaw (and its clones like MoltBot and Clawdbot), NanoClaw, Hermes, and Picoclaw—with each platform getting a tailored set of skills covering areas such as security feed verification, config drift detection, supply-chain install verification, and runtime traffic monitoring.

The core capabilities include a one-command suite installer with integrity verification, file integrity protection (drift detection and auto-restore for critical agent files like SOUL.md and IDENTITY.md), live security advisories via automated NVD CVE polling, self-check security audits to detect prompt injection markers, and SHA256 checksum verification for all skill artifacts. Installation is aimed at AI agents themselves—typically run via npx clawhub@latest install clawsec-suite—with cross-platform notes for Linux/macOS and Windows PowerShell.

A key feature is the security advisory feed, continuously populated from NIST's National Vulnerability Database. The feed monitors CVEs related to the supported platforms plus prompt-injection and agent-security patterns, and enriches raw CVSS scores with exploitability context—such as whether public exploits exist, weaponization status, and attack prerequisites—to help agents prioritize real-world threats over theoretical ones. Advisories follow defined JSON schemas for both NVD CVEs and community-submitted reports.