ClawSec

What is ClawSec?

ClawSec is a security audit platform designed for AI agent skills, including Claude Code and MCP servers. It provides a five‑tier analysis pipeline with a Trust Score system for risk assessment. The platform has audited over 33,000 skills, identifying 20+ malicious and 160+ suspicious items. It targets threats such as prompt injection, data exfiltration, crypto wallet theft, and AI alignment attacks. ClawSec is free to use.

How to use ClawSec?

—

Key features of ClawSec

• Five‑tier analysis: static analysis, pattern matching, LLM semantic review, Firecracker sandbox execution, and LLM final review
• Trust Score system for overall risk assessment
• Detection of prompt injection, data exfiltration, crypto wallet theft, and AI alignment attacks
• Audited over 33,000 AI agent skills
• Identified 20+ malicious and 160+ suspicious skills

Use cases of ClawSec

• Audit Claude Code skills for security vulnerabilities before deployment
• Evaluate MCP servers for malicious or suspicious behavior
• Assess the risk of AI agent tools using a structured, multi‑stage analysis pipeline

FAQ from ClawSec

How many skills has ClawSec audited so far?

ClawSec has audited more than 33,000 AI agent skills, finding over 20 malicious and 160 suspicious items.

What types of threats does ClawSec detect?

ClawSec detects prompt injection, data exfiltration, crypto wallet theft, and AI alignment attacks.

Is ClawSec free to use?

Yes, the platform is free to use.

What does the Trust Score represent?

The Trust Score is a system for overall risk assessment of a skill, based on the results of the five‑tier analysis.

What analysis methods are used in ClawSec’s pipeline?

The pipeline includes static analysis, pattern matching, LLM semantic review, Firecracker sandbox execution, and a final LLM review.