Show HN: A security scanner for AI Agent Skills

Category: security

Tags: security, ai-agents, static-analysis, sandbox, vulnerability-scanner

Score: 7.0/10 (Innovation: 7, Technical: 7, Documentation: 8, Utility: 6)

SkillWard is a security scanner for AI Agent Skills that combines static analysis, LLM evaluation, and sandbox execution to identify potential risks. It's interesting because it addresses a novel and growing attack surface in the AI agent ecosystem by providing runtime verification in isolated Docker containers, moving beyond purely static or heuristic analysis.

Target audience: devops, security engineers, ai developers

Repository: https://github.com/Fangcun-AI/SkillWard/tree/main · Python · NOASSERTION · 29 stars

View on Hacker News