SkillSpector

Tracked

Security scanner for AI agent skills by NVIDIA that detects vulnerabilities, malicious patterns, and security risks across 64 vulnerability patterns in 16 categories.

Author NVIDIA Open Sourced 2026-03-21 Last Commit Unknown

Overview

SkillSpector is a security scanner for AI agent skills built by NVIDIA. It answers "Is this skill safe to install?" by detecting vulnerabilities, malicious patterns, and security risks before installation. Research shows 26.1% of skills contain vulnerabilities and 5.2% show likely malicious intent.

Key Features

  • 64 vulnerability patterns across 16 categories including prompt injection, data exfiltration, privilege escalation, and supply chain risks.
  • Two-stage analysis: fast static analysis with optional LLM semantic evaluation.
  • Multi-format input: scan Git repos, URLs, zip files, directories, or single files.
  • Risk scoring with 0-100 score, severity labels, and clear recommendations.

Use Cases

  • Audit AI agent skills before installing them in Claude Code, Codex CLI, or Gemini CLI.
  • Scan skill repositories for supply chain attacks and malicious patterns.
  • Integrate security scanning into CI/CD pipelines for agent skill development.

Technical Details

  • NVIDIA official project, Apache 2.0 licensed.
  • Live vulnerability lookups via OSV.dev with offline fallback.
  • Outputs terminal, JSON, Markdown, and SARIF reports.