NVIDIA/SkillSpector

Security scanner for AI agent skills. Detect vulnerabilities, malicious patterns, security risks, prompt injection, data exfiltration, and supply-chain risks in Claude Code, Codex, and MCP skills before you install them.

What it solves

SkillSpector is a security scanner designed to detect vulnerabilities, malicious patterns, and security risks in AI agent skills (such as those used by Claude Code, Codex CLI, and Gemini CLI) before they are installed. It addresses the risk of executing untrusted skills that may contain prompt injections, data exfiltration attempts, or malicious code.

How it works

SkillSpector employs a two-stage analysis process: fast static analysis combined with an optional semantic evaluation powered by an LLM. It can scan various inputs, including Git repositories, URLs, zip files, and local directories. The tool identifies 71 vulnerability patterns across 17 categories, such as prompt injection, supply chain risks, and privilege escalation. It also performs live vulnerability lookups via OSV.dev for real-time CVE data and provides a risk score (0-100) with severity labels and recommendations.

Who it’s for

It is intended for users and developers of AI agents who want to ensure the skills they install are safe, as well as organizations implementing a verified skills pipeline to gate installations based on security audits.

Highlights

  • Comprehensive Detection: Covers 71 patterns across 17 categories, including prompt injection, memory poisoning, and tool misuse.
  • Multi-format Support: Scans Git repos, URLs, zip files, and single files.
  • LLM-Enhanced Analysis: Supports multiple LLM providers (OpenAI, Anthropic, Bedrock, NVIDIA, etc.) for deeper semantic evaluation.
  • MCP Server Integration: Can be run as a Model Context Protocol server, allowing AI agents to use it as a tool to gate their own skill installations.
  • Baseline Suppression: Allows users to accept known findings to focus only on new issues during re-scans.

Related

  • Project
    cisco-ai-defense/skill-scanner

    A security scanner for AI Agent Skills that detects prompt injection, data exfiltration, and malicious code using multi-engine analysis.

  • Project
    asamassekou10/ship-safe

    An AI security scanner that detects vulnerabilities in application code, AI agents, and supply chains, specifically targeting risks like prompt injection and agent hijacking.

  • Project
    snyk/agent-scan

    A CLI security scanner that discovers AI agents (Claude, Cursor, etc.), their MCP servers, and skills on your machine, and checks them for prompt injection, malicious code, and other vulnerabilities.

  • Project
    openai/codex-security

    A CLI and TypeScript SDK for finding, validating, and fixing security vulnerabilities in code using AI, featuring support for multiple LLM providers and bulk scanning.

  • Project
    bruc3van/agent-skills-guard

    A visual management and security auditing tool for Claude Code skills, providing a curated marketplace and a multi-layer scanning engine to prevent malicious code execution.