Overview

SkillSpector is a powerful security scanner designed to protect your AI agent ecosystem. It proactively detects vulnerabilities, malicious patterns, and security risks within AI agent skills before they are installed, ensuring the integrity and safety of your Claude Code, Codex, and MCP environments. With a significant percentage of analyzed skills containing vulnerabilities or exhibiting malicious intent, SkillSpector provides a crucial layer of defense against potential threats.

Key Features

• Comprehensive Vulnerability Detection: Identifies 71 distinct vulnerability patterns across 17 categories, including prompt injection, data exfiltration, privilege escalation, supply chain risks, excessive agency, and more.
• Multi-Format Input Support: Accepts skills from Git repositories, URLs, zip files, directories, or individual files for flexible scanning.
• Advanced Analysis Pipeline: Employs a two-stage analysis process combining fast static analysis with optional LLM-powered semantic evaluation for deeper insights.
• Real-time Threat Intelligence: Integrates live vulnerability lookups against OSV.dev for up-to-date CVE data, with an automatic offline fallback.
• Actionable Reporting: Delivers risk scores (0-100), severity labels, clear recommendations, and multiple output formats (Terminal, JSON, Markdown, SARIF) for easy integration and understanding.
• Flexible Suppression: Allows for baseline and false-positive suppression to focus on new or unaddressed issues, ensuring reports remain relevant.
• LLM Integration: Seamlessly integrates with various LLM providers (OpenAI, Anthropic, Bedrock, Ollama, etc.) for enhanced semantic analysis, or can operate in a faster, static-only mode.
• MCP Server Capability: Can be run as a Model Context Protocol server, enabling AI agents to use SkillSpector as a tool to gate skill installations in real-time.

Typical Use Cases

• Pre-installation Security Audits: Scan any AI agent skill before integrating it into your development workflow or production environment to prevent the introduction of malicious code or vulnerabilities.
• CI/CD Pipeline Integration: Utilize SARIF output for automated security checks within your continuous integration and continuous deployment pipelines, ensuring only vetted skills proceed.
• Agent Runtime Protection: Deploy SkillSpector as an MCP server to act as a runtime guardrail, actively preventing the installation of unsafe skills by agents like Claude Code, Codex CLI, and Gemini CLI.
• Research and Development: Analyze novel agent skills to understand potential security implications and contribute to a safer AI ecosystem.
• Threat Hunting: Investigate suspicious or unexpected agent behavior by scanning the skills involved for known malicious patterns and vulnerabilities.