SkillSpector
Security scanner for AI agent skills to detect vulnerabilities and malicious patterns.
Catch prompt injection, data exfiltration, and supply-chain risks in Claude Code, Codex, and MCP skills.
SkillSpector is a security scanner built to evaluate AI agent skills before you install them. It targets developers and security professionals using AI assistants who need to vet implicit trust risks in third-party agent tools.
What it is
The tool executes a two-stage analysis combining fast static checks with optional LLM semantic evaluation across 71 vulnerability patterns. It integrates live OSV.dev vulnerability lookups and supports multiple output formats like SARIF and JSON for CI/CD pipelines.
Who it's for
It suits security engineers and developers integrating agent frameworks like Claude Code or Model Context Protocol who require supply-chain protection. It does not suit teams lacking CLI proficiency or those unwilling to configure LLM provider credentials for semantic analysis.
Key features
Multi-Format Input Scans
Scan Git repositories, URLs, zip files, local directories, or single files directly.
71 Vulnerability Patterns
Detect prompt injection, data exfiltration, taint tracking, and MCP tool poisoning.
Two-Stage Analysis Engine
Combine rapid static parsing with optional semantic LLM evaluation for deep inspection.
Baseline False-Positive Suppression
Suppress known findings via glob rules or fingerprints so re-scans highlight only new issues.
Up and running in 3 steps
Our honest verdict
What's great
- Scans agent skills across multiple formats including Git repos, URLs, and zip files.
- Performs live vulnerability queries against OSV.dev for real-time CVE data.
- Supports multiple output formats including terminal, JSON, Markdown, and SARIF reports.
- Provides a baseline mechanism to suppress known false positives across re-scans.
Where it falls short
- Requires Python 3.12+ and command-line management for installation and execution.
- Deep semantic analysis requires active API keys and paid access to supported LLM providers.
- Container usage requires familiarity with Docker image builds and local volume mounting.
Pricing & license
Free under its Apache-2.0 license, with the only running costs dependent on your own hardware or API usage fees if you enable paid LLM providers for semantic analysis.
Open remote AI-training roles
880 open remote AI-training roles, updated November 15, 2026; most pay $20–$175 an hour. Zekai lists them free; the application goes to the employer’s platform.
Alternatives
These reviewed AI tools do the same job instantly — from our scored directory.
Use it, or have us set it up
FAQ
What AI agent frameworks does SkillSpector support?
It scans skills designed for systems like Claude Code, Codex CLI, Gemini CLI, and Model Context Protocol (MCP).
Can I run SkillSpector without an active internet connection?
Yes, live vulnerability lookups feature an automatic offline fallback when OSV.dev is unreachable.
How do I ignore known findings during a scan?
You can generate a baseline file using `skillspector baseline` and reference it with the `--baseline` flag during scans.
