Does this scanner only look for malicious code?

No, it uniquely uses AI semantic analysis to detect malicious natural language instructions that might bypass traditional code-based security filters.

Which AI agents are compatible with this scanner?

The scanner supports Claude, Codex, Gemini, and the general 'agents' (Skills.sh/Amp) architecture.

Can I scan a skill before I install it locally?

Yes, you can use the command `/security-scanner ` to audit any public repository before adding it to your local environment.

How do I prevent the scanner from flagging my own trusted tools?

You can define trusted plugins, skills, and marketplaces in a `security-scanner.local.md` configuration file to exclude them from audits.

What languages are supported for security reports?

While the default is Japanese, you can configure the `report_language` setting to any language (English, French, etc.) and the AI will generate the report accordingly.

Security Scanner

Name: Security Scanner
Author: lev-os

bylev-os

•

Seguridad y Pruebas

Audits Claude Code plugins and skills for malicious code and harmful natural language instructions using AI semantic analysis.

The Security Scanner is a critical auditing tool designed to protect your AI development environment by inspecting plugins and skills for security vulnerabilities. Unlike traditional scanners, it utilizes AI semantic analysis to identify both malicious code execution and deceptive natural language instructions that could lead to prompt injection or unauthorized data access. It supports a wide range of AI agents—including Claude, Codex, and Gemini—and allows developers to audit public GitHub repositories before installation, ensuring that third-party extensions meet safety standards before they ever touch a local environment.

Características Principales

01Customizable trusted source lists for marketplaces and specific plugins

02Multi-language security report generation (English, Japanese, etc.)

03Cross-platform support for Claude, Codex, Gemini, and Skills.sh

04Semantic AI analysis of code and natural language instructions

05Pre-installation auditing via public GitHub URL scanning

062 GitHub stars

Casos de Uso

01Vetting third-party Claude skills from GitHub before local deployment

02Identifying hidden prompt injection risks in complex natural language skills

03Automating security audits for local plugin directories in CI/CD workflows

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add lev-os/agents security-scanner

For use in Claude.ai and ChatGPT