security-scan

To scan your Claude Code configuration for security vulnerabilities, you can run a security scan that audits files like CLAUDE.md, settings.json, MCP servers, hooks, and agent definitions for risks like hardcoded secrets and excessive permissions.

Common security risks in CLAUDE.md and settings.json files include hardcoded API keys, prompt injection vulnerabilities, and overly permissive tool access. A configuration security scan identifies these misconfigurations to ensure a safer development environment.

Yes, you can automatically fix hardcoded secrets and misconfigurations in your Claude Code setup by using the auto-fix option. This feature offers safe, automated remediation for common security issues found during the configuration audit.

Yes, AgentShield works with Claude Code MCP servers, hooks, and agent definitions. It analyzes these specific configuration components to detect excessive permissions and potential injection risks within your development environment.

The best way to audit Claude Code hooks for prompt injection risks is to perform a comprehensive configuration scan using AgentShield. This process identifies potential injection vulnerabilities and supports outputting the audit results in JSON or Markdown formats.

When scanning Claude Code configurations for excessive permissions, the scan supports various output formats including JSON and Markdown. This allows you to easily review the identified security flaws and apply safe auto-fixable solutions.