security-scan

To scan Claude Code configurations for security vulnerabilities, you can apply this skill to your project. It audits CLAUDE.md, settings.json, and MCP files for risky permissions, secrets, and prompt injection vectors using AgentShield.

Security risks in CLAUDE.md and MCP files include exposed secrets, insecure permissions, and prompt injection patterns. This skill scans these files alongside hooks and agent definitions to identify data exfiltration risks and unauthorized access vectors.

You need to run a security audit when onboarding a new Claude Code project, after modifying CLAUDE.md or settings.json, changing MCP configurations, or during periodic security hygiene checks to ensure configurations remain secure before deployment.

Yes, you can automatically remediate insecure permissions and secrets found in Claude Code settings. After scanning configurations with AgentShield, this skill can apply safe fixes directly via the CLI to resolve detected vulnerabilities.

Yes, this security scan checks Claude Code hooks and agent definitions for data exfiltration. It validates these components alongside MCP configurations to prevent unauthorized access and ensure your project setup is secure.

The best way to prevent prompt injection in Claude Code agents is to audit configurations for injection patterns. This skill scans CLAUDE.md and agent definitions to detect and remediate prompt injection vectors before deployment.