security-scan

Scanning Claude settings for hardcoded secrets requires running a static security scan over the .claude configuration directory to detect API keys, tokens, and passwords, returning a grade with actionable remediation guidance for any CRITICAL findings.

Auditing overly permissive allow rules involves reviewing the Claude settings and skills directory to highlight HIGH severity configurations, allowing you to assess whether broad permissions like expected SSH or API calls match your actual intent.

A configuration audit is needed during scheduled security hygiene checks and immediately after adding new credentials, skills, or MCP servers to your environment to prevent accidental over-permission and reduce exposure.

Remediating security risks after adding new MCP servers requires a static Agentshield scan over the .claude configuration to identify hardcoded secrets and unsafe permissions, producing categorized findings with recommended fixes.

The security scan does not require any external dependencies or components to run, applying a static analysis approach directly over your Claude settings and skills directory to generate grades and findings.

Claude configurations get flagged for unsafe permissions when overly permissive allow rules are detected, enabling you to assess intent and restrict access to prevent accidental over-permission across your skills directory.