security-scan

Auditing AI configuration files for hardcoded secrets involves scanning settings, hooks, and MCP server definitions to detect excessive permissions and injection risks. This skill operates across project-level directories to identify misconfigurations and validate system integrity.

Hardening settings.json and MCP definitions against prompt injection is achieved through a multi-agent adversarial analysis pipeline. This approach simulates attacker perspectives to validate system integrity and provides targeted security hardening recommendations.

Remediation for excessive permissions in AI environment configuration files can be automated. The system provides safe, automated fixes for common misconfigurations and permission issues detected during the vulnerability auditing process.

Scanning project-level configuration directories for security flaws requires the ecc-agentshield dependency. This prerequisite supports the multi-agent adversarial analysis pipeline used to validate system integrity and enforce security best practices.

Adversarial security analysis detects misconfigurations in CLAUDE.md and settings files such as hardcoded secrets, overly permissive tool access, and potential prompt injection vectors. It validates system integrity by simulating attacker perspectives.