Three-stage security checks: local rules → git diff → deep review, and it also comes with false-positive filtering. Anthropic really understands enterprise-level security needs, and automatic deployment in virtual environments is also hassle-free.

Anthropic launches Claude Code automatic vulnerability scanning and fixing plugin, which can intercept and fix vulnerabilities in real-time during code writing sessions. Once enabled, PR fix suggestions decrease by 30-40%. The security check occurs in three stages: local rule matching during editing, comparing git differences at the end of the session, and in-depth review at submission, with false positives excluded. Enterprises can customize rules, requiring Claude Code CLI 2.1.144+ and local Python 3.8+, with automatic deployment of virtual environment and installation of agent SDK upon first activation.