Definition

Official anthropic plugin for claude-code that runs a multi-agent vulnerability scan inside the coding session and optionally produces patch files for human review — never auto-applied.

Key Points

  • Public beta ~July 22–23, 2026; install /plugin install claude-security@claude-plugins-official
  • Requires Claude Code ≥v2.1.154, paid plan + dynamic workflows, Python ≥3.9.6
  • Jobs: full-repo scan, change/PR/commit scan, suggest patches
  • Distinct from hosted claude-security Enterprise product and from /security-review / security-guidance
  • Adversarial verification panel (REACHABILITY/IMPACT/DEFENSES) before findings ship (secondary MarkTechPost detail)

Sources