Claude Code Skills Just Got a MASSIVE Upgrade

Anthropic released a new 'skill creator' plugin for Claude Code that addresses key pain points with custom skills: no systematic testing, unreliable skill triggering, and no way to benchmark improvements. The skill creator enables A/B testing (with vs. without a skill), parallel multi-agent test runs, and description optimization to improve trigger reliability. It distinguishes between two skill types—capability uplift (making Claude better at something) and encoded preference (enforcing a specific workflow)—and applies different evaluation strategies to each. Installation is straightforward via the /plugin command, and the tool provides visibility into skill performance without requiring any code.