Everything Claude Code Harness

This is a comprehensive agent harness system that's been battle-tested over 10+ months of daily use. It adds specialized subagents, slash commands like /plan and /security-scan, lifecycle hooks, language-specific rules, and a skills system that lets Claude research before coding instead of hallucinating APIs. The instincts feature is clever: it exports session-learned patterns with confidence scores so the agent gets smarter over time. Works across Claude Code, Cursor, and other IDEs. Installation requires a manual step for rules since plugins can't auto-distribute them, but the install script handles multiple language stacks at once. If you're running AI agents in production and want more structure than raw prompting, this gives you the scaffolding.