everything-claude-code

Hey, I'd like to build an advanced performance optimization system specifically designed for AI agent harnesses. The goal is to create a robust, modular framework that enhances the capabilities of various AI agents, like Claude Code, Codex, Cursor, Gemini, and others.

The core idea is to develop a system, primarily in JavaScript/TypeScript with Node.js, that acts as an intelligent layer for these agents. It should focus on providing and managing "skills" (reusable functions or tools), establishing "instincts" (rules and guided behaviors), optimizing "memory" (context management and persistence, possibly leveraging an in-browser SQL database like `sql.js`), and integrating strong "security" measures from the ground up.

Architecturally, I envision a highly extensible, plugin-based system where new agents and their unique configurations can be easily integrated. It should support a research-first development approach, allowing for continuous experimentation and improvement of agent intelligence. Python would also be integrated for scripting utilities and potentially a simple dashboard or monitoring component. The system needs to be highly configurable, likely through YAML or similar files, to define agent behaviors, rules, and environments.

Ultimately, I'm looking for a production-ready framework that enables AI agents to be more performant, secure, and capable through a unified, optimized harness.