secure-agent/README.md

This is an AI Coding Agent built on security first principles

Architecture

• the interface and orchestration runs natively
• all tool use and filesystem access done by the llm is sandboxed using Podman with krun microvm runtime.

todo

• sandbox
• generic llm sdk
• UI
  • TUI
  • web GUI
• tools:
  • bash
  • read file
  • write file
  • git push
  • git pull