Details can be found on my github: hchoi41/agentsetup: This git contains my device’s local agent setup

This uses the multi-LLM orchestration workflow explained in my previous post: Commander’s Intent.

This agent_setup is a portable toolkit for configuring AI coding agents on different computers: the local hub stores the current skills, instructions, governance rules, and scripts; approved files are filtered and published to GitHub; a new machine clones the repository and runs Get-AgentSetup.ps1 to install them and rebuild the local agent connections. It also includes a multi-LLM workflow where Codex creates or fixes work, Copilot reviews it, and Claude coordinates or resolves disagreements, with tests and safety scans used as quality gates.

I use it for my daily workflows, which are centered more on cowork/documentation-type work involving research, analysis, and some coding (even if I may be using Codex or Claude Code). I do not unleash agents for fully autonomous workflows, and this agent setup is not intended for that use case because I do not think AI agents are capable of fully autonomous workflows yet, at least not for the types of work that I do. Alternatively, I may simply not have found a way to automate them end to end without quality slipping.