louis-codex-stack

What problem does it solve?

Managing multiple specialized AI agents for development work often leads to role confusion, security risks from leaked secrets, misallocated tasks, and conflicting work between parallel agents. This Skill eliminates those issues by providing clear, actionable rules for multi-agent orchestration.

Core Features & Use Cases

• Clear Agent Role Definitions: Assigns specific, non-overlapping responsibilities to Codex (main orchestrator, final implementation), Hermes (parallel experiments, research, summaries), Warpish (terminal/router), Taskie/CDCH (visual command center), OpenRouter (provider vault), and local model workers.
• Security and Cost Guardrails: Enforces rules to keep secrets out of public repos, prioritizes cheap/local models for repetitive work, and restricts write/destructive actions unless explicitly approved.
• Collaboration Workflows: Defines standard processes for parallel work, including requirements for Hermes to summarize completed work, changes made, and pending Codex approval items.
• Use Case: When running a bounded experiment testing a new local model via Hermes while Codex handles core repository validation, this Skill ensures clear task boundaries, no duplicate work, and proper approval flows.

Quick Start

Use the louis-codex-stack skill to define the correct roles and collaboration rules for running a parallel Hermes experiment on a new local model while Codex works on core repo validation.