r7-Harness

A simplified version of my own agent setup on OMP, for Claude and Codex.

GitHub Repo

Overview

r7-Harness is a simplified version of my own agent setup, built on OMP.

It only supports Claude and Codex models, but the other OMP providers can be added back. It runs on WSL with Windows Terminal or r7-Shell.

This repo is an unofficial modification of OMP. It isn't officially associated with OMP, Codex, or their developers.

Personality

You pick your agent's name, provider and model when you install it, and it gets its own separate local profile. How it behaves is up to you in its personality file. Your identity, instructions, credentials, sessions, memory, and project details stay on your own machine.

r7-Harness terminal workspace

Included Features

Compact Interface

Session titles, turn timing, usage and status, with an input box that stays put.

Separate Profiles

Each agent gets its own identity, config, login and sessions.

Instructions and Skills

Global and project instructions, plus skills.

Themes and Fonts

Idle and working themes that follow what the agent is doing, and each model can have its own, plus a list of good terminal fonts.

Reply Style

Short replies with colored markers and numbered questions, every color can be changed.

Safeguards

It won't change anything until setup checks pass, tools are limited in what they can change, and subagents can't edit the same files.

Pinned Build

Built on OMP 18.8.3 with the parts I don't use taken out, and it's easy to roll back to an earlier version.