harness-setup

Coach developers through configuring Claude Code or Codex agent harnesses.

1|Updated Mar 17, 2026
One-click install
npx skills add https://github.com/BedirT/harness-setup --skill harness-setup-bedirt
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: harness-setup
Source: https://github.com/BedirT/harness-setup/tree/main
Command: npx skills add https://github.com/BedirT/harness-setup --skill harness-setup-bedirt

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) and scripts (resource) components.

What problem does it solve?

An interactive coaching workflow that helps developers configure a complete agent harness for any project, ensuring best practices and preventing risky, AI-generated instruction content.

Core Features & Use Cases

  • Eight-phase interactive coaching workflow: project exploration, CLAUDE.md/AGENTS.md authoring with context budget, deterministic verification with hooks and back-pressure, MCP/CLI tool auditing, verification loops and self-correction, sub-agent orchestration, long-running infrastructure, and skills/docs finalization.
  • Dual-tool support: supports Claude Code and Codex workflows with templates, docs, and progressive disclosure to minimize context load.
  • Human-guided, not AI-generated instructions: your team writes CLAUDE.md or AGENTS.md; the coach provides skeletons, guidance, and review to prevent downstream errors.
  • Structured scaffolding: provides references, templates, and scripts to accelerate harness setup.

Quick Start

Begin by evaluating your project's needs, then select the harness phases to enable and let the coach guide you through Phase 1.

Frequently Asked Questions about harness-setup

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I set up a Claude Code or Codex agent harness for my project?

To set up a Claude Code or Codex agent harness, this Skill coaches you through an eight-phase workflow covering project exploration, CLAUDE.md or AGENTS.md authoring, hooks, MCPs, and sub-agent orchestration.

What should I include in my CLAUDE.md or AGENTS.md file to configure an agent harness?

A CLAUDE.md or AGENTS.md file for an agent harness should include a context budget, deterministic verification rules, and progressive disclosure references. The coach provides skeletons and guidance to help your team write these instructions safely.

How do I create deterministic verification loops with hooks for an AI agent?

You create deterministic verification loops by configuring hooks and back-pressure mechanisms within your agent harness. The coaching workflow guides you through setting up these self-correction loops to validate AI outputs reliably.

Does this agent harness setup workflow support both Claude Code and Codex?

Yes, the agent harness setup workflow supports both Claude Code and Codex. It provides dual-tool templates, documentation, and progressive disclosure rules to minimize context load across either platform.

What is progressive disclosure in the context of building an agent harness?

Progressive disclosure in an agent harness is the practice of structuring docs and rules to minimize context load. It allows the AI to access detailed instructions only when needed, maintaining a tight instruction budget.

Why use a guided coaching workflow instead of AI-generated instructions for an agent harness?

A guided coaching workflow prevents risky, AI-generated instruction content by having your team write the harness rules directly. The coach provides skeletons and reviews to ensure best practices and prevent downstream errors.