🚀 Level Up Your Dev Workflow: How to Configure OpenHarness for Killer Project Specs

 

Ever feel like handing a raw Business Requirement Document (BRD) to your AI coding assistant results in a chaotic mess? 😅 While GitHub Copilot is absolute magic at writing code, throwing massive, unstructured requirements at it can lead to missed context, sloppy architecture, and endless debugging loops.

Enter OpenHarness 🦄—the multi-agent framework designed to take your high-level ideas, slice them up with specialized AI personas, and generate crystal-clear documentation before a single line of code is written.

Here’s your step-by-step guide to setting up OpenHarness on Windows and why pairing it with GitHub Copilot is a complete game-changer. 👇

---

🛠️ Step-by-Step OpenHarness Setup (Windows Edition)

Step 1: Clone the Repository 📂

First, grab the official repository to get all the underlying scripts and configurations.

git clone https://github.com/HKUDS/OpenHarness

Step 2: Installation ⚡

Because we’re on Windows, you can kick off a slick one-click installation via PowerShell:

# One-click install (PowerShell)
iex (Invoke-WebRequest -Uri 'https://raw.githubusercontent.com/HKUDS/OpenHarness/main/scripts/install.ps1')

Alternatively, if you prefer using pip:

pip install openharness-ai

Step 3: Locate Your Virtual Environment 🔍

Once installed, OpenHarness automatically creates an isolated virtual environment inside your user folder:

C:\Users\<YourUserName>\.openharness-venv

Step 4: Activate the Environment 🐍

Open your favorite terminal (Command Prompt, PowerShell, or VS Code terminal), navigate to the script directory, and activate the virtual environment:

cd C:\Users\<YourUserName>\.openharness-venv\Scripts\
.\activate

Step 5: Launch OpenHarness CLI 🖥️

Run the CLI command to make sure everything is fired up and working properly:

openh

You should see the interactive OpenHarness CLI screen ready for action! 🎉

Step 6: Connect to GitHub Copilot 🔐

Let’s hook up OpenHarness to leverage your GitHub Copilot models. Run:

openh setup

1. Select GitHub as your provider.
2. The CLI will prompt you to authenticate via your browser.
3. Log in with your corporate or personal GitHub account, grant device authorization, and boom—you’re connected to Copilot’s models! 🔑

---

🏗️ Generating Complete Architecture & Specs

Now for the fun part: turning a vague requirement into clean, actionable project documentation.

1️⃣ The Business Analysis Pass 📊

Feed your BRD into OpenHarness and prompt it to parse out the critical domain specifics:

Prompt Goal:
Identify: Actors, Functional & Non-Functional Requirements, Integrations, Security, and Reporting.
Generate: Epics, Features, and User Stories.
Output: BRD Summary, Feature List, User Story Backlog, and Acceptance Criteria.

2️⃣ The Architecture & Design Pass 🏛️

Next, tell OpenHarness to construct the complete technical blueprint based on your tech stack:

Tech Stack:
🔹 Frontend: Angular
🔹 Backend: Spring Boot
🔹 Database: Oracle
🔹 Security: LDAP + OAuth2
🔹 Deployment: OpenShift
Generate: HLD (High-Level Design), LLD (Low-Level Design), Sequence Diagrams, Module Designs, and API Inventories.
Output Folder: /architecture

Once OpenHarness dumps these crisp Markdown (.md) spec files into your project, point GitHub Copilot at the documentation and ask it to write the code. Copilot will perform infinitely better because it now has full context! 🎯

---

🤔 Why OpenHarness + GitHub Copilot?

If GitHub Copilot already writes code, why do you need OpenHarness?

Think of GitHub Copilot as a master developer, and OpenHarness as the entire engineering department. When you have complex systems, a single prompt isn’t enough. OpenHarness orchestrates specialized AI Agents, each focusing on a distinct phase of the software development lifecycle:

Agent Role 👤	Core Responsibility 📋
Business Analyst Agent	BRD Summaries & Domain Mapping
Product Owner Agent	Features & User Stories
Architect Agent	HLD, LLD & Tech Stack Alignments
Design / API Agent	Swagger / OpenAPI Contracts & Diagrams
Developer Agent	Code Generation & Scaffolding
Testing / QA Agent	Test Cases & Integration Tests
Security Agent	Vulnerability & Architecture Reviews

---

💡 The Recommended Workflow

Instead of making Copilot do all the heavy lifting from scratch, divide and conquer:

📋 Business Requirement Document (BRD)
                  │
                  ▼
        🤖 OpenHarness Agents
        ├── 👔 BA Agent        ➡️ Generates User Stories & Acceptance Criteria
        ├── 📐 Architect Agent ➡️ Generates HLD, LLD, & API Contracts
        │
                  ▼
      🤖 GitHub Copilot Agent
        └── 💻 Dev Agent       ➡️ Generates Spring Boot & Angular Code
        │
                  ▼
        🤖 OpenHarness Agents
        ├── 🧪 QA Agent        ➡️ Generates Unit & Integration Tests
        └── 🛡️ Security Agent  ➡️ Conducts Code & Doc Reviews

TL;DR 📌

• Use OpenHarness to generate well-structured .md documentation, architecture blueprints, and user stories.
• Feed those .md docs into GitHub Copilot to write production-grade code.

Happy coding! 🚀✨

How to Configure OpenClaude with GitHub Copilot 🚀

Want to harness the power of GitHub Copilot right inside OpenClaude? Integrating these two tools gives you a seamless environment for planning, coding, testing, and asking complex questions about your repository.

Follow this quick guide to set up GitHub Copilot as your primary LLM in OpenClaude. 💡

---

Step 1: Install OpenClaude 📦

First, grab the latest version of OpenClaude directly from npm.

npm install -g @gitlawb/openclaude@latest

Check out the OpenClaude GitHub Repository for additional docs and source code.

---

Step 2: Connect GitHub Copilot 🔑

Once the installation is complete, you need to configure OpenClaude to talk to your GitHub Copilot models.

Run the onboarding command:

/onboard-github

Next, follow the interactive authentication steps:

1. Select the Login via Browser option when prompted.
2. A unique device code will appear in your terminal. Copy it! 📋
3. Open the provided GitHub authorization link in your browser, enter your code, and authenticate your account.

---

Step 3: Verify and Start Coding 🎯

After authorizing, confirm that your Copilot models are successfully loaded into OpenClaude by checking the model list in your terminal:

/models

You should now see the GitHub Copilot model available and active!

What’s Next? 🛠️

You’re all set! You can now leverage GitHub Copilot’s capabilities directly inside OpenClaude for:

• 🏗️ Planning project architectures
• 💻 Development & code generation
• 🧪 Testing & debugging
• ❓ Q&A on your codebase

---

Have you tried using OpenClaude with other model providers? Let us know your favorite setup in the comments below! 👇

🚀 How to Install GSTACK for GitHub Copilot (Step-by-Step Guide)

Supercharge your GitHub Copilot setup with GSTACK! Whether you are working on Linux, macOS, or Windows, here is the exact guide to getting it up and running smoothly.

---

📥 Step 1: Clone and Run Initial Setup

Open your terminal and run the one-liner below to fetch the shallow clone and start the setup script:

git clone --single-branch --depth 1 https://github.com/garrytan/gstack ~/.copilot/skills/gstack && cd ~/.copilot/skills/gstack && ./setup

---

⚙️ Step 2: Running Setup on Windows

Since the ./setup script is a Bash script, Windows users will need a Bash-compatible environment to execute it properly. Pick whichever option fits your workflow best:

🔹 Option 1: Git Bash (Simplest & Recommended)

1. Open Git Bash.
2. Navigate to your installation path and give the script execution permissions:

cd /c/Users/YOUR_USERNAME/.copilot/skills/gstack
chmod +x setup
./setup

---

🔹 Option 2: Run directly from PowerShell

If you prefer staying inside PowerShell, call Bash directly using:

bash -lc "cd /c/Users/YOUR_USERNAME/.copilot/skills/gstack && ./setup"

---

🔹 Option 3: Using WSL (Windows Subsystem for Linux)

If you operate primarily inside WSL, run:

wsl bash -lc "cd /mnt/c/Users/YOUR_USERNAME/.copilot/skills/gstack && ./setup"

---

⚠️ Troubleshooting & Prerequisites

💡 Prerequisite Check: If the script fails with an error stating bun is required but not installed, install Bun first before retrying the setup.
You can install Bun via terminal:

curl -fsSL https://bun.sh/install | bash

Once installed, re-run ./setup and you'll be good to go! 🎉

Open github copilot window and cross verify hitting /office-ho

🚀 Setting Up Addy Osmani’s agent-skills in VS Code with GitHub Copilot on Windows

If you are looking to bring agentic workflows and structured AI engineering to your local development environment, Addy Osmani's agent-skills project is a fantastic place to start.

In this guide, we will walk through setting up agent-skills step-by-step on Windows inside VS Code with GitHub Copilot, enabling you to run specialized commands for specification, planning, building, testing, and shipping code.

---

📋 Prerequisites & Tools

Before diving in, make sure you have:

• 💻 Windows OS with Git installed.
• 🛠️ VS Code configured with the GitHub Copilot and GitHub Copilot Chat extensions.

---

🛠️ Step 1: Clone the agent-skills Repository 📥

First, pull down the official agent-skills repository to your local machine:

# Create or navigate to your source directory
cd C:\Github-repo\

# Clone the agent-skills repository
git clone https://github.com/addyosmani/agent-skills.git

---

📁 Step 2: Create Your Target Workspace 🏗️

Next, create a new directory for the workspace or project where you plan to build with Copilot:

# Create your new project folder
mkdir C:\vscode-agentic-framework
cd C:\vscode-agentic-framework

---

⚙️ Step 3: Copy the Agentic Framework into .github/ 📂

To allow GitHub Copilot to recognize your agent skills and commands in VS Code, create a .github directory inside your project folder and copy the core skill files into it:

1. Create a .github folder in C:\vscode-agentic-framework\:

mkdir C:\vscode-agentic-framework\.github

2. Copy the following folders and files from C:\Github-repo\agent-skills\ into C:\vscode-agentic-framework\.github\:

• 🤖 agents/
• 📝 AGENTS.md
• ⚡ commands/
• 📖 docs/
• 🧪 evals/
• 🪝 hooks/
• 📚 references/
• 📜 scripts/
• 🛠️ skills/

---

📄 Step 4: Add Your Product Requirements 📝

Agentic workflows rely on clear context. Create a dedicated documents directory in your project root to hold your product specification or requirements:

# Create a documents folder
mkdir C:\vscode-agentic-framework\documents

Place your requirements file (e.g., product_requirement.txt or prd.txt) inside C:\vscode-agentic-framework\documents\.

---

🚀 Step 5: Execute the Agentic Workflow Sequentially

Once your workspace context is in place, open C:\vscode-agentic-framework in VS Code. You can now leverage Copilot Chat to execute the standard lifecycle commands sequentially.

📊 Agentic Workflow Commands

Action	Command	Key Principle	File Reference Context
1. Define what to build	/spec	Spec before code	@file:spec.toml @file:prd.txt
2. Plan how to build it	/plan	Small, atomic tasks	@file:planning.toml
3. Build incrementally	/build	One slice at a time	@file:build.toml
4. Prove it works	/test	Tests are proof	@file:test.toml
5. Review before merge	/review	Improve code health	@file:review.toml
6. Audit performance	/webperf	Measure before you optimize	@file:webperf.toml
7. Simplify the code	/code-simplify	Clarity over cleverness	@file:code-simplify.toml
8. Ship to production	/ship	Faster is safer	@file:ship.toml

---

💡 Best Practices

• 🎯 Keep tasks focused: Never skip straight to /build without running /spec and /plan first.
• 🔁 Iterate in slices: Run small validation loops (/build ➔ /test ➔ /review) before finalizing features.
• 📌 Maintain explicit context: Always reference your relevant TOML configurations and PRD files using Copilot's @file attachment mechanism.

Happy building! 🎉