Prompt Refiner is a lightweight Windows desktop application designed to enhance and refine user prompts by leveraging large language models (LLMs) via the GitHub Copilot SDK. It provides a streamlined tool for users to input a basic prompt, process it through an LLM, and receive a polished, comprehensive version.
Key Features:
Real-time streaming output: View refined prompts as they are generated in real time.
Dynamic model selection: Choose from available models to best suit your needs.
Headless CLI support: Operate seamlessly in scriptable environments or pipelines.
Debug mode: Troubleshoot issues without exposing sensitive data.
Copy-able error dialogs: Access detailed error information for easier troubleshooting.
Audience & Benefit:
Ideal for developers, technical writers, educators, and anyone working with AI-driven tools. Prompt Refiner helps users generate more precise, effective prompts quickly, saving time and improving outcomes in prompt-based workflows. It can be installed via winget for convenient setup.
README
Prompt Refiner
A lightweight Windows desktop application (.NET 8, WinForms) that takes a rough prompt, sends it to a large language model via the GitHub Copilot SDK, and returns a polished, comprehensive version. Supports both a graphical interface and a headless CLI.
Supported Platforms
Platform
Status
Windows 10/11 (x64)
✅ Supported
macOS
❌ Not supported (WinForms)
Linux
❌ Not supported (WinForms)
Prerequisites
Before using Prompt Refiner, ensure you have:
GitHub Copilot CLI — installed and available on PATH.
# Install via npm
npm install -g @anthropic-ai/sdk
# Or via the GitHub CLI extension
gh extension install github/gh-copilot
Active GitHub Copilot subscription — any tier (Free, Individual, Business, Enterprise). The SDK bills against your Copilot premium request quota.
Authentication — sign in once before first use:
gh auth login
The app auto-discovers credentials from (in order): COPILOT_GITHUB_TOKEN, GH_TOKEN, GITHUB_TOKEN environment variables, or the Copilot CLI's cached login.
This verifies the GitHub Copilot CLI and GitHub CLI are installed and checks for auth tokens.
Usage
GUI Mode
Double-click PromptRefiner.exe (or run it with no arguments):
Type or paste a rough prompt into the input area.
Select a model from the dropdown (default: claude-opus-4.6). The list is fetched live from Copilot on startup.
Click Refine ➜ (or press Ctrl+Enter).
A streaming output window opens showing the refined prompt in real time.
Edit the result if needed, then click 📋 Copy to copy to clipboard.
CLI Mode
Run with arguments to operate headlessly — ideal for scripts and pipelines.
# Refine a prompt directly
PromptRefiner.exe -p "write a python script to sort a list"
# Read prompt from a file
PromptRefiner.exe -f prompt.txt
# Pipe from stdin
echo "help me debug my code" | PromptRefiner.exe
# Use a specific model
PromptRefiner.exe -p "design a REST API" -m gpt-4o
# Save output to a file (also prints to stdout)
PromptRefiner.exe -p "explain recursion" -o refined.txt
# List all available models
PromptRefiner.exe --list-models
# Show help
PromptRefiner.exe --help
CLI flags:
Flag
Description
-p, --prompt
Prompt text to refine
-f, --file
Read prompt from a file
-m, --model
Model to use (default: claude-opus-4.6)
-o, --output
Write refined prompt to file
--list-models
Print available models and exit
--check-prereqs
Verify prerequisites and exit
-h, --help
Show usage help
Exit codes:0 = success, 1 = error. Status messages go to stderr; refined output goes to stdout.
Build from Source
git clone https://github.com/PranavPeshwe/prompt-refiner.git
cd prompt-refiner
dotnet build prompter2.sln
Running Tests
# Unit tests only (no network, no auth needed)
.\Run-Tests.ps1 -Filter Unit
# Startup benchmark tests
.\Run-Tests.ps1 -Filter Startup
# Integration tests (requires Copilot auth + subscription)
.\Run-Tests.ps1 -Filter Integration
# All tests
.\Run-Tests.ps1 -Filter All
# Standalone startup measurement (10 iterations with bar graph)
.\Measure-Startup.ps1 -Runs 10
