DEV Community

Thejan Vishmitha
Thejan Vishmitha

Posted on

PR Whisperer: A Copilot-Powered CLI for Writing Better Pull Requests

#devchallenge #githubchallenge #cli #githubcopilot

GitHub Copilot CLI Challenge Submission

This is a submission for the GitHub Copilot CLI Challenge

What I Built

I built PR Whisperer (prw), a command-line tool that helps developers write better pull requests using GitHub Copilot CLI.

Instead of being “AI for the sake of AI”, PR Whisperer focuses on real developer workflows:

  • reviewing your own changes objectively
  • generating clear, structured PR descriptions
  • working even when you’re not on a feature branch

The goal was to make Copilot feel like a thoughtful PR partner rather than just a text generator.

PR Whisperer supports multiple review contexts:

  • traditional PRs (branch vs base)
  • pre-commit reviews (staged changes)
  • same-branch workflows (last commit vs previous commit)

All from the terminal, with zero browser context switching.

Demo

Repository:

https://github.com/HarryVishmitha/prw-whisper

Example commands:

prw doctor
prw review --commit --save
prw describe --commit --save
prw init
Enter fullscreen mode Exit fullscreen mode

What happens when you run it

  • prw review generates a structured PR review (summary, risk, issues, suggestions, checklist)
  • prw describe drafts PR titles and a clean GitHub-ready PR description
  • --save writes Markdown reports to .prw/reports/
  • prw init runs copilot init and generates repository instructions automatically

Screenshots and CLI output examples are available in the repository README.

Sample Screenshots

My Experience with GitHub Copilot CLI

This project was built around GitHub Copilot CLI — not just with it.

I used Copilot CLI in three intentional ways:

  1. AI-powered reviews

    Copilot analyzes Git diffs and generates structured, human-readable PR feedback.

  2. PR description generation

    Copilot drafts PR titles and PR descriptions using a consistent template

    (What / Why / How / Testing / Risk).

  3. Repository initialization

    prw init wraps copilot init to generate .github/copilot-instructions.md, documenting architecture, conventions, and CLI behavior.

I intentionally avoided calling Copilot when it wasn’t needed:

  • PR Whisperer detects “no diff” scenarios and guides the user instead of wasting AI calls
  • All Copilot usage is non-interactive and scripted, making it reliable and repeatable

This made Copilot feel less like a chatbot and more like a well-integrated CLI collaborator.

Why This Project Is Different

Most PR tools assume:

“You always work on feature branches.”

PR Whisperer supports:

  • solo developers
  • same-branch workflows
  • pre-commit reviews
  • quick self-reviews before opening a PR

It’s not flashy — it’s practical, and designed to fit into how developers actually work.

Final Thoughts

Building PR Whisperer showed me that the real strength of GitHub Copilot CLI isn’t just code generation — it’s augmenting developer judgment.

This project reflects how I’d genuinely want to use Copilot in my own workflow:

  • intentional
  • context-aware
  • respectful of developer time

Thanks for the challenge — this was a great opportunity to explore what AI-powered CLI tools can really be.

Top comments (0)