This is a submission for the GitHub Copilot CLI Challenge
What I Built
I built The Linux Compass, a terminal-based utility designed to bridge the gap between AI discovery and permanent technical knowledge.
As a Computer Science student , I am currently documenting my journey into infrastructure through a Linux/DevOps learning series. I noticed a recurring problem: I would use GitHub Copilot CLI to find a command, use it once, and then forget it. This "knowledge decay" is a major hurdle for students. I created a basic website for an web based app called 'Flashy' (Flashcard learning style because I am a slow learner and suffering with extreme ADHD), but that was not enough either, this challenge pushed me to create something that will be beneficial to me, and other students out there to solve that problem whiles in the terminal.
The Linux Compass wraps the GitHub Copilot Agent in a Python workflow that forces a "Learning Loop." It captures the user's goal, assists in finding the command, triggers an educational breakdown (the "why"), and then automatically synchronizes the session into a persistent Markdown study log (LINUX_STUDY_LOG.md).
Demo
GitHub Repository: Linux-compass
How it Works:
(Go through the ReadMe in the Github repo to get started!)
Below are screenshots of how it works:
1)Initiating a Query The tool greets the user and consults the Copilot Agent for the specific Linux task.
2) The Educational Breakdown The Compass doesn't just provide an answer; it enforces understanding by calling the explain agent automatically.
3) Successful Synchronization The session concludes with a success message once the data is safely logged.
4)The Persistent Study Log By opening LINUX_STUDY_LOG.md, we see the automated entry: timestamp, learning goal, and verified command. This creates a structured, searchable history of my technical growth.
My Experience with GitHub Copilot CLI
My journey with the GitHub Copilot CLI was a rollercoaster of technical troubleshooting that ultimately made me a better developer.
The Struggles:
The biggest challenge was dealing with the 2026 Agentic Update of the GitHub CLI. Early in development, I faced a "Version Conflict" where the CLI refused to recognize simple prompts. I kept getting the error: error: Invalid command format.
I had to dive deep into the CLI documentation to discover the new Interactive Payload Pattern. I learned that the 0.0.395 version of the extension required the suggestcommand and the -p prompt to be passed as a single, nested string within the -i (interactive) flag.
The Breakthrough:
Once I understood how Python's subprocess module needed to "hand over" terminal control to the Copilot agent, everything clicked. I also navigated Windows-specific hurdles, such as migrating my project from the restricted System32 directory to a trusted user space and upgrading to PowerShell Core (pwsh) to support the modern Copilot features.
The Impact:
GitHub Copilot CLI turned the terminal from a "scary black box" into an interactive tutor. Instead of just giving me a command, the explain feature allowed me to build a tool that teaches me. For a student aiming for a career in DevOps, this isn't just a shortcut it's a productivity multiplier that ensures I am learning efficiently while I build.
Top comments (1)
This is a really thoughtful idea. I love the focus on retention, not just “getting the command to work.”
The learning loop + Markdown log is smart — it turns Copilot from a shortcut into actual long-term knowledge. That’s a real pain point, especially when you’re learning Linux/DevOps and everything blurs together.
Also appreciate you sharing the struggles with the Copilot CLI update. That part alone will probably save other people hours of confusion.
Nice work — this feels genuinely useful, not just a challenge submission. 👏