Mahesh

Posted on Feb 10 • Edited on Feb 12

RepoSentry — Take your app to production, understand flow, security vulnerabilities. — Why read if you can visualize it.

#devchallenge #githubchallenge #cli #githubcopilot

GitHub Copilot CLI Challenge Submission

This is a submission for the GitHub Copilot CLI Challenge

What I Built

RepoSentry is an open-source CLI that answers the question every developer eventually hits:

“How do I understand this codebase fast enough to ship safely?”

Point it at any repository and it generates a full intelligence report:

Docs suite (README, API docs, setup, contributing, changelog, FAQ)
Architecture diagrams (Mermaid: dependency graph, data flow, ERD, API flow)
Security audit (pattern scan + Copilot analysis + threat model diagram)
CI/CD suggestions (GitHub Actions + Docker / Compose + production guide)
API testing assets (Postman collection + shell script + coverage report)
Performance anti-pattern scan + audit
Team templates (PR template, issue templates, CODEOWNERS, onboarding)
A weighted Health Score (A+ → F) with history + compare view

The goal is to make RepoSentry something you can clone and use immediately on any repo:

onboarding new teammates
evaluating OSS dependencies before adopting them
standardizing baseline docs/security/CI across multiple projects
creating a “before vs after” health score as you improve quality

RepoSentry is also intentionally demo-friendly (for judges): it has a preview server UI, progress spinners, and a clear final score summary.

Demo

GitHub repo: https://github.com/MaheshDoiphode/reposentry
npm package: https://www.npmjs.com/package/reposentry

Zero-wait judge demo (no analysis needed)

To make the demo instant, this repo includes a pre-generated .reposentry/ folder checked into git.

npm i -g reposentry
git clone https://github.com/MaheshDoiphode/reposentry.git
cd reposentry
reposentry serve
# open http://localhost:3000

Quick demo (3 minutes)

# 1) install
npm i -g reposentry

# 2) run analysis in any repo
cd path/to/any-repo
reposentry analyze --force

# 3) open the UI
reposentry serve
# open http://localhost:3000

“Judge mode” demo (shows uniqueness fast)

1) Run RepoSentry once

reposentry analyze --force

2) Make a small improvement (add a README section, add CI, add tests)

3) Run again

reposentry analyze --force
reposentry compare

4) Open the compare UI

reposentry serve
# click “Compare Scores”

Screenshots

Full analysis run output:
Copilot-powered fixes:

My Experience with GitHub Copilot CLI

This challenge asked us to build an application using GitHub Copilot CLI.

RepoSentry uses Copilot CLI in two ways:

1) Copilot CLI as the development partner

I used Copilot CLI during development to:

iterate on the CLI UX (commands, flags, interactive mode)
design a modular “engine” architecture (docs / security / CI / tests / etc.)
harden real security issues (output overwrite protection, server path traversal defenses, markdown sanitization)
write focused tests (Vitest) for the tricky parts

The biggest win: I could stay in the terminal, ask for a plan, refine prompts, and immediately turn those results into code with fast iteration.

2) Copilot CLI inside the product (the core superpower)

RepoSentry is not just “built with Copilot CLI” — it’s powered by it.

Under the hood, RepoSentry:

scans your repository (languages/frameworks, routes/models/imports, git history)
builds a structured prompt context (file tree + detected signals)
calls Copilot CLI to generate outputs for each engine
writes results into a consistent .reposentry/ folder

Example outputs you can open immediately:

HEALTH_REPORT.md + analysis.json
ARCHITECTURE.md + diagrams/*.mmd
security/SECURITY_AUDIT.md + security/threat-model.mmd
infrastructure/ci.yml and deployment guidance

Reliability + safety choices (important for real repos)

RepoSentry never asks Copilot to write into your repo directly.
It writes generated files only into the configured output directory.
The preview server blocks path traversal and escapes raw HTML in markdown.
If Copilot CLI isn’t installed/logged-in, RepoSentry still runs and produces placeholder outputs.

How to Run It (for judges)

npm i -g reposentry

# Instant preview (works immediately in this repo because .reposentry/ is committed)
git clone https://github.com/MaheshDoiphode/reposentry.git
cd reposentry
reposentry serve

# Generate fresh output for any other project
# cd path/to/any-repo
# reposentry analyze --force
# reposentry serve

If Copilot CLI isn’t configured yet:

Install Copilot CLI: npm i -g @github/copilot (or winget install GitHub.Copilot on Windows)
Authenticate: run copilot once and use /login

Some Demo Screenshots of the reposentry repo itself

Top comments (7)

Gaëtan Compigni • Feb 12

Thanks for the work, very good project.
It would be great if it were compatible with other AIs just by defining keys in env for those who don't have a copilot subscription.

Mahesh • Feb 12

Thanks for the feedback.. Sounds good.. I will add some common ai providers openai, gemini, ollama too for running it locally..

Allan • Feb 10

It looks cool from the outside but have you tried this on a really large legacy codebase yet? What happened?

Mahesh • Feb 10

I tried it on some ANT app projects I found on internet.. it was very old.. Tried on it, and it worked nicely.. Although I found that ANT java project copilot is not that good, i mean sure copilot works perfectly for modern tech... hell i created this project only with github copilot cli just in like 5 hrs or something(only the planning step took 4 hrs 😉)

Delacruz • Feb 10

heyy, have you tried it out ?? or if anyone else has tried this, drop your results... I’d love to see how it performs!

Nisha • Feb 10

I tried it out on the same github repo.... it works perfectly..

Absolutely loved the Front end website it creates for serving .. the theme, the font and stuff..... but best thing i loved is still the mermaid diagrams ..