YushaCyber

What it is

A learning platform for cybersecurity, built as a single Flask application and organised into fourteen registered blueprints — authentication, dashboard, roadmap, CTF, labs, resources, profiles, leaderboard, admin, analytics, community, AI endpoints, an in-browser terminal and a missions interface.

It is the largest thing in this archive by some distance: around 3.4 MB of Python, plus the templates, stylesheets and client JavaScript that go with it.

Why I built it

I wanted the thing I was learning from to be the thing I was building. Working through security material and building the platform that teaches it are the same activity done twice, and doing both meant every topic had to be understood well enough to turn into an exercise someone else could complete.

How it works

The application factory registers each area of the platform as its own blueprint under its own URL prefix, so a feature is a directory rather than another branch in a growing file. Models sit behind SQLAlchemy with Alembic migrations, sessions are handled by Flask-Login, and the whole app is containerised with a Dockerfile and a Compose file.

Two decisions in it are security decisions rather than convenience ones, which matters for a platform about security:

• CSRF protection is global, via Flask-WTF, rather than applied per form.
• Rendered markdown is sanitised with bleach, against an explicit tag-and-attribute allow-list, before it reaches a page. Lesson content and CTF challenge descriptions are markdown, so the lesson viewer is an injection surface.

What is actually built

Four things worth naming precisely, because they are the parts most easily overstated:

• Progression is real. A user row carries xp and level; challenges and mission objectives carry an xp_reward; an achievement engine reads those totals and a leaderboard ranks them.
• The CTF arena has a schema behind it — challenge categories, challenges, per-user solves and staged hints — rather than a page of static puzzles.
• The AI mentor is a client, not a model. “CyberMentor” is a chat endpoint in front of a provider interface with OpenAI and Anthropic implementations and a mock. It needs a key in the environment; without one the feature reports itself unavailable. Nothing about it runs locally, and nothing about it is trained here.
• The browser terminal and the missions UI are their own blueprints, added after the platform’s first shape and registered alongside the rest.

What I learned

That the tests are where a platform like this actually gets built. The repository carries 48 test files, and the largest of them are the ones covering the lab content itself — SQL injection, XSS, CSRF, file-upload security, Wireshark, nmap, network reconnaissance, forensics, SOC workflow. Writing a test for a lab forces the lab to have a correct answer, which forces the topic to be understood rather than paraphrased.

Status

Active development. It runs locally and in Docker; it is not deployed, and it has no users to speak of. Nothing here should be read as a live service.