Every lab starts with a concept.
Animated 5-beat tutorials in the sizzle-reel visual language — 10 hands-on labs plus 6 Model reels on machine learning and prompt engineering. Pick a discipline, watch the system flow, then press Spawn lab when you're ready to break it for real.
Linux Terminal Basics
Hunt for a hidden flag using ls, find, grep, and cat in a minimal Ubuntu shell.
Networking & Packet Analysis
Inspect a captured pcap to identify the attacker's IP and the port used for a reverse shell.
DNS TXT Chain Reconnaissance
Follow a base64-encoded TXT-record chain across an internal zone to recover the flag.
Classic SQL Injection: Login Bypass
Bypass authentication on a naive PHP/SQLite form via tautology-based SQL injection.
Stored XSS Cookie Theft
Exploit a stored XSS bug to steal the admin session cookie — the cookie value is the flag.
Password Cracking Fundamentals
Identify and crack three password hashes (MD5, SHA-1, bcrypt) using john and hashcat.
Burp Suite Cookie Tampering
Use Burp's intercepting proxy in a noVNC desktop to tamper a base64 role cookie.
Web Recon: Hidden Assets
Enumerate a misconfigured web server: backups, leaked config, and a dev vhost ending in dotenv disclosure.
SOC Log Triage & Intrusion Timeline
Pivot across four log sources to reconstruct an end-to-end intrusion: brute force, login, data read, exfil.
AI Prompt Injection
Attack a FastAPI chatbot told never to reveal a secret. Probe for OWASP LLM01 vulnerabilities.
ML 101 · Train Your First Classifier
Watch a no-code trainer turn labelled rows into a model, then grade it server-side against held-out answers it never sees.
Classify the Malicious Sessions
Binary classification on network sessions: four features per row — pkt_rate, bytes, failed logins, entropy — to flag benign vs. malicious.
Triage Vulnerabilities by Severity
Step up from binary to multiclass: low / medium / high. See why one global accuracy can hide the class your model keeps missing.
Forecast Resolution Time
Swap classification for regression: predict the hours to clear a support backlog on a continuous scale, graded by RMSE.
Prompt Engineering Foundations
Turn a fuzzy language model into a reliable component. Write a prompt, then watch an LLM judge grade it against a hidden rubric.
Prompt Engineering · Reliable JSON
Pull a name and email out of free text and return strict JSON. The reliability lives in the constraints — make every one explicit.