▸ HANDS-ON DEFENSIVE SECURITY · OPEN SOURCE
9 small, local-first labs that teach the defensive skills teams need as they adopt AI agents, cloud identity, and modern software supply chains. Read the guide, run the script, break the fixture, and learn what the checks miss.
Most training stops at slides or needs a cloud account, a vulnerable VM, and an afternoon of setup. CyberSHIELD Labs fit in a lunch break: clone the repo, run one command, and see a real check find a real misconfiguration.
Standard-library Python only. No pip install, no Docker, no API keys, no cloud bill. If you have Python 3.9 or later, you can start now.
Every lab uses bundled synthetic data. Nothing touches production systems, and the one network lab refuses to talk to anything except your own machine.
Each lab spells out what its checks cannot see. Learning where a control stops is as important as learning what it catches.
Work through them in order or jump to the area you need. Every lab has a guide, a runnable script, a synthetic fixture, and a takeaway you can apply at work.
Flag instruction-like text in untrusted content and tool requests outside an allowlist.
$ python3 -m labs.ai_application_safety.analyze
→ OPEN LAB
Audit a provider-neutral policy for wildcard actions, wildcard resources, and public principals.
$ python3 -m labs.cloud_iam_posture.audit
→ OPEN LAB
Hunt synthetic auth logs for login-failure bursts and MFA push-fatigue patterns.
$ python3 -m labs.identity_threat_hunting.hunt
→ OPEN LAB
Detect and redact email addresses, phone numbers, and a demo key pattern in local text.
$ python3 -m labs.sensitive_data_protection.scan
→ OPEN LAB
Validate simulated backup evidence for encryption, immutability, and recent restore tests.
$ python3 -m labs.backup_resilience.validate
→ OPEN LAB
Check an SBOM-like inventory for unpinned versions, missing digests, and unknown sources.
$ python3 -m labs.software_supply_chain.audit
→ OPEN LAB
Statically review a Dockerfile for floating base images, root users, and embedded secrets.
$ python3 -m labs.container_configuration.audit
→ OPEN LAB
Estimate strength after discounting common words, keyboard runs, and repeated patterns.
$ python3 -m labs.password_strength.estimate
→ OPEN LAB
Check a local practice server for exposed .git/, backups, and admin paths. Loopback only.
$ python3 -m labs.web_exposure.check --demo
→ OPEN LABLab 9 starts its own throwaway practice server on 127.0.0.1, checks it for commonly exposed paths, and shuts it down again. Nothing leaves your machine.
GETTING STARTED GUIDE ↗Each lab follows the same three-step loop, so the habit carries over to real reviews.
One command against the bundled synthetic fixture. Read every finding and ask why it fired.
Fix the risky settings in a copy of the fixture, or add new ones, and rerun. Watch findings appear and disappear.
Every guide lists what the check misses. Try to sneak a risky input past it, then decide how a real control would catch it.
The labs are small on purpose: easy to read in one sitting, easy to extend, and easy to run in a classroom or a team session.
Go beyond definitions. See exactly what a wildcard IAM grant, a root container, or an MFA push-fatigue pattern looks like in data.
→ START WITH LAB 1Each script is under 200 lines. Read how the check works, then extend it for the configs and logs your team actually ships.
→ EXTEND A LABNo accounts to provision and no lab environment to tear down. Hand out the repo link and every attendee is running in minutes.
→ REVIEW SAFETY & SCOPEPractice treating retrieved content as untrusted, keeping tool permissions narrow, and redacting data before it reaches a prompt or a log.
→ OPEN THE AI SAFETY LABOpen source under Apache 2.0. Runs entirely on your machine. Built by CybersecurityOS.