Secure Agent Framework
A secure-by-default Python agent framework — 10 controls on the agent's own execution seams, each mapped to OWASP Agentic / MAESTRO and proven by a runnable attack it blocks.
PythonOWASP AgenticMAESTROA2A
~$ ./janatan
I build AI agents.
Then I break them.
I'm Janatan Tajik, a software developer and offensive-security engineer, and the founder of Janreth. I think like an attacker across apps, cloud, blockchain, and AI, and I build the systems I break.
janatan@janreth: ~
~$ ▋
What I do
Where I focus across the modern attack surface.
Offensive security & red teaming
I test systems the way an attacker would: penetration testing, red teaming, and exploit development.
Blockchain & smart-contract security
Adversarial review of on-chain systems and smart contracts, before they ship.
AI & agentic-AI security
Red-teaming AI agents and the systems around them, the newest attack surface.
Security tooling & development
I build the tools, not just run them, from offensive tooling to multi-agent systems.
The company
I founded Janreth
Janreth is a cybersecurity company. Its first product red-teams AI agents and maps the risk to the GCC's actual obligations. It's the productization of everything on this site.
Visit JanrethFeatured projects
A few things I've built and broken. Code on GitHub.
Blockchain in Rust
A from-scratch Bitcoin-style blockchain in Rust — SHA-256 proof-of-work, ECDSA-signed transactions, and multi-node HTTP peers syncing the longest valid chain.
RustProof-of-WorkECDSAP2P
Linux Kernel Blocker
A kernel module that hooks the Linux reboot() syscall and silently blocks it.
CLinux KernelSyscall Hooking
Let's talk
Open to collaborations, security work, and conversations about agentic-AI risk. If that's your world, reach out.