Allow-listed commands only
Agents run only the commands in their tools.yaml. No free-form shell.
Watch three AI security agents and their Chief audit a domain. They map what it exposes, check its HTTPS security and check its email protection, then write a report each.
A teaching demo extracted from Cyber360, the agentic cybersecurity framework by Rohas Nagpal.
Chief: Ready. Confirm your authority over the target, then run the audit.
SimulationRecorded sample output for the reserved domain example.com. Nothing is scanned from this page.
An AI agent is an AI model set up to pursue a goal across several steps on its own. It decides what to do next, runs a tool, reads the result and goes round again. Here is S18 working out whether example.com can be spoofed.
Can someone send email pretending to be example.com?
These agents are not separate AI models. Each one is a folder of instructions, tools and knowledge that turns the model already running in Claude Code or Codex into a specialist. Tap a layer to open it.
Simplified from the H1 Reconnaissance agent.
An agent that can run commands needs limits it cannot talk its way around. Cyber360 builds three of them into every agent.
Agents run only the commands in their tools.yaml. No free-form shell.
Every command's raw output is saved, and every finding links to the line that proves it.
; <<>> DiG 9.18 <<>> TXT _dmarc.example.com ;; status: NOERROR, ANSWER: 1 ;; ANSWER SECTION: _dmarc.example.com. 300 IN TXT "v=DMARC1; p=none" ;; Query time: 24 msec
A blocked or failed check is reported as unknown, never as secure.
The plugin installs the Chief and all three agents. Point it only at a domain you own or have written permission to test. Reports and evidence stay on your computer.
setup the plugin from https://github.com/rohasnagpal/demo-domain-audit and run a full domain audit against example.com