AGENT 01
GPT-5.6 Sol · Cyber
OpenAI
independent campaign
multi-agent penetration testing
Five agents. One target.
Authorize one target. Five models each run the same pentest independently, so different approaches get a real chance to find what one agent could miss.
Approved-target testing only. Every agent stays inside the scope you authorize.
ai-hardening / campaign5 independent runs
01
GPT-5.6 Sol · Cyber
OpenAI
02
Claude Fable 5
Anthropic
03
Kimi K3
Moonshot AI
04
GLM-5.2
Z.AI
05
Gemini 3.1 Pro
Google
target
approved
the agent roster
Different models take different paths through the same scope.
Each agent receives the same approved target, credentials, and time budget. The model changes; the rules do not.
AGENT 02
Claude Fable 5
Anthropic
independent campaign
AGENT 03
Kimi K3
Moonshot AI
independent campaign
AGENT 04
GLM-5.2
Z.AI
independent campaign
AGENT 05
Gemini 3.1 Pro
independent campaign
one authorization, five runs
01
Approve one scope
Set the target, credentials, and authorization once. Every agent receives the same boundaries.
02
Run five independent pentests
Each model gets a separate run against the target rather than inheriting another model’s output.
03
Review one campaign
Review findings from every completed agent in one workspace, with evidence kept inside the existing issue flow.
ai hardening
Put five agents on your target.
Define the approved scope once, launch the full roster, and review the campaign from your dashboard.
Open AI Hardening