How the assessment works
-
Inventory
The agents in scope: what each one reaches and what it can do.
-
Threat model
Trust boundaries, tool permissions, and the damage if an agent is turned against you.
-
Testing
We test your agents against real attacks, with your permission: prompt injection, tool misuse, data leaks. You see what worked and what held.
-
Findings
Rated by severity, with a remediation plan in order of priority.
What you get
- An inventory of the agents in scope
- A threat model per agent
- Test results
- Findings rated by severity
- A remediation plan, in order of priority
- A one-page summary for leadership
Your team can run the plan on its own. If you'd like a hand, we can also help put the fixes in place.