Offensive testing
that doesn't wait for the annual audit
Offensive agents map your applications and perimeter, test authentication, permissions and business logic, and only report what they managed to exploit.
✓Included in the complete platform and available on its own
IDOR on other customers' orders
GET /api/orders/:id · CVSS 8.1 · CWE-639
Request · test customer B session
GET /api/orders/88412 HTTP/2 Host: tienda.acme.example Cookie: session=<customer B> Accept: application/json
Response · customer A's order
HTTP/2 200
{ "id": 88412,
"customer": "A",
"address": "C/ Example 12, Zaragoza",
"items": 7, "total": 64.20 }Agent-offensive
The ID is sequential and the API does not check the order belongs to the session. Tested with two test accounts, without touching real customer data.
Fix
Filter by the session's customer_id in OrdersController.show
Authorised scope
Signed by M. Ortega · valid until 31 Dec
Targets
tienda.acme.example✓
api.acme.example✓
pedidos.acme.example✓
203.0.113.0/28✓
pagos.acme.exampleExcluded
Window
Monday to Friday, 22:00 to 06:00 · max 20 requests per second
Allowed techniques
- Authentication and sessions
- Access control between users
- Payment and coupon logic
- Injection (SQL, commands, templates)
- Denial of service
- Social engineering
Validated findings
Round of 14 Oct · after deployment 2026.10.14-2
Routes tested
214
Hypotheses dismissed
1.106
Exploited
2
Fixed this month
6
Offensive agents test your applications the way an attacker would, every time they change. You only get what they managed to exploit, with the proof to reproduce it. No lists of possible vulnerabilities.
What it does
GET /api/orders/88412 · session B
HTTP/2 200
{ "id": 88412,
"customer": "A",
"address": "Zaragoza" }
Continuous testing
Every deployment or perimeter change can trigger a new round of tests, without waiting for the next window.
Validated findings
Each finding comes with the request, the response and the steps to reproduce it. No lists of possible vulnerabilities.
Business logic
The agents test permissions between users, payment flows and object access, not just known signatures.
Controlled scope
You set domains, time windows and allowed techniques. The agents check scope before every request.
In the console
Platform screens with data from Acme, the fictional company in the live demo.
IDOR on other customers' orders
GET /api/orders/:id · CVSS 8.1 · CWE-639
Request · test customer B session
GET /api/orders/88412 HTTP/2 Host: tienda.acme.example Cookie: session=<customer B> Accept: application/json
Response · customer A's order
HTTP/2 200
{ "id": 88412,
"customer": "A",
"address": "C/ Example 12, Zaragoza",
"items": 7, "total": 64.20 }Agent-offensive
The ID is sequential and the API does not check the order belongs to the session. Tested with two test accounts, without touching real customer data.
Fix
Filter by the session's customer_id in OrdersController.show
Authorised scope
Signed by M. Ortega · valid until 31 Dec
Targets
tienda.acme.example✓
api.acme.example✓
pedidos.acme.example✓
203.0.113.0/28✓
pagos.acme.exampleExcluded
Window
Monday to Friday, 22:00 to 06:00 · max 20 requests per second
Allowed techniques
- Authentication and sessions
- Access control between users
- Payment and coupon logic
- Injection (SQL, commands, templates)
- Denial of service
- Social engineering
How it works
01Set the scope
You list applications, environments, test credentials and the windows when testing is allowed.
- ✓Routes and forms mapped
- ✓Attack hypotheses
- ✓Exploitation within scope
02The agents test
They map the surface, form attack hypotheses and try to exploit them within scope.
- PT-412 · IDORExploited
- PT-405 · CouponRetesting
- PT-401 · Sign-inFixed
03You fix, it retests
The finding becomes a case for your team and is retested automatically after the fix.
Connected to the rest of the platform
Exposure management tells it which assets to test first, and every exploited finding raises the priority of alerts on that asset in the Agentic SOC.
See the complete platform [ → ]Other modules

