HackMyAgent Attack Mode

Attack Payloads

55 adversarial test payloads across 5 categories. Each payload tests a specific OASB-1 control and includes severity, CWE mapping, and remediation guidance.

CRITICAL17
HIGH19
MEDIUM11
LOW5
INFO3

Prompt InjectionPI

Attempts to manipulate agent behavior via malicious input

Tests OASB controls:3.13.23.3

JailbreakingJB

Attempts to bypass safety guardrails and restrictions

Tests OASB controls:3.14.1

Data ExfiltrationDE

Attempts to extract sensitive information from the agent

Tests OASB controls:4.35.28.2

Capability AbuseCA

Attempts to misuse agent tools and capabilities

Tests OASB controls:2.22.34.2

Context ManipulationCM

Attempts to poison agent context or memory

Tests OASB controls:8.18.2