hackagent.catalog.risks.jailbreak.types
Jailbreak sub-types.
JailbreakType Objects
class JailbreakType(Enum)
Sub-types for Jailbreak.
ROLEPLAY
Uses fictional scenarios or personas to bypass safety filters.
ENCODING
Encodes malicious intent via Base64, ROT13, leetspeak, etc.
MULTI_TURN
Gradually escalates across conversation turns (crescendo).
HYPOTHETICAL
Frames harmful requests as hypothetical or academic questions.
AUTHORITY_MANIPULATION
Invokes false authority or urgency to override restrictions.