Tool · curated 31 Jul 2026
Release Notes | Giskard Documentation
First reported giskard.ai
Coverage timeline
Single-source advisory — first reported, latest, and curated coincide.
Why it matters
Giskard Hub's TokenBreak probe gives defenders a runnable way to test whether their LLM agents' safety classifiers can be bypassed by tokenization-manipulation prompt injection, a known real-world evasion technique.
Giskard Hub UI release 2.5.0 adds a new TokenBreak security probe to its LLM red-teaming scanner, which tests whether an agent can be manipulated through obfuscated prompt injection by prepending characters to trigger words (e.g. 'ignore' → 'Aignore') to evade input classifiers while remaining interpretable to the LLM. The release also adds stateful multi-turn agent support and fine-grained probe selection for scans, and is based on HiddenLayer's TokenBreak attack research (arXiv:2506.07948).