Robustness Leaderboard

Model × category.

Robustness is the fraction of attack traces the agent left not compromised, so higher is safer. The numbers below are placeholder fixture data, not a claimed benchmark; each cell opens that run in the Live Attack Replay.

Weakest0.29Model CASI06 · Memory & Context Poisoning

.50 to .79Category names → ThreatsSOURCE · fixture

Swipe · 7 categories · model column stays pinned

Robustness per model and OWASP Agentic category. Fixture data, not a claimed benchmark. Each value cell links to that run in Live Attack Replay.
Model	ASI01 Agent Goal Hijack	ASI02 Tool Misuse and Exploitation	ASI03 Identity and Privilege Abuse	ASI04 Agentic Supply Chain Vulnerabilities	ASI05 Unexpected Code Execution (RCE)	ASI06 Memory & Context Poisoning	ASI10 Rogue Agents	Overall
Model A	0.92	0.81	0.79	0.74	0.55	0.38	0.86	0.72
Model B	0.88	0.69	0.63	0.55	0.44	0.52	0.79	0.64
Model C	0.71	0.48	0.52	0.41	0.33	0.29	0.63	0.48

Cell → its run in Live Attack Replay