ResearchBenchmarksRug Pull Prevention on LLM Skill-Security corpus 250 tampered artifactsRug Pull Prevention on LLM Skill-Security corpus 250 tampered artifactsFollow95.1Empirical AccuracySIGIL92.70893.32993.9594.571May 6, 2026Evaluation resultsMethodMethodLinksEmpirical AccuracyResistance StatusSIGIL2026.0595.1—AgentFuzz2026.0592.8—