A certifying, model-free, minimal-intervention runtime shield for LLM agents: it repairs an unsafe tool-call trajectory with the least-cost enactable edit instead of blocking, and certifies every intervention for re-verification without the model.
python security ltl temporal-logic formal-methods ai-safety interpretability runtime-verification model-free tamper-evident tsetlin-machine guardrails llm runtime-enforcement llm-agents agent-safety minimal-intervention shield-synthesis edit-automata certifying-algorithms
-
Updated
Jun 28, 2026 - Python