Skip to content
InfoResearchPreprintLLM-specific

Guarding the Kingdom of Skill-Based Agents in the AI Dynasty of Computation: Control-Flow Integrity for Agent Skills

Published
Record updated
View JSON

Summary

SkillCFI carries control-flow integrity over to skill-based LLM agents. Its compiler derives agent instructions and an executable contract from one skill, and a deterministic runtime enforcer admits tool requests only on trusted evidence. On eight self-written skills across 770 sessions, it blocked all 240 violations that a permission allowlist admitted, though no public skills were accepted as written.