Method and system for controlling the execution frontier of an artificial intelligence agent skill protected unit

Abstract

This invention discloses a method and system for controlling the execution frontier of protected units in an artificial intelligence agent skill. The construction end generates the execution position, allowed successors, and a plan commitment and packet hierarchy for mapping protected execution units. The state control end atomically generates unit usage control results based on the current frontier. The controlled execution component verifies the result, enabling the unit bound to the current position and packet hierarchy to enter a usable state, complete processing, and generate state transition evidence that cannot be effectively signed by a normal caller; in one embodiment, the unit is a skill-encrypted unit formed using authentication encryption. The state control end verifies the evidence and generates the next frontier based on the committed plan, preventing the old frontier, old control results, and caller-reported processing data from advancing the state. This invention is applicable to Skills, plugins, MCP tools, and workflow nodes.

Images