📄 中文摘要
传统软件依赖合同（如API、类型系统和断言）来规范和强制执行正确行为，而AI代理则通过提示和自然语言指令进行操作，缺乏正式的行为规范。这一差距导致了代理AI部署中的漂移、治理失败和频繁的项目失败。提出了代理行为合同（ABC）这一形式化框架，将设计-by-合同原则引入自主AI代理。ABC合同C = (P, I, G, R)将前提条件、保持不变性、治理政策和恢复机制作为一等公民的运行时可强制执行的组件进行规范。定义了(p, delta, k)-满足性，这是一种考虑概率的合同合规性概念。
📄 English Summary
Agent Behavioral Contracts: Formal Specification and Runtime Enforcement for Reliable Autonomous AI Agents
The research introduces Agent Behavioral Contracts (ABC), a formal framework that applies Design-by-Contract principles to autonomous AI agents. Traditional software relies on contracts such as APIs, type systems, and assertions to specify and enforce correct behavior, while AI agents operate on prompts and natural language instructions without formal behavioral specifications. This gap leads to drift, governance failures, and frequent project failures in agentic AI deployments. An ABC contract C = (P, I, G, R) specifies Preconditions, Invariants, Governance policies, and Recovery mechanisms as first-class, runtime-enforceable components. The concept of (p, delta, k)-satisfaction is defined, which provides a probabilistic notion of contract compliance.