Unverified50% confidenceFactExact time
In its best-practices document for Agent design, Anthropic recommends that a tool's error response should include sufficient context to enable the model to self-correct without re-querying the user.
1
Sources
50%
Confidence
Long-term
Relevance
8/31/2026
First Seen
Sources
Related Entities
Related Claims
Unverified工具设计应遵循动作幂等、错误明确的原则，否则模型调用失败无法恢复，工具描述质量直接决定Agent成败70% similarUnverified判断AI方案是否可靠的实用问题包括团队是否有针对幻觉的监控机制以及出错时的降级策略，这些问题的缺席本身就是话术包装的指示信号69% similarUnverifiedAnthropic在其Claude的设计文档中将可纠正性机制称为'corrigibility'，视为安全AI系统的核心属性之一68% similarUnverifiedAnthropic在请求体中额外要求 anthropic-version 请求头，对系统提示的传递方式有独立规范68% similarUnverified该项目引入主动消歧机制：当不确定性超过阈值时智能体不直接输出，而是主动向用户提出澄清问题68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/827561API
curl https://kongchang.com/api/v1/knowledge/claims/827561MCP
get_claim(id=827561)