Unverified50% confidenceFactExact time
当模型输出工具调用语法后会停止响应，Harness执行工具并将输出追加到对话历史末尾，然后向同一模型重新发起请求
1
Sources
50%
Confidence
Long-term
Relevance
7/9/2026
First Seen
Sources
60行Python揭秘Claude Code Harness原理：AI编程工具性能差异的核心
bilibili7k的每日搬运7/7/2026
Related Claims
UnverifiedWhen an LLM model decides to call a tool, it generates an assistant message containing a tool_use block; after the tool finishes executing, the result is appended as a new message in tool_result format76% similarUnverifiedSystem Prompt在对话开始前由开发者注入，向模型声明身份、行为边界和回复风格，在每次对话中被放入上下文窗口最前端69% similarUnverified自定义工具（Custom Tool）可以将重复流程封装起来实现一键复用，绕过重复推理69% similarUnverified多轮对话中System消息设定行为基调和约束，user消息代表用户输入，assistant消息是模型的历史回复67% similarUnverifiedContext Mode使用Pre-tools钩子拦截每个工具调用，并在隔离的子进程中执行，只将提炼后的输出传入对话上下文67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/400638API
curl https://kongchang.com/api/v1/knowledge/claims/400638MCP
get_claim(id=400638)