CASE EVIDENCE / A RECORD

bedriyan / Medkit 使用 Claude Opus 4.7 (max) 处理软件工程任务执行

bedriyan / Medkit 公开的代码代理与软件工程案例，来源为公开代码库，复核于 2026-06-27T08:40:26Z。

原始记录：Medkit uses Claude Opus 4.7 as an attending grader for browser-based clinical training

Chinese Brief

中文案例导读

bedriyan / Medkit 公开的代码代理与软件工程案例，来源为公开代码库，复核于 2026-06-27T08:40:26Z。 Model Atlas 将它标记为 A 类证据，因为它同时具备具体使用者、具体任务、公开原始证据和可访问产物。 Model Atlas 不把 benchmark、教程、发布说明或集合页包装成真实案例。

厂商Anthropic / Claude

模型Claude Opus 4.7 (max)

任务类型代码代理与软件工程

审核状态auto_approved

任务

真实任务背景

这是一个围绕软件工程任务执行的真实任务，公开材料可以回溯到具体使用者和具体产物。原始资料写作：A browser-based ER and polyclinic clinical training simulator where learners speak with AI patients, order tests, diagnose, treat, and receive post-session assessment.

代码代理与软件工程公开代码库A 类可核验real_case

公开产物

公开材料提供公开代码、README 或项目配置，可用于核验任务结果、项目形态和模型绑定关系。原始资料写作：The public repository describes an OSCE-style simulator that produces AI patient interactions and an attending-style grading report covering communication, history-taking, and clinical reasoning with guideline citations.

模型作用

Claude Opus 4.7 (max) 在该案例中承担软件工程任务执行相关的生成、分析、编排或实现角色。原始资料写作：The README states that an attending physician powered by Claude Opus 4.7 watches and grades decisions after each session, and that the project was built using Claude Code with Opus 4.7.

风险边界

当前判断基于公开材料；若产物下线、仓库变更或模型参与比例仅来自作者自述，需要在引用前重新复核。原始资料写作：Evidence is from the project author's public GitHub README; it is a hackathon/open-source artifact, not an independently audited customer story. Medical cases are described as synthetic and not clinical claims.