Model Economy 0.6.1
English
Model Economy 0.6.1 tightens the local safety boundary and prepares a reproducible Skills-only directory package.
Highlights:
- Plain doctor now checks only managed files and Codex version availability; it no longer invokes the broader Codex diagnostic.
- Managed role, configuration, and state artifacts fail closed when their filesystem safety cannot be established.
- doctor --smoke remains an explicit authenticated opt-in that may consume usage; model and role identity remain unverified.
- A deterministic Skills-only packager now rejects app/MCP manifest fields and removes directory-unsupported screenshot metadata and files.
- Plugins Directory candidate materials now include current field limits, three starter prompts, five positive tests, three negative tests, and final submission gates.
This release does not add telemetry, a remote service, an authentication flow, or a fixed Token-saving claim. The OpenAI Plugins Directory package is prepared but has not been formally submitted.
简体中文
Model Economy 0.6.1 收紧了本地安全边界，并准备了可重复生成的 Skills-only 官方目录候选包。
主要变化：
- 普通 doctor 现在只检查受管理文件与 Codex 版本可用性，不再调用范围更广的 Codex 诊断。
- 六角色、配置和状态工件无法确认文件系统安全性时会失败关闭。
- doctor --smoke 仍是可能产生用量的显式认证运行；模型和角色身份继续保持未验证。
- 新增确定性的 Skills-only 打包器，拒绝 app/MCP manifest 字段，并移除目录不支持的截图元数据与文件。
- 官方目录候选材料补齐当前字段限制、三条 Starter Prompt、五个正向测试、三个负向测试和最终提交门槛。
本版本不引入遥测、远程服务或认证流程，不承诺固定 Token 节省比例。OpenAI Plugins Directory 候选包已经准备，但尚未正式提交。