## 新功能
- 新增可配置的宽限期，用于从可选 MCP 服务器发现工具。（#41199）
- 扩展现在可以在 MCP 工具结果传递给模型之前检查或替换这些结果。（#41202）
- 插件目录现在会合并每个仓库的配置，并报告无效的项目市场，而不会隐藏有效插件。（#41208）
## Bug 修复
- 在 TUI 会话之间保留恢复的权限配置文件，并防止 `/cd` 削弱沙箱限制。（#41192）
- 在切换模型或回退到其他模型时，确保工具可用性和推理力度保持正确。（#41195、#41206）
- 改进远程沙箱强制机制，使用执行器的实际主目录、操作系统和路径约定。（#41196、#41204、#41207、#41209）
- 在应用服务器响应中保留结构化的 MCP 工具和资源错误。（#41196）
- 将嵌套子代理的令牌使用量计入根目标预算。（#41183）
- 防止过时的 Guardian 分类在权限状态变更后继续授权操作。（#41196）
## 维护工作
- 为升级后的 stdin 审查和远程执行器 MCP 发现增加遥测数据。（#41189、#41205）
- 在缓慢或高度并发的 CI 环境下，提升 Guardian WebSocket 和核心 fixture 测试的稳定性。（#41191、#41194）
## 更新日志
完整更新日志：https://github.com/openai/codex/compare/rust-v0.150.0...rust-v0.151.0
- #41183 将子代理令牌使用量计入根目标 @copyberry
- #41189 为 stdin 审查大小检查添加埋点 @copyberry
- #41191 提升 Guardian WebSocket 测试的稳定性 @copyberry
- #41192 在 TUI 会话中保留恢复的权限配置文件 @copyberry
- #41193 报告远程插件同步所影响的能力 @copyberry
- #41194 强化核心测试 fixture 启动断言 @copyberry
- #41195 在 `ToolRouter` 中完成特定于模型的工具计划 @copyberry
- #41196 改进沙箱、MCP 错误和缓存的授权结果 @copyberry
- #41199 使可选 MCP 启动宽限期可配置 @copyberry
- #41202 允许扩展处理 MCP 工具结果 @copyberry
- #41204 将执行器主目录传递到沙箱上下文中 @copyberry
- #41205 追踪执行器 MCP 发现遥测数据 @copyberry
- #41206 使 Ultra 推理回退模型具备模型感知能力 @copyberry
- #41207 将执行器操作系统传递到回合环境中 @copyberry
- #41208 在目录请求中遵循每个仓库的插件配置 @copyberry
- #41209 使拒绝读取匹配逻辑与执行器路径语义保持一致 @copyberry
## New Features
- Added a configurable grace period for discovering tools from optional MCP servers. (#41199)
- Extensions can now inspect or replace MCP tool results before they reach the model. (#41202)
- Plugin catalogs now combine per-repository configuration and report invalid project marketplaces without hiding valid plugins. (#41208)
## Bug Fixes
- Preserved restored permission profiles across TUI turns and prevented `/cd` from weakening sandbox restrictions. (#41192)
- Kept tool availability and reasoning effort correct when switching models or falling back to another model. (#41195, #41206)
- Improved remote sandbox enforcement using the executor’s actual home directory, operating system, and path conventions. (#41196, #41204, #41207, #41209)
- Preserved structured MCP tool and resource errors in app-server responses. (#41196)
- Counted nested subagent token usage toward root goal budgets. (#41183)
- Prevented stale Guardian classifications from authorizing actions after permission state changes. (#41196)
## Chores
- Added telemetry for escalated stdin reviews and remote executor MCP discovery. (#41189, #41205)
- Stabilized Guardian WebSocket and core fixture tests under slow or highly concurrent CI. (#41191, #41194)
## Changelog
Full Changelog: https://github.com/openai/codex/compare/rust-v0.150.0...rust-v0.151.0
- #41183 Account subagent token usage toward root goals @copyberry
- #41189 Instrument stdin review size checks @copyberry
- #41191 Stabilize Guardian WebSocket tests @copyberry
- #41192 Preserve restored permission profiles in TUI sessions @copyberry
- #41193 Report affected capabilities from remote plugin syncs @copyberry
- #41194 Harden core test fixture startup assertions @copyberry
- #41195 Finalize model-specific tool plans in `ToolRouter` @copyberry
- #41196 Improve sandboxing, MCP errors, and cached approvals @copyberry
- #41199 Make the optional MCP startup grace configurable @copyberry
- #41202 Let extensions process MCP tool results @copyberry
- #41204 Propagate executor home directories into sandbox contexts @copyberry
- #41205 Track executor MCP discovery telemetry @copyberry
- #41206 Make Ultra reasoning fallback model-aware @copyberry
- #41207 Propagate executor OS into turn environments @copyberry
- #41208 Honor per-repository plugin configuration in catalog requests @copyberry
- #41209 Align deny-read matching with executor path semantics @copyberry