推荐插件 · 第 116 页

fork provider 会创建一个进程内子级，并以父级已完成的对话轮次作为初始内容。

The fork provider creates an in-process child seeded with the parent's completed conversation turns.

面向 evolution 管理的技能树的原生 ctx.skills provider——只读；变更通过 tool-skill-manage 进行。

Native ctx.skills provider for the evolution-managed skill tree — read-only; mutations go through tool-skill-manage.

面向插件作者和维护者的作用域注册库，用于构建按 agent 或按组隔离贡献的注册表或事件接口。

The scoped-registration library for plugin authors and maintainers building registries or event surfaces that isolate contributions per agent or per group.

通过单例 ctx.tokenMeter 服务进行可感知重放的 token 计量。

Replay-aware token measurement through the singleton ctx.tokenMeter service.

面向模型的 todo_write 工具：代理的整个任务列表，每次调用时整体替换。

The model-facing todo_write tool: the agent's whole task list, replaced wholesale on each call.

面向模型的 todo_write 工具基于 DeepSeek Harness 会话日志：整表替换、按会话归属以及 todos 投影，供选择、配置或调试该工具的用户与维护者参考。

The model-facing todo_write tool over the DeepSeek Harness session log: whole-list replacement, per-session ownership, and the todos projection, for users and maintainers choosing, configuring, or debugging the tool.

agent preset 携带的一行，用于说明模型看到其工具的哪种形式：native（每个 schema）、code（仅 run_code 加生成的 TypeScript SDK），或两者兼有。

The row an agent preset carries to say which form of its tools the model sees: native (every schema), code (only run_code plus a generated TypeScript SDK), or both.

基础压缩后端：BasicCompactionEngine 实现 @deepseek-ai/dsh-compaction 服务定义，具备可复用的 ctx.tokenMeter 压力、token 预算保留，以及通过直接一次性 ctx.llm.stream() 进行摘要……

The basic compaction backend: a BasicCompactionEngine implementing the @deepseek-ai/dsh-compaction Service Definition with reusable ctx.tokenMeter pressure, token-budget retention, and summarization as a direct one-shot ctx.llm.stream()…

面向用户与维护者的持久会话存储接缝，用于选择持久化后端、恢复会话，或基于共享服务契约构建后端。

The durable session-storage seam for users and maintainers choosing a persistence backend, resuming sessions, or building a backend against the shared service contract.

面向用户与维护者的持久会话存储接缝，用于选择持久化后端、恢复会话，或基于共享服务契约构建后端。

The durable session-storage seam for users and maintainers choosing a persistence backend, resuming sessions, or building a backend against the shared service contract.

DeepSeek Harness 的实时语音能力接缝。

A realtime voice capability seam for DeepSeek Harness.

用于 Host 和浏览器包中的循环双端队列，支持均摊常数时间的队列操作、被移除条目的即时释放以及有界空闲存储。

Circular deque for Host and browser packages that need amortized constant-time queue operations, immediate release of removed entries, and bounded vacant storage.

Web GUI 宿主的工作区目录选择器是一个能力接缝。

The web GUI host's workspace-directory picker is a capability seam.

Web GUI 宿主的 workspace 目录选择接缝：原生后端与浏览后端所实现的服务契约、能力词汇及错误码。

Workspace-directory picking seam for the web GUI host: the service contract, capability vocabulary, and error codes the native and browse backends implement.

用于投影状态的不可变仅追加列表，具有有界追加复制、按插入顺序迭代以及 Zod 检查点校验。

Immutable append-only lists for projection state, with bounded append copying, insertion-order iteration, and Zod checkpoint validation.

可选的全局命名工具 send_message、interrupt_agent 和 list_agents 是 ctx.subagents 之上的轻量适配器。

The optional, globally named send_message, interrupt_agent, and list_agents tools are thin adapters over ctx.subagents.

用于部署的 stdio JSON-RPC 服务插件，使进程外 SDK 客户端能够在 DeepSeek Harness 运行时中打开会话并驱动 agent。

The stdio JSON-RPC serving plugin for deployments that let out-of-process SDK clients open sessions and drive agents in a DeepSeek Harness runtime.

面向模型的 pwsh(command) 由单个 owner 作用域的 ctx.terminals shell 支撑。

Model-facing pwsh(command) backed by one owner-scoped ctx.terminals shell.

面向用户和维护者的提供商中立模型调用服务，支持流式请求、注册提供商适配器或解析模型元数据。

The provider-neutral model-call service for users and maintainers streaming requests, registering provider adapters, or resolving model metadata.

面向用户和维护者的提供商中立模型调用服务，支持流式请求、注册提供商适配器或解析模型元数据。

The provider-neutral model-call service for users and maintainers streaming requests, registering provider adapters, or resolving model metadata.

面向用户和维护者的提供商中立模型调用服务，支持流式请求、注册提供商适配器或解析模型元数据。

The provider-neutral model-call service for users and maintainers streaming requests, registering provider adapters, or resolving model metadata.

后台审查编排：当对话需要审查时（依据节奏与完成通道）进行锁定，构建审查请求，并在对话结束时执行经校验的计划所要求的写入操作。

Background review orchestration: it latches when a conversation deserves a review (the cadence and completion channels), builds the review request, and executes the writes the validated plan calls for at conversation end.

面向配置标题提供方或调试辅助 LLM 请求的用户和维护者的共享模型驱动标题生成策略。

Shared model-backed title generation policy for users and maintainers configuring title providers or debugging auxiliary LLM requests.

模型挂载动态包的宿主侧部分：定义注册表、用于宿主侧的 node:vm 沙箱与 fiber 生命周期、invoke 处理程序表，以及浏览器页面执行的 run 往返流程。

The host half of model-mounted dynamic packages: the definition registry, the node:vm sandbox and fiber lifecycle for host halves, the invoke handler table, and the run round trip a browser page carries out.