dominds 1.26.2 → 1.26.3

package/dist/tools/prompts/control/en/scenarios.md

@@ -57,11 +57,11 @@ delete_reminder({
 - If the information should synchronize the whole team's current effective state, put it in Taskdoc `progress` instead
 - If the note is durable knowledge rather than an active current-work cue, move it to `personal_memory` instead
 
-## Scenario 2: Side Dialog is complete, and the assignment header requires replyTellask
+## Scenario 2: Side Dialog is complete, and the task header requires replyTellask
 
 ### Scenario Description
 
-The current Side Dialog is finished, and the assignment header explicitly says "when complete, call `replyTellask`".
+The current Side Dialog is finished, and the task header explicitly says "when complete, call `replyTellask`".
 
 ### Example
 
@@ -74,7 +74,7 @@ replyTellask({
 
 ### Key Points
 
-- Do not replace this with a plain final message; direct-reply fallback is only a temporary runtime transition safeguard, not a valid planned completion method
+- Do not replace this with a plain final message; formal delivery must use the reply tool named by Dominds, or the other dialog may not receive a formal reply
 - Put the final deliverable body directly in `replyContent`
 - If the header says `replyTellaskSessionless`, use the same shape with that exact function name
 
@@ -97,13 +97,13 @@ tellaskBack({
 
 - This example uses `tellaskBack` because tellasker input is specifically required
 - If team SOP / role ownership already identifies the responsible executor, directly use `tellask` / `tellaskSessionless` for that owner instead of mapping every unfinished state to `tellaskBack`
-- Do not use `replyTellask*` for intermediate clarifications
+- Do not use reply tools for intermediate clarifications
 
 ## Scenario 4: Tellasker answered the ask-back, so use replyTellaskBack to close
 
 ### Scenario Description
 
-You previously sent a `tellaskBack`, the tellasker has now replied, and runtime exposes `replyTellaskBack`.
+You previously sent a `tellaskBack`, the tellasker has now replied, and Dominds shows `replyTellaskBack`.
 
 ### Example
 

package/dist/tools/prompts/control/en/tools.md

@@ -23,7 +23,7 @@
 
 The **tool descriptions themselves** for these functions intentionally stay minimal and spec-like. This section carries the smallest practical lookup for when they appear and how to choose among them.
 
-| Function                  | Minimal parameter contract   | When runtime exposes it                                                                                     | Effect                                                         |
+| Function                  | Minimal parameter contract   | When Dominds shows it                                                                                       | Effect                                                         |
 | ------------------------- | ---------------------------- | ----------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------- |
 | `replyTellask`            | `{ replyContent: string }`   | Current Side Dialog comes from a sessioned `tellask` and is ready for final delivery                        | Delivers the final result for the current tellask session      |
 | `replyTellaskSessionless` | `{ replyContent: string }`   | Current Side Dialog comes from a one-shot `tellaskSessionless` and is ready for final delivery              | Delivers the final result for the current one-shot tellask     |
@@ -32,11 +32,11 @@ The **tool descriptions themselves** for these functions intentionally stay mini
 
 ### Minimal Usage Rules
 
-- Focus on doing the current task correctly first; only move into `reply*` closure when final tellasker delivery is actually ready
-- Call whichever `reply*` runtime currently exposes; do not switch to another reply variant by yourself
-- If the assignment header explicitly names a reply function, follow that exact name
+- Focus on doing the current task correctly first; send the final reply only when the final content is ready and Dominds names a reply tool
+- Call whichever reply tool Dominds currently shows; do not switch to another reply tool by yourself
+- If the task header explicitly names a reply tool, follow that exact name
 - Put only the final deliverable body in `replyContent`; do not wrap it in meta-explanations like "I am now calling replyTellask"
-- If you produce final deliverable content instead of the reply tool, runtime may inject a temporary `role=user` reminder telling you to use the correct reply function; do not rely on direct-reply fallback, which is only a temporary runtime transition safeguard, not the formal reply mechanism
+- If you write final deliverable content but do not send it through the reply tool named by Dominds, Dominds may temporarily remind you to use that tool; do not treat plain text as the formal reply path, because the other dialog may not receive a formal reply that way
 
 ### 1. add_reminder
 
@@ -45,8 +45,8 @@ Add reminder.
 Use when:
 
 - Adding a new temporary current-work item
-- Before `clear_mind`, the Main Dialog first records undocumented discussion details the next course needs to know into Taskdoc, then creates continuation-package notes; a Side Dialog directly maintains sufficiently detailed continuation-package reminders. When the current course is already under caution/critical remediation, Side Dialog reminder length has no technical limit and rough bridge notes are acceptable
-- Record only manually maintained current-work / continuation details; do not put runtime-maintained environment state such as background process status, in-flight background asks, or session attachment state into manual reminders
+- Before `clear_mind`, the Main Dialog first records undocumented discussion details the next course needs to know into Taskdoc, then creates continuation-package notes; a Side Dialog directly maintains sufficiently detailed continuation-package reminders. When Dominds has already warned that context is tight or critical, Side Dialog reminders have no fixed length limit and rough bridge notes are acceptable
+- Record only manually maintained current-work / continuation details; do not put environment state automatically maintained by Dominds, such as background process status, in-flight background asks, or session attachment state, into manual reminders
 
 **Parameters:**
 
@@ -92,7 +92,7 @@ Use when:
 - Compressing / merging existing reminders
 - Rewriting pre-clear resume info into continuation-package notes
 - Removing details that have already been promoted into Taskdoc
-- Removing or rewriting system environment state that was mistakenly captured in a manual reminder; runtime-managed reminders, panels, and tool outputs are the single source of truth for that state
+- Removing or rewriting system environment state that was mistakenly captured in a manual reminder; Dominds-managed reminders, panels, and tool outputs are the authoritative place for that state
 
 **Parameters:**
 
@@ -360,7 +360,7 @@ message: <error message>
 
 - Normal reminders should stay concise, fresh, and directly actionable; often 1-3 items total
 - For a continuation package, use structured notes by default: next step, key pointers, run/verify, easy-to-lose volatile details
-- If the current course is already under caution/critical remediation: the Main Dialog first records undocumented discussion details the next course needs to know into the appropriate Taskdoc sections, then keeps necessary continuation-package reminders; a Side Dialog must not maintain Taskdoc or draft Taskdoc update proposals, and should directly maintain sufficiently detailed continuation-package reminders with no technical length limit. Rough multi-reminder bridge notes are acceptable and should be reconciled as the first step only after the system actually starts the new course
+- If Dominds has already warned that context is tight or critical: the Main Dialog first records undocumented discussion details the next course needs to know into the appropriate Taskdoc sections, then keeps necessary continuation-package reminders; a Side Dialog must not maintain Taskdoc or draft Taskdoc update proposals, and should directly maintain sufficiently detailed continuation-package reminders with no fixed length limit. Rough multi-reminder bridge notes are acceptable and should be reconciled as the first step only after Dominds actually starts the new course
 - Keep only details still not covered by Taskdoc; do not repeat team-shared status. If the team needs “where we are now / which decisions are in effect / what is next / which blockers still hold”, write it back to Taskdoc `progress`
-- Do not manually record Dominds runtime-maintained environment state, such as whether background processes are still running, in-flight background asks/collaboration, or browser/session attachment state. Manual copies go stale easily and conflict with the runtime-maintained single source of truth, creating cognitive noise
+- Do not manually record environment state automatically maintained by Dominds, such as whether background processes are still running, in-flight background asks/collaboration, or browser/session attachment state. Manual copies go stale easily and conflict with the Dominds-managed status, creating cognitive noise
 - Do not paste long logs, large tool outputs, or raw material into reminders

package/dist/tools/prompts/control/zh/index.md

@@ -29,7 +29,7 @@ control 是 Dominds 的**对话控制工具集**，用于管理对话状态、
 - **提醒管理**：提醒分 `dialog` / `task` / `agent` 三个 scope；默认是 `task`，用于同一差遣牒任务下的手头工作；`dialog` 只用于真正对话局部的提醒；`agent` 只用于紧急、短期、全局刺眼提醒
 - **差遣牒操作**：追加、替换或删除任务契约章节（goals/constraints/progress）；其中 `progress` 是全队共享、准实时、可扫读的任务公告牌
 - **上下文维护**：在不丢关键恢复线索的前提下降低认知负载
-- **回复路由**：在支线/回问语境下，区分 `tellaskBack`、`replyTellask*` 与普通文本的职责边界
+- **回复路由**：在支线/回问语境下，区分回问诉请者、发送最终回贴与普通文本的职责边界
 
 ## 快速导航
 
@@ -56,9 +56,9 @@ control 是 Dominds 的**对话控制工具集**，用于管理对话状态、
 - 标记待处理事项
 - 追踪当前下一步/阻塞
 - 记录阻塞问题
-- 在 `clear_mind` 前承载接续包；若当前程已被系统置于吃紧/告急处置态，可先带多条粗略提醒项过桥
+- 在 `clear_mind` 前承载接续包；若 Dominds 已提醒当前程吃紧/告急，可先带多条粗略提醒项过桥
 
-提醒不用于手工复制 Dominds 系统环境会自动维护的状态，例如后台进程状态、后台进行中诉请/协作、浏览器/会话附件状态等。这些状态以 runtime 自动维护的提醒/面板/工具输出为 single source of truth；手工记录很容易过时并造成认知干扰。
+提醒不用于手工复制 Dominds 会自动维护的系统状态，例如后台进程状态、后台进行中诉请/协作、浏览器/会话附件状态等。这些状态以 Dominds 自动维护的提醒、面板和工具输出为准；手工记录很容易过时并造成认知干扰。
 
 scope 规则：
 

package/dist/tools/prompts/control/zh/principles.md

@@ -42,7 +42,7 @@
 - `agent` 提醒会在所有由你主理的后续对话里继续可见
 - 可随时添加、修改、删除
 - 默认应保持少量、可扫读、可直接指导下一步
-- 准备 `clear_mind` 时，主线对话先把应由下一程知会的未落文档讨论细节补进差遣牒，再压缩成结构化接续包提醒项；支线对话直接维护足够详尽的接续包提醒项。若当前程已被系统置于吃紧/告急处置态，支线提醒项长度没有技术限制，也可先保留多条粗略提醒项过桥
+- 准备 `clear_mind` 时，主线对话先把应由下一程知会的未落文档讨论细节补进差遣牒，再压缩成结构化接续包提醒项；支线对话直接维护足够详尽的接续包提醒项。若 Dominds 已提醒上下文吃紧/告急，支线提醒项没有固定长度限制，也可先保留多条粗略提醒项过桥
 
 **与 memory 的区别：**
 | 特性 | dialog reminder | task reminder | agent reminder | personal memory |
@@ -103,17 +103,17 @@
 - 不要把智能体队友当成真人同事：祂不会因为你又发一条诉请而被打扰。同一个队友 + 同一个 `sessionSlug` = 接着同一件事说，并更新那件正在做的事；`tellaskSessionless` 或不同 `sessionSlug` = 另一件独立任务
 - 只有当必须向诉请者补需求、澄清目标、裁决取舍、确认验收口径、提供缺失输入，或现有规程无法明确判责时：调用 `tellaskBack({ tellaskContent })`
 - 需要人类亲自登录 / GUI / 验证码 / 高风险授权时：调用 `askHuman({ tellaskContent })`
-- 当前支线已经完成，且当前 assignment 明确要求 `replyTellask`：调用 `replyTellask({ replyContent })`
-- 当前支线已经完成，且当前 assignment 明确要求 `replyTellaskSessionless`：调用 `replyTellaskSessionless({ replyContent })`
-- 当前是在回复一条诉请者发来的 `tellaskBack` 续诉请，且 runtime 暴露了 `replyTellaskBack`：调用 `replyTellaskBack({ replyContent })`
-- 普通文本不是跨对话完成交付通道；如果你直接产出可作为最终回贴的内容而没调 reply 工具，runtime 可能临时插入一条 `role=user` 的提醒要求你改用正确的 reply 函数。不要依赖 direct-reply fallback；它只是运行时临时过渡兜底，不是正式回复机制
+- 当前支线已经完成，且当前任务开头明确要求 `replyTellask`：调用 `replyTellask({ replyContent })`
+- 当前支线已经完成，且当前任务开头明确要求 `replyTellaskSessionless`：调用 `replyTellaskSessionless({ replyContent })`
+- 当前是在回复一条诉请者发来的 `tellaskBack` 续诉请，且 Dominds 显示了 `replyTellaskBack`：调用 `replyTellaskBack({ replyContent })`
+- 普通文本不是跨对话完成交付通道；如果你直接写出了可回给诉请者的最终内容但没按 Dominds 点名的回复工具发送，Dominds 可能临时提醒你改用那个工具。不要把普通文本当成正式回贴路径；对方可能收不到正式回贴。
 
 ### 心智负担最小化规则
 
-- 先专注把当前任务做对；`reply*` 只在最终对诉请者交付时用于收口
-- 不要靠记忆硬选 reply 变体；以当前 assignment 头部和 runtime 当前暴露的函数名为准
-- `reply*` 函数自身说明文案故意保持极简，只承载最小规格；情景判断看本手册的 principles / scenarios
-- 若 runtime 只暴露一个 `reply*`，那就是当前应调用的唯一完成路径
+- 先专注把当前任务做对；只有最终内容准备好、且 Dominds 点名了回复工具时，才发送最终回贴
+- 不要靠记忆硬选回复工具；以当前任务开头和 Dominds 当前点名/显示的回复工具为准
+- 回复工具自身说明文案故意保持极简，只承载最小规格；情景判断看本手册的 principles / scenarios
+- 若 Dominds 只点名/显示一个回复工具，那就是当前应调用的唯一完成路径
 - `tellaskBack` 只在“无法按现有规程明确判责，或必须回问诉请者”时成立，不是所有阻塞的默认第一动作
 
 ## 最佳实践
@@ -122,7 +122,7 @@
 
 - **当前手头工作**：当前下一步、阻塞、关键定位信息
 - **易丢细节**：临时路径、id、命令、样例输入
-- **换程接续**：在 `clear_mind` 前整理接续包；若当前程已被系统置于吃紧/告急处置态，可先保留多条粗略提醒项
+- **换程接续**：在 `clear_mind` 前整理接续包；若 Dominds 已提醒当前程吃紧/告急，可先保留多条粗略提醒项
 - **任务延续提示**：如果在当前差遣牒任务内都应该继续看到它，用 `task`
 - **全局刺眼提示**：如果它应该在所有由你主理的后续对话里继续看到，而且是短期、紧急、全局刺眼提醒，用 `agent`
 
@@ -136,10 +136,10 @@
 
 - 保持简洁：默认提醒项常见 1–3 条，优先 `update_reminder` 压缩/合并
 - 区分载体：需要向全队同步的当前有效状态、关键决策、下一步与仍成立阻塞，写入 `progress` 这一准实时任务公告牌；提醒项只留个人/当前对话恢复所需细节
-- 避免重复系统状态：后台进程状态、后台进行中诉请/协作、浏览器/会话附件状态等 Dominds 系统环境会自动维护的信息，不要写进手工提醒项；这些状态以 runtime 自动维护的提醒/面板/工具输出为 single source of truth，手工副本容易过时并造成认知干扰
+- 避免重复系统状态：后台进程状态、后台进行中诉请/协作、浏览器/会话附件状态等 Dominds 会自动维护的信息，不要写进手工提醒项；这些状态以 Dominds 自动维护的提醒、面板和工具输出为准，手工副本容易过时并造成认知干扰
 - 面向全队：`progress` 应保持可扫读、以“当前仍有效”为准，不要退化成个人日志、流水账、临时便签或历史残影堆积；少量新增用 `mind_more`，需要清旧/重排/压缩时先调用 `recall_taskdoc`，再用 `change_mind` 并携带返回的 `content_hash` 作为 `previous_content_hash`
 - 按需整理：`mind_more` 不是默认记账动作。若同一主题已有多条阶段性记录，优先用 `change_mind` 合并成当前摘要；把细节放进 rtws 正式文档，并在差遣牒里保留文档定位 pointer。若会冲掉原有内容，必须有直接人类确认，或已按人类认可 SOP/验收标准充分考虑原有内容后再改写
-- 换程前收束：主线对话先把尚未落实到文档、且下一程需要知会的讨论细节写入差遣牒合适章节，再整理结构化接续包提醒项；支线对话不要维护差遣牒，也不要整理差遣牒更新提案，直接维护足够详尽的接续包提醒项。若系统已把当前程切到吃紧/告急处置态，则先保留多条粗略提醒项过桥也可以；当前程只做落文档/保信息 + `clear_mind`（支线只做保信息 + `clear_mind`），系统真正开启新一程后第一步再收敛
+- 换程前收束：主线对话先把尚未落实到文档、且下一程需要知会的讨论细节写入差遣牒合适章节，再整理结构化接续包提醒项；支线对话不要维护差遣牒，也不要整理差遣牒更新提案，直接维护足够详尽的接续包提醒项。若 Dominds 已提醒上下文吃紧/告急，则先保留多条粗略提醒项过桥也可以；当前程只做落文档/保信息 + `clear_mind`（支线只做保信息 + `clear_mind`），Dominds 真正开启新一程后第一步再收敛
 - 拒绝原料堆积：不要把长日志/大段 tool output 直接塞进提醒项
 - 文档分层：差遣牒写“现在应如何同步/推进”，正式文档写“为什么、怎么做、详细证据和完整过程”。差遣牒需要引用正式文档时，写稳定路径/章节名/相关命令，而不是复制整段内容
 

package/dist/tools/prompts/control/zh/scenarios.md

@@ -57,11 +57,11 @@ delete_reminder({
 - 如果信息需要向全队同步当前有效状态、关键决策、下一步或仍成立阻塞，应写入 Taskdoc `progress`
 - 如果内容本质上是长期知识而不是当前手头工作提示，应改存到 `personal_memory`
 
-## 场景 2：支线已完成，按 assignment 头部要求调用 replyTellask
+## 场景 2：支线已完成，按任务开头要求调用 replyTellask
 
 ### 场景描述
 
-当前支线处理完毕，assignment 头部明确写着“完成任务时必须调用 `replyTellask`”。
+当前支线处理完毕，任务开头明确写着“完成任务时必须调用 `replyTellask`”。
 
 ### 示例
 
@@ -73,9 +73,9 @@ replyTellask({
 
 ### 关键点
 
-- 不要再发一条普通最终消息代替；direct-reply fallback 只是运行时临时过渡兜底，不是可规划使用的收口方式
+- 不要再发一条普通最终消息代替；正式回贴必须走 Dominds 点名的回复工具，否则对方可能收不到正式回贴
 - `replyContent` 直接放最终交付正文
-- 若 assignment 头部写的是 `replyTellaskSessionless`，则同结构替换函数名
+- 若任务开头写的是 `replyTellaskSessionless`，则同结构替换函数名
 
 ## 场景 3：当前未完成，需要回问诉请者
 
@@ -90,20 +90,20 @@ tellaskBack({
   tellaskContent: '还缺少生产环境端口与部署入口信息。请补充这两项后我再继续给出最终方案。',
 });
 
-// 等诉请者补充后，runtime 会在当前对话里继续推进
+// 等诉请者补充后，Dominds 会在当前对话里继续推进
 ```
 
 ### 关键点
 
 - 这里只是“必须向诉请者补输入”的例子，所以用 `tellaskBack`
 - 若团队规程 / SOP / 职责卡已经能明确执行负责人，应直接 `tellask` / `tellaskSessionless` 对应负责人，而不是机械因为“未完成态”就用 `tellaskBack`
-- `tellaskBack` 只负责把问题问回去，不负责最终交付
+- `tellaskBack` 只负责把问题问回去，不负责最终交付；中间澄清不要调用回复工具
 
 ## 场景 4：收到 ask-back 续诉请后，用 replyTellaskBack 收口
 
 ### 场景描述
 
-你之前发过 `tellaskBack`，诉请者现在补回了所需信息，runtime 暴露了 `replyTellaskBack`。
+你之前发过 `tellaskBack`，诉请者现在补回了所需信息，Dominds 显示了 `replyTellaskBack`。
 
 ### 示例
 

package/dist/tools/prompts/control/zh/tools.md

@@ -23,7 +23,7 @@
 
 这些函数的**工具自身说明**刻意保持最小规格风格；这里负责补充“何时会出现、出现后怎么选”的最小速查。
 
-| 函数                      | 最小参数契约                 | runtime 何时暴露                                                     | 调用后语义                          |
+| 函数                      | 最小参数契约                 | Dominds 何时显示                                                     | 调用后语义                          |
 | ------------------------- | ---------------------------- | -------------------------------------------------------------------- | ----------------------------------- |
 | `replyTellask`            | `{ replyContent: string }`   | 当前支线承接的是 sessioned `tellask`，且已进入可交付完成态           | 把最终结果回复给当前 `tellask` 会话 |
 | `replyTellaskSessionless` | `{ replyContent: string }`   | 当前支线承接的是 one-shot `tellaskSessionless`，且已进入可交付完成态 | 把最终结果回复给当前一次性诉请      |
@@ -32,11 +32,11 @@
 
 ### 最小使用规则
 
-- 先专注把当前任务做对；只有真到最终对诉请者交付时，才进入 `reply*` 收口
-- 看见哪一个 `reply*` 被 runtime 暴露，就调用哪一个；不要自行改选别的 reply 变体
-- assignment 头部若已点名 reply 函数名，以那个名字为准
+- 先专注把当前任务做对；只有最终内容准备好、且 Dominds 点名了回复工具时，才发送最终回贴
+- 看见 Dominds 当前显示哪一个回复工具，就调用哪一个；不要自行改选别的回复工具
+- 任务开头若已点名回复工具名，以那个名字为准
 - `replyContent` 只放最终交付正文，不要再包一层“我现在调用了 replyXXX”
-- 如果你直接产出最终回贴内容而没调 reply 工具，runtime 可能插入一条 `role=user` 的 runtime reminder 纠正你；不要依赖 direct-reply fallback，它只是运行时临时过渡兜底，不是正式回复机制
+- 如果你直接写出了最终回贴内容但没按 Dominds 点名的回复工具发送，Dominds 可能临时提醒你改用那个工具；不要把普通文本当成正式回贴路径，对方可能收不到正式回贴
 
 ### 1. add_reminder
 
@@ -45,8 +45,8 @@
 **适用：**
 
 - 添加新的临时手头工作条目
-- 若准备 `clear_mind`，主线对话先把应由下一程知会的未落文档讨论细节补进差遣牒，再新建接续包提醒项；支线对话直接维护足够详尽的接续包提醒项。若当前程已被系统置于吃紧/告急处置态，支线提醒项长度没有技术限制，先记粗略过桥信息也可以
-- 只记录需要你手工维护的手头工作/接续细节；不要在手工提醒项里记录后台进程状态、后台进行中诉请、会话附件状态等 Dominds 系统环境会自动维护的信息
+- 若准备 `clear_mind`，主线对话先把应由下一程知会的未落文档讨论细节补进差遣牒，再新建接续包提醒项；支线对话直接维护足够详尽的接续包提醒项。若 Dominds 已提醒上下文吃紧/告急，支线提醒项没有固定长度限制，先记粗略过桥信息也可以
+- 只记录需要你手工维护的手头工作/接续细节；不要在手工提醒项里记录后台进程状态、后台进行中诉请、会话附件状态等 Dominds 会自动维护的信息
 
 **参数：**
 
@@ -90,9 +90,9 @@ deleted_at: <删除时间戳>
 **适用：**
 
 - 压缩/合并现有提醒项
-- 把换程前需要保留的信息整理成接续包；若当前程已被系统置于吃紧/告急处置态，也可先保留多条粗略提醒项
+- 把换程前需要保留的信息整理成接续包；若 Dominds 已提醒当前程吃紧/告急，也可先保留多条粗略提醒项
 - 删除已写入差遣牒、无需在提醒项重复保留的内容
-- 删除/改写手工提醒项里误记录的系统环境状态；这类状态以 runtime 自动维护的提醒/面板/工具输出为 single source of truth
+- 删除/改写手工提醒项里误记录的系统环境状态；这类状态以 Dominds 自动维护的提醒、面板和工具输出为准
 
 **参数：**
 
@@ -353,7 +353,7 @@ message: <错误消息>
 
 - 默认提醒项应保持短、新、能直接指导下一步，常见 1–3 条
 - 若用于接续包，默认优先结构化内容：下一步行动、关键定位、运行/验证、容易丢的临时细节
-- 若已吃紧/告急：主线对话先把尚未落实到文档、且下一程需要知会的讨论细节写入差遣牒合适章节，再保留必要的接续包提醒项；支线对话不要维护差遣牒，也不要整理差遣牒更新提案，直接维护足够详尽的接续包提醒项，长度没有技术限制；多条粗略提醒项也可以，系统真正开启新一程后第一步再收敛整理
+- 若 Dominds 已提醒上下文吃紧/告急：主线对话先把尚未落实到文档、且下一程需要知会的讨论细节写入差遣牒合适章节，再保留必要的接续包提醒项；支线对话不要维护差遣牒，也不要整理差遣牒更新提案，直接维护足够详尽的接续包提醒项，没有固定长度限制；多条粗略提醒项也可以，Dominds 真正开启新一程后第一步再收敛整理
 - 接续包只保留差遣牒仍未覆盖的细节；不要重复团队共享状态。要向全队同步“现在到哪了 / 哪些决策已生效 / 下一步是什么 / 哪些阻塞仍成立”，请写回 Taskdoc `progress`
-- 不要手工记录 Dominds 系统环境会自动维护的状态，例如后台进程是否仍在运行、后台进行中的诉请/协作、浏览器/会话附件状态等；手工记录很容易过时，并与 runtime 自动维护的 single source of truth 冲突，造成认知干扰
+- 不要手工记录 Dominds 会自动维护的系统状态，例如后台进程是否仍在运行、后台进行中的诉请/协作、浏览器/会话附件状态等；手工记录很容易过时，并与 Dominds 自动维护的状态冲突，造成认知干扰
 - 不要把长日志、大段 tool output、原始材料直接塞进提醒项

package/dist/tools/prompts/mcp_admin/en/principles.md

@@ -37,15 +37,15 @@ MCP (Model Context Protocol) is a protocol for connecting to external services.
 
 ### 2. MCP Lease
 
-MCP uses a lease mechanism to manage runtime resources such as HTTP connections and stdio processes. Lease ownership is about who holds a runtime instance; it does not define global MCP tool registration or visibility.
+MCP uses a lease mechanism to manage connection/process resources such as HTTP connections and stdio processes. Lease ownership is about who holds a connection/process; it does not define global MCP tool registration or visibility.
 
 **Lifecycle:**
 
-- **Acquire lease**: Establish or hold a runtime instance
-- **Hold lease**: Keep using that runtime instance
+- **Acquire lease**: Establish or hold a connection/process
+- **Hold lease**: Keep using that connection/process
 - **Release lease**: Stop/disconnect and release resources
 
-`mcp_restart` changes the target server from `enabled: false` back to `enabled: true`, then tries to start it. When it succeeds, it replaces the global MCP runtime and clears all dialog leases on the old runtime; you do not need to call `mcp_release` first. To force a server into the disabled state, use `mcp_disable`.
+`mcp_restart` changes the target server from `enabled: false` back to `enabled: true`, then tries to start it. When it succeeds, it replaces the global MCP service connection and clears all dialog leases on the old connection/process; you do not need to call `mcp_release` first. To force a server into the disabled state, use `mcp_disable`.
 
 ### 3. Environment Variables
 

package/dist/tools/prompts/mcp_admin/en/scenarios.md

@@ -17,7 +17,7 @@
 
 When an MCP service has issues, needs to refresh the connection, or is currently disabled with `enabled: false`, enable and restart the MCP service.
 
-`mcp_restart` writes `enabled: true` when the server is currently `enabled: false`, then tries to start it. After it succeeds, it clears every dialog lease on the old runtime; you do not need to call `mcp_release` first.
+`mcp_restart` writes `enabled: true` when the server is currently `enabled: false`, then tries to start it. After it succeeds, it clears every dialog lease on the old connection/process; you do not need to call `mcp_release` first.
 
 ### Example
 

package/dist/tools/prompts/mcp_admin/en/tools.md

@@ -21,7 +21,7 @@
 
 ### 1. mcp_restart
 
-Enable and rebuild an MCP service using the current `.minds/mcp.yaml` config. If the target server is currently `enabled: false`, this writes `enabled: true` before trying to start it. After a successful restart, Dominds replaces the global MCP runtime/tool registration and clears all dialog leases held on the old runtime. If restart fails, the old runtime/leases are kept so troubleshooting does not tear down a still-working connection.
+Enable and rebuild an MCP service using the current `.minds/mcp.yaml` config. If the target server is currently `enabled: false`, this writes `enabled: true` before trying to start it. After a successful restart, Dominds replaces the global MCP tool registration and clears all dialog leases held on the old connection/process. If restart fails, the old connection/process and leases are kept so troubleshooting does not tear down a still-working connection.
 
 **Parameters:**
 
@@ -39,7 +39,7 @@ ok: restarted <MCP service identifier>
 
 ### 2. mcp_release
 
-Release the current dialog's leased MCP runtime instance for a server. This stops/releases the underlying HTTP connection or stdio process, but does not define global tool registration/visibility.
+Release the current dialog's leased MCP connection/process for a server. This stops/releases the underlying HTTP connection or stdio process, but does not define global tool registration/visibility.
 
 **Parameters:**
 
@@ -64,7 +64,7 @@ ok: no active lease for <MCP service identifier> (or server is truely-stateless)
 
 ### 3. mcp_disable
 
-Disable an MCP service and write `enabled: false` for that server in `.minds/mcp.yaml`. This does not wait for a replacement service to become available: it unconditionally clears the loaded runtime/leases. The disabled server remains visible as a zero-tool MCP toolset, with its manual clearly marked disabled.
+Disable an MCP service and write `enabled: false` for that server in `.minds/mcp.yaml`. This does not wait for a replacement service to become available: it unconditionally clears the loaded MCP connection/leases. The disabled server remains visible as a zero-tool MCP toolset, with its manual clearly marked disabled.
 
 **Parameters:**
 

package/dist/tools/prompts/mcp_admin/zh/principles.md

@@ -37,15 +37,15 @@ MCP（Model Context Protocol）是连接外部服务的协议。
 
 ### 2. MCP 租约
 
-MCP 使用租约机制管理运行时资源，例如 HTTP 连接和 stdio 进程。租约只表示谁持有某个运行时实例，不决定 MCP 工具的全局注册或可见性。
+MCP 使用租约机制管理连接/进程资源，例如 HTTP 连接和 stdio 进程。租约只表示谁持有某个连接/进程，不决定 MCP 工具的全局注册或可见性。
 
 **生命周期：**
 
-- **获取租约**：建立或持有一个运行时实例
-- **持有租约**：继续使用该运行时实例
+- **获取租约**：建立或持有一个连接/进程
+- **持有租约**：继续使用该连接/进程
 - **释放租约**：停止/断开并释放资源
 
-`mcp_restart` 会把目标 server 的 `enabled: false` 改回 `enabled: true`，然后尝试启动；成功时会替换全局 MCP runtime 并清掉旧 runtime 的所有对话 lease；不需要先逐个 `mcp_release`。如果需要强制禁用某个 server，使用 `mcp_disable`。
+`mcp_restart` 会把目标 server 的 `enabled: false` 改回 `enabled: true`，然后尝试启动；成功时会替换全局 MCP 服务连接，并清掉旧连接/进程的所有对话 lease；不需要先逐个 `mcp_release`。如果需要强制禁用某个 server，使用 `mcp_disable`。
 
 ### 3. 环境变量
 

package/dist/tools/prompts/mcp_admin/zh/scenarios.md

@@ -17,7 +17,7 @@
 
 当 MCP 服务出现故障、需要刷新连接，或当前被 `enabled: false` 禁用时，启用并重启 MCP 服务。
 
-`mcp_restart` 会把 `enabled: false` 写回 `enabled: true` 后尝试启动；成功后会清理旧 runtime 的全部对话 lease，不需要先 `mcp_release`。
+`mcp_restart` 会把 `enabled: false` 写回 `enabled: true` 后尝试启动；成功后会清理旧连接/进程 的全部对话 lease，不需要先 `mcp_release`。
 
 ### 示例
 

package/dist/tools/prompts/mcp_admin/zh/tools.md

@@ -21,7 +21,7 @@
 
 ### 1. mcp_restart
 
-按当前 `.minds/mcp.yaml` 配置启用并重建 MCP 服务。如果目标 server 当前是 `enabled: false`，会先写回 `enabled: true` 再尝试启动。重启成功后会替换全局 MCP runtime/tool 注册，并清理旧 runtime 上所有对话持有的 lease；重启失败时保留旧 runtime/lease，避免排障过程中把仍可用的连接拆掉。
+按当前 `.minds/mcp.yaml` 配置启用并重建 MCP 服务。如果目标 server 当前是 `enabled: false`，会先写回 `enabled: true` 再尝试启动。重启成功后会替换全局 MCP 工具注册，并清理旧连接/进程上所有对话持有的 lease；重启失败时保留旧连接/进程和 lease，避免排障过程中把仍可用的连接拆掉。
 
 **参数：**
 
@@ -39,7 +39,7 @@ ok: restarted <MCP 服务标识符>
 
 ### 2. mcp_release
 
-释放当前对话为某个 server 持有的 MCP 运行时实例。它会停止/释放底层 HTTP 连接或 stdio 进程，但不决定工具的全局注册或可见性。
+释放当前对话为某个 server 持有的 MCP 连接/进程。它会停止/释放底层 HTTP 连接或 stdio 进程，但不决定工具的全局注册或可见性。
 
 **参数：**
 
@@ -64,7 +64,7 @@ ok: no active lease for <MCP 服务标识符> (or server is truely-stateless)
 
 ### 3. mcp_disable
 
-禁用 MCP 服务并将 `.minds/mcp.yaml` 中对应 server 写为 `enabled: false`。该操作不等待新服务可用：会无条件清理已加载 runtime/lease。禁用后的 server 仍作为 0 工具 MCP toolset 可见，并在手册中明确标记为 disabled。
+禁用 MCP 服务并将 `.minds/mcp.yaml` 中对应 server 写为 `enabled: false`。该操作不等待新服务可用：会无条件清理已加载 MCP 连接/lease。禁用后的 server 仍作为 0 工具 MCP toolset 可见，并在手册中明确标记为 disabled。
 
 **参数：**
 

package/dist/tools/team_mgmt-mcp-manual.js

@@ -76,14 +76,14 @@ function renderMcpManualPresentBlock(language, serverId, manual) {
     if (manual.contentFile && manual.contentFile.trim() !== '') {
         out += fmtList([
             language === 'zh'
-                ? `运行时手册文件前缀（\`contentFile\`）：\`${manual.contentFile}\``
-                : `Runtime manual file prefix (\`contentFile\`): \`${manual.contentFile}\``,
+                ? `手册文件前缀（\`contentFile\`）：\`${manual.contentFile}\``
+                : `Manual file prefix (\`contentFile\`): \`${manual.contentFile}\``,
         ]);
         if ((!manual.content || manual.content.trim() === '') && manual.sections.length === 0) {
             out += fmtList([
                 language === 'zh'
-                    ? `该 toolset 的最终 \`man({ "toolsetId": "${serverId}" })\` 正文会在运行时从该前缀下的 topic 文件加载；这里不直接内嵌正文。`
-                    : `The final \`man({ "toolsetId": "${serverId}" })\` body is loaded at runtime from topic files under this prefix; it is not inlined here.`,
+                    ? `该 toolset 的最终 \`man({ "toolsetId": "${serverId}" })\` 正文会从该前缀下的 topic 文件加载；这里不直接内嵌正文。`
+                    : `The final \`man({ "toolsetId": "${serverId}" })\` body is loaded from topic files under this prefix; it is not inlined here.`,
             ]);
         }
     }
@@ -189,8 +189,8 @@ function renderRawMcpAutomaticBaselineSection(language, entry) {
             ? '未配置 `servers.<serverId>.manual`；以下为标准 Raw MCP 接入的自动基础说明。'
             : '`servers.<serverId>.manual` is not configured; below is the automatic baseline for this standard Raw MCP integration.',
         language === 'zh'
-            ? `运行时状态：status=${statusText}，transport=${transportText}。`
-            : `Runtime status: status=${statusText}, transport=${transportText}.`,
+            ? `当前连接状态：status=${statusText}，transport=${transportText}。`
+            : `Current connection status: status=${statusText}, transport=${transportText}.`,
         language === 'zh'
             ? '说明：MCP 工具名称、description 与参数 schema 已由 server 通过 `tools/list` 自动提供，并随工具函数说明进入 LLM 上下文；手册应列出工具名清单以澄清“哪些工具属于该 toolset”，但不要复制参数 schema 或长篇工具契约。'
             : 'Note: MCP tool names, descriptions, and parameter schemas are already provided by the server through `tools/list` and enter LLM context as function-tool metadata; manuals should list tool names to clarify which tools belong to this toolset, but should not copy parameter schemas or long tool-contract prose.',
@@ -515,8 +515,8 @@ async function renderMcpManual(language) {
         return (fmtHeader('.minds/mcp.yaml') +
             fmtList([
                 '每个 MCP `serverId` 注册一个 toolset，toolset 名称 = `serverId`（不加 `mcp_` 前缀）。成员通过 `members.<id>.toolsets` 选择能用哪些 MCP toolset。',
-                '支持热重载：编辑 `.minds/mcp.yaml` 后通常无需重启 Dominds；必要时用 `mcp_restart`。`mcp_restart` 会把目标 server 的 `enabled: false` 写回 `enabled: true` 后尝试启动；成功后替换全局 runtime 并清掉旧 runtime 的全部对话 lease。',
-                '默认按“每个对话租用一个 MCP 运行时实例”运行（更安全）：某个对话首次使用该 server 时，可能为它启动/持有一个运行时实例（HTTP 连接或 stdio 进程），并添加 sticky reminder。确认当前对话不再需要该运行时实例后，用 `mcp_release` 释放。租约只表达运行时资源归属，不决定该 server 的全局工具注册/可见性；如确实是无状态服务器，可配置 `truely-stateless: true` 允许跨对话共享。若需停用某 server，用 `mcp_disable` 写入 `enabled: false`；禁用后仍会保留 0 工具 toolset 与 Raw MCP 基线/手册可见性。',
+                '支持热重载：编辑 `.minds/mcp.yaml` 后通常无需重启 Dominds；必要时用 `mcp_restart`。`mcp_restart` 会把目标 server 的 `enabled: false` 写回 `enabled: true` 后尝试启动；成功后替换全局 MCP 工具注册，并清掉旧连接上的全部对话 lease。',
+                '默认按“每个对话租用一个 MCP 连接/进程”运行（更安全）：某个对话首次使用该 server 时，可能为它启动/持有一个连接或进程（HTTP 连接或 stdio 进程），并添加 sticky reminder。确认当前对话不再需要该连接/进程后，用 `mcp_release` 释放。租约只表达连接/进程资源归属，不决定该 server 的全局工具注册/可见性；如确实是无状态服务器，可配置 `truely-stateless: true` 允许跨对话共享。若需停用某 server，用 `mcp_disable` 写入 `enabled: false`；禁用后仍会保留 0 工具 toolset 与 Raw MCP 基线/手册可见性。',
                 'stdio 配置格式：`command` 必须是字符串（可执行命令），参数放在 `args`（string[]，可省略，默认空数组）。`cwd` 可选（字符串）：用于固定相对路径解析目录。',
                 'HTTP headers 支持三种值：字面量字符串、`{ env: NAME }`、`{ prefix: "Bearer ", env: NAME }`；认证 token 建议从环境变量读取，不要写死在 YAML。',
                 '用 `tools.whitelist/blacklist` 控制暴露的工具。顶层 `transform` 是 tools/prompts/resources/resource skills 的默认 ID 转换；内层 `tools.transform`、`prompts.transform`、`resources.transform`、`resources.skills.transform` 会覆盖外层转换，不会叠加；显式 `transform: []` 表示该层不转换。',
@@ -524,7 +524,7 @@ async function renderMcpManual(language) {
                 '高频坑（stdio 路径）：若未设置 `cwd`，相对路径按 Dominds 进程工作目录（通常 rtws 根目录）解析；建议显式配置 `cwd` 或直接使用绝对路径。`cwd` 必须存在且是目录。',
                 '可选手册字段：`servers.<serverId>.manual` 是增强说明，不是标准 MCP 接入的必填项。',
                 '如果你要控制最终 `man({ "toolsetId": "<serverId>" })` 给 LLM 看的正式手册，请使用 `manual.contentFile` 指向 topic 文件目录前缀。',
-                '如果你只是想在 `team_mgmt` 的 MCP 章节里补充团队管理说明，也可以使用 inline `content`（总说明）+ `sections`（章节列表）。这类 inline 内容会展示在 `team_mgmt` 指南里，但不会替代运行时 `contentFile` 手册。',
+                '如果你只是想在 `team_mgmt` 的 MCP 章节里补充团队管理说明，也可以使用 inline `content`（总说明）+ `sections`（章节列表）。这类 inline 内容会展示在 `team_mgmt` 指南里，但不会替代 `contentFile` 手册。',
                 '重要：未配置 `manual` 并不代表 MCP toolset 不可用；但 Dominds 会在 Problems 中给 warning，提醒补充整体定位说明。标准工具元数据来自 MCP `tools/list`，智能体应继续依据各工具函数说明判断并使用。',
                 '团队管理者建议：配置并验证 MCP 后，优先依赖 MCP server 暴露的工具 description/参数；同时给每个 MCP server 的手册写一段简短整体定位，并列出工具名清单，明确“哪些工具属于这个 toolset”。只有当团队需要补充更完整的使用场景、安全边界、故障处置或协作规范时，再写 `manual.contentFile` 或更详细的 inline `content + sections`。',
                 '工具列表写法：列工具名即可，必要时补一句用途归类；不要复制参数 schema 或长篇工具 description，因为这些契约由系统提示里的函数工具定义提供。',
@@ -602,8 +602,8 @@ async function renderMcpManual(language) {
     return (fmtHeader('.minds/mcp.yaml') +
         fmtList([
             'Each MCP `serverId` registers one toolset, and the toolset name is exactly `serverId` (no `mcp_` prefix). Members choose MCP access via `members.<id>.toolsets`.',
-            'Hot reload: edits usually apply without restarting Dominds; use `mcp_restart` when needed. `mcp_restart` writes `enabled: true` when the target server is currently `enabled: false`, then tries to start it; after success it replaces the global runtime and clears all dialog leases on the old runtime.',
-            "Default is per-dialog MCP runtime leasing (safer): a dialog's first use may start/hold one runtime instance for that server (an HTTP connection or stdio process), and first use adds a sticky reminder. Call `mcp_release` when you're sure the current dialog no longer needs that runtime instance. This lease is about runtime resource ownership only; tool registration/visibility still follows the latest global server instance. If the server is truly stateless, set `truely-stateless: true` to allow cross-dialog sharing. To disable a server, use `mcp_disable` to write `enabled: false`; disabled servers remain visible as zero-tool toolsets with Raw MCP baseline/manual visibility.",
+            'Hot reload: edits usually apply without restarting Dominds; use `mcp_restart` when needed. `mcp_restart` writes `enabled: true` when the target server is currently `enabled: false`, then tries to start it; after success it replaces the global MCP tool registration and clears all dialog leases on the old connection/process.',
+            "Default is per-dialog MCP connection/process leasing (safer): a dialog's first use may start/hold one connection or process for that server (an HTTP connection or stdio process), and first use adds a sticky reminder. Call `mcp_release` when you're sure the current dialog no longer needs that connection/process. This lease is about connection/process resource ownership only; tool registration/visibility still follows the latest global server connection. If the server is truly stateless, set `truely-stateless: true` to allow cross-dialog sharing. To disable a server, use `mcp_disable` to write `enabled: false`; disabled servers remain visible as zero-tool toolsets with Raw MCP baseline/manual visibility.",
             'Stdio shape: `command` must be a string executable; parameters go in `args` (string[], optional, defaults to empty). Optional `cwd` (string) fixes the working directory used for relative paths.',
             'HTTP headers support three value forms: literal strings, `{ env: NAME }`, and `{ prefix: "Bearer ", env: NAME }`; keep auth tokens in env instead of hardcoding them in YAML.',
             'Use `tools.whitelist/blacklist` for exposure control. Top-level `transform` is the default ID transform for tools/prompts/resources/resource skills; nested `tools.transform`, `prompts.transform`, `resources.transform`, and `resources.skills.transform` override the enclosing transform instead of stacking with it. Explicit `transform: []` means no transform for that layer.',
@@ -611,7 +611,7 @@ async function renderMcpManual(language) {
             'High-frequency pitfall (stdio paths): if `cwd` is omitted, relative paths are resolved from Dominds process cwd (usually rtws root). Prefer setting `cwd` explicitly or use absolute paths. `cwd` must exist and be a directory.',
             'Optional manual field: `servers.<serverId>.manual` is enhancement guidance, not required for standard MCP use.',
             'To control the final formal manual shown to the LLM by `man({ "toolsetId": "<serverId>" })`, use `manual.contentFile` and point it at the topic-file directory prefix.',
-            'If you only want extra team-management guidance inside the `team_mgmt` MCP chapter, you may also use inline `content` (overview) + `sections` (chapter list). That inline content is shown in the `team_mgmt` guide, but it does not replace runtime `contentFile` manuals.',
+            'If you only want extra team-management guidance inside the `team_mgmt` MCP chapter, you may also use inline `content` (overview) + `sections` (chapter list). That inline content is shown in the `team_mgmt` guide, but it does not replace `contentFile` manuals.',
             'Important: an unconfigured `manual` does not mean the MCP toolset is unavailable, but Dominds reports a Problems warning to encourage an overall positioning note. Standard tool metadata comes from MCP `tools/list`; continue by reading each function-tool description.',
             'Team-manager recommendation: after MCP config is validated, rely on MCP-provided descriptions/arguments first; also write a short overall positioning note for each MCP server manual and list tool names to make clear which tools belong to this toolset. Add `manual.contentFile` or more detailed inline `content + sections` only when the team needs fuller guidance about use cases, safety boundaries, failure handling, or coordination norms.',
             'Tool-list writing rule: list tool names, with a short purpose group when useful; do not copy parameter schemas or long tool descriptions, because those contracts are provided by function-tool definitions in the system prompt.',

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "dominds",
-  "version": "1.26.2",
+  "version": "1.26.3",
   "description": "Dominds CLI and aggregation shell for the LongRun AI kernel/runtime packages.",
   "type": "commonjs",
   "publishConfig": {
@@ -50,12 +50,13 @@
     "@modelcontextprotocol/sdk": "^1.27.1",
     "async-mutex": "^0.5.0",
     "openai": "^6.35.0",
+    "selfsigned": "^5.5.0",
     "ws": "^8.19.0",
     "yaml": "^2.8.2",
     "zod": "^4.3.6",
     "@longrun-ai/codex-auth": "0.13.0",
-    "@longrun-ai/shell": "1.16.0",
-    "@longrun-ai/kernel": "1.16.0"
+    "@longrun-ai/kernel": "1.16.0",
+    "@longrun-ai/shell": "1.16.0"
   },
   "devDependencies": {
     "@types/node": "^25.3.5",

package/webapp/dist/assets/{_basePickBy-DUivPKf7.js → _basePickBy-DTJFxmut.js}

@@ -1,5 +1,5 @@
-import { e as isSymbol, c as baseFlatten, g as baseIteratee, k as keys, h as baseFindIndex, j as baseEach, l as arrayMap, m as hasPath, n as castPath, t as toKey, o as baseGet } from "./_baseUniq-hcs024vD.js";
-import { aU as isObject, aC as baseRest, aV as isIterateeCall, aW as keysIn, aX as eq, aY as isArrayLike, aZ as isArray, a_ as identity, a$ as isIndex, b0 as assignValue } from "./index-BOO_zFrI.js";
+import { e as isSymbol, c as baseFlatten, g as baseIteratee, k as keys, h as baseFindIndex, j as baseEach, l as arrayMap, m as hasPath, n as castPath, t as toKey, o as baseGet } from "./_baseUniq-CVu4lo4P.js";
+import { aU as isObject, aC as baseRest, aV as isIterateeCall, aW as keysIn, aX as eq, aY as isArrayLike, aZ as isArray, a_ as identity, a$ as isIndex, b0 as assignValue } from "./index-DYO4ZL8y.js";
 var reWhitespace = /\s/;
 function trimmedEndIndex(string) {
   var index = string.length;
@@ -193,4 +193,4 @@ export {
   map as m,
   toFinite as t
 };
-//# sourceMappingURL=_basePickBy-DUivPKf7.js.map
+//# sourceMappingURL=_basePickBy-DTJFxmut.js.map