Merge remote-tracking branch 'origin/master' into mergebot/pr883

# Conflicts: # packages/client/tsdown.client.ts # packages/client/ui-conversation/src/client/chat/ChatView.tsx # packages/client/ui-conversation/src/client/chat/PendingCard.tsx # packages/client/ui-conversation/src/client/chat/ToolRow.tsx # packages/client/ui-conversation/src/client/toolviews/todo-row.module.css # packages/client/ui-conversation/src/client/toolviews/todo-row.tsx # packages/client/ui-conversation/tests/chat-apply.spec.tsx # packages/client/ui-conversation/tests/coverage-tails.spec.tsx # packages/client/ui-conversation/tests/todo-panel.spec.tsx # packages/client/ui-primitives/src/icons/index.tsx # packages/client/ui-primitives/tests/icons.spec.tsx # packages/client/ui-question/README.i18n.yaml # packages/client/ui-question/src/client/QuestionComposer.module.css # packages/client/ui-question/src/client/QuestionComposer.tsx
2026-07-30 01:01:14 +08:00
parent 7639f4cb68 a58bc6f1e9
commit d9eb3545d8
@@ -1,6 +1,6 @@
 # Bilingual-pair consistency record (docs/i18n/README.md): the git blob hash of each
 # side as of the last confirmed-consistent state. Both languages carry equal authority;
 # after editing either side, bring the other along and re-record with:
-#   pnpm run verify-translation-pairing --write
+#   pnpm run verify-translation-pairing --write .agents/notes/README.md
 README.md: 3cfbb5154713046846a3bfcb2ccea62c0e4cb6c0
-README.zh.md: ddecac79519219c4a76cf9ba19edea312eea9d0d
+README.zh.md: a3369a94c6761e27567b1408d98a81665443f2ef
@@ -11,7 +11,7 @@
 - **生命周期**（顶层文件夹）是 Agent Note 的状态，Agent Note 随状态变化在文件夹之间移动：
  - **`proposed/`**：实施前评审的提案；尚未构建（或仅部分构建）。
  - **`implemented/`**：决策已交付。文件记录做了什么决定、否决了什么，并**与实际交付的内容保持同步**：当代码后续移动文件、重命名包（package）或更改键名/默认值时，Agent Note 在同一个变更中同步更新（仅限事实——路径、名称、结构——而非决策本身）。见 [implemented/AGENTS.md](implemented/AGENTS.md)。
-  - **`rejected/`**：提案经过讨论后被否决。仅当其决策依据仍能避免一种诱人且影响重大的错误时保留；否则删除完整的三个配对文件。
+  - **`rejected/`**：提案经过讨论后被否决。仅当其决策依据仍能避免一种诱人且影响重大的错误时保留；否则删除完整的英文、中文和伴随记录三文件组。
 - **类别**（嵌套文件夹）是决策的*种类*——见下方[分类](#classification)。

 文件名中的日期是该主题**首次提出**的时间（以 git 历史为准）。Agent Note 之间的交叉引用使用相对 Markdown 链接（`[topic](../../implemented/architecture/2026-…-….md)`），从不使用纯文字或编号，这样既可机械检查，也能在文件夹间移动时保持有效。
@@ -28,7 +28,7 @@
 |---|---|
 | `feature` | 面向用户或模型的新功能。 |
 | `bug-fix` | 修正缺陷或弥补事故复盘（postmortem）发现的缺口。 |
-| `simplification` | 在不增加功能的前提下移除代码、行为或对外表面积。 |
+| `simplification` | 在不增加功能的前提下移除代码、行为或对外范围。 |
 | `architecture` | 关于**交付源码**的结构性决策：包之间的关系、运行时词汇。 |
 | `process` | 代码**周边**的工具、策略或工作流——门禁、包管理器、vendor 化——不涉及运行时行为。 |
 | `testing` | 测试基础设施与策略。 |
@@ -39,13 +39,13 @@

 当一份 implemented Agent Note 记录的交付决策已经完整落地，且其决策依据不太可能再指导未来工作时，将其归档。如果其中的备选方案、归属边界、否定性保证、持久化语义或协议语义、安全规则，或者重新引入条件仍有价值，则继续作为活跃记录保留。绝不归档 proposed Agent Note：过时的提案应转为 rejected。仅当 rejected Agent Note 仍能避免一种可能发生的错误时保留；否则一并删除其英文、中文和伴随记录文件。请使用经过校准的 [`dsh-archive-agent-notes`](../skills/dsh-archive-agent-notes/SKILL.md) 工作流，不要根据字数、存续时间或目标配额来判断。

-归档路径编码为 `archived/{class}/yyyy-mm-dd-topic-title.md`；其中有意省略 `implemented`，因为只有 implemented Agent Note 可以进入归档。归档变更会移动完整的英文、中文和伴随记录三个文件，保留 `Status: implemented`，在两种语言的文件中紧接该状态行插入相同的 `Archived: YYYY-MM-DD` 行，重新记录伴随文件，并修复或删除入站链接。归档时只允许对内容做这些更改。
+归档路径编码为 `archived/{class}/yyyy-mm-dd-topic-title.md`；其中有意省略 `implemented`，因为只有 implemented Agent Note 可以进入归档。归档变更会移动完整的英文、中文和伴随记录三个文件，保留 `Status: implemented`，在两种语言的文件中紧接该状态行插入相同的 `Archived: YYYY-MM-DD` 行，重新记录伴随记录，并修复或删除入站链接。归档时只允许对内容做这些更改。

 封存后，每组归档文件都永久冻结。禁止编辑、翻译、重新格式化、更新、移动或删除，也不得将其视为当前行为的权威依据。文档门禁会跳过归档源文件，包括其中的出站链接；当活跃文档有意引用历史时，仍可链接到归档 Agent Note。[`verify-archived-agent-notes`](../../scripts/verify-archived-agent-notes.ts) 强制执行封闭的类别目录树、完整的三文件配对、归档元数据、伴随记录 hash，以及仅追加的冻结内容 manifest。[归档政策 Agent Note](implemented/process/2026-07-26-frozen-agent-note-archive.md) 记录了设计依据。

 ## 何时需要写一份

-每个非平凡变更都必须在同一 PR（Pull Request）中新增或更新至少一份 Agent Note。如果变更修改了行为、架构、跨文件或跨包契约、流程或工具、测试策略、磁盘、协议或配置格式，或者其他维护者可能合理重新审视的决策，就属于非平凡变更。对未来重大工作的提案从 `proposed/` 开始；已经做出的决策从 `implemented/` 开始。选择与决策匹配的类别文件夹（见[分类](#classification)）。
+每个非平凡变更都必须在同一 PR（Pull Request）中新增或更新至少一份 Agent Note。如果变更修改了行为、架构、跨文件或跨包契约、流程或工具、测试策略、磁盘存储格式、协议格式（wire format）或配置格式，或者其他维护者可能合理重新审视的决策，就属于非平凡变更。对未来重大工作的提案从 `proposed/` 开始；已经做出的决策从 `implemented/` 开始。选择与决策匹配的类别文件夹（见[分类](#classification)）。

 更新已经拥有该决策的 Agent Note 即可满足规则；不要创建重复记录。只有不涉及行为、契约、结构、流程或理由变化的纯机械性或局部编辑才可豁免。Agent Note 永远不会被编辑为一个*不同的决策*：用新 Agent Note 取代旧记录，并让两个记录保持互相链接，除非后续依据下方规则完全合并旧记录。编辑 `implemented/` Agent Note 以跟踪其现有决策的所在位置是必需的，而非禁止的；见 [implemented/AGENTS.md](implemented/AGENTS.md)。

@@ -112,7 +112,7 @@ Status: <status>

 ### 曾考虑的替代方案——必需

-每份 Agent Note 都必须包含 `## Alternatives considered` 章节：每个真实的替代方案及其落选原因，每个替代方案用一个加粗引导的段落，或对争议较大的替代方案用 `### Why not <X>?` 子节。记录决策时不记录它击败了什么，就是在邀请反复争论——正是这些 Agent Note 存在的意义所要防止的。
+每份 Agent Note 都必须包含 `## Alternatives considered` 章节：每个真实的替代方案及其落选原因，每个替代方案用一个加粗引导的段落，或对争议较大的替代方案用 `### Why not <X>?` 子节。记录决策时不记录它击败了什么，就是在邀请反复争论——这正是 Agent Note 旨在防止的问题。

 替代方案是记录下来的，不是凭空编造的。日期早于 2026-07-05 且替代方案无法从记录中重建的 Agent Note，在该章节位置放置以下精确注释，门禁仅对格式规范之前的文件接受此注释：

@@ -122,8 +122,8 @@ Status: <status>

 ### 在生命周期之间移动

-将文件在生命周期文件夹之间移动意味着在同一个变更中更新 `Status:` 行并满足目标文件夹的骨架要求——否则门禁会失败。具体而言，`proposed/` → `implemented/` 将 `## Proposal` 改写为现在时态的 `## Decision`，将 `## Acceptance criteria` 和 `## Risks` 折入 `## Consequences`（或折入一个现在时态的 `## Testing`/`## Verification` 章节，用于描述现在锁定该行为的内容），并用实际交付的内容替换计划——即 [implemented/AGENTS.md](implemented/AGENTS.md) 所要求的改写，使之机械化。`proposed/` → `rejected/` 仅在 `Status:` 行添加原因并冻结文件。
+将文件在生命周期文件夹之间移动意味着在同一个变更中更新 `Status:` 行并满足目标文件夹的骨架要求——否则门禁会失败。具体而言，`proposed/` → `implemented/` 将 `## Proposal` 改写为现在时态的 `## Decision`，将 `## Acceptance criteria` 和 `## Risks` 折入 `## Consequences`（或折入一个现在时态的 `## Testing`/`## Verification` 章节，用于描述现在锁定该行为的内容），并用实际交付的内容替换计划——也就是将 [implemented/AGENTS.md](implemented/AGENTS.md) 所要求的改写变成可机械检查的规则。`proposed/` → `rejected/` 仅在 `Status:` 行添加原因并冻结文件。

 ### 中文对侧文件

-`.zh.md` 对侧文件按 [i18n 契约](../../docs/i18n/README.md)逐章节镜像其英文兄弟文件的结构；机器检查的头部标记（`# Agent Note: ` 和 `Status:` 行）保持英文原样不翻译。格式门禁跳过 `.zh.md` 文件——配对门禁负责它们的一致性。
+`.zh.md` 对侧文件按 [i18n 契约](../../docs/i18n/README.md)逐章节镜像其英文对侧文件的结构；机器检查的头部标记（`# Agent Note: ` 和 `Status:` 行）保持英文原样不翻译。格式门禁跳过 `.zh.md` 文件——配对门禁负责它们的一致性。
@@ -1,6 +1,6 @@
 # Bilingual-pair consistency record (docs/i18n/README.md): the git blob hash of each
 # side as of the last confirmed-consistent state. Both languages carry equal authority;
 # after editing either side, bring the other along and re-record with:
-#   pnpm run verify-translation-pairing --write
-2026-06-14-session-persistence.md: 75e13b860f621ed407849b3b4c62ff7287ab4812
-2026-06-14-session-persistence.zh.md: a6bd400a053779c742940236737447d1687622de
+#   pnpm run verify-translation-pairing --write .agents/notes/implemented/architecture/2026-06-14-session-persistence.md
+2026-06-14-session-persistence.md: 137b2b01126214629952812f3dd3b71985a3acda
+2026-06-14-session-persistence.zh.md: 2846ee92349c297fb3a173ba9dd3e2ff3cd9ee1a
@@ -29,7 +29,7 @@ Key choices recorded here because they are durable, contested, and surprising:

 Each key choice above records its rejected alternative where the choice is stated: a **chunk-filtered canonical log** (Codex's `policy.rs` shape) — breaks the contiguous-seq contract; **truncating a crashed turn** — silently destroys a long autonomous run's real work; an **in-log `session/meta` event as line 0** — metadata is not replayable state; **finite fractional `createdAt` values** — have no producer and diverge from integer Unix-millisecond storage and query columns; **adopting a non-pristine unversioned SQLite file** — can overwrite unrelated objects or identity; **hard-injecting `sessionPersistence` into the loop** — would pend non-persistent demos forever.

-Format versioning: the header carries a `version`; `load` rejects any non-current version (no migration — the pre-release session format is pinned at `SESSION_FORMAT_VERSION = 0` and absorbs shape churn, per the AGENTS.md pre-release stance). Stated honestly: append-only + flush is robust to partial trailing writes (tolerated on load) but not to fsync-less power loss mid-line; a DB/WAL backend is the stronger option later.
+Format versioning: the header carries a `version`; `load` rejects any non-current version. The pre-release session format stays pinned at `SESSION_FORMAT_VERSION = 0` and carries no broad compatibility promise, while the coordinator may own an explicit narrow import upgrade when persisted user data requires it ([pre-identity message recovery](../bug-fix/2026-07-28-load-pre-identity-session-messages.md)). Stated honestly: append-only + flush is robust to partial trailing writes (tolerated on load) but not to fsync-less power loss mid-line; a DB/WAL backend is the stronger option later.

 ## Consequences

@@ -29,7 +29,7 @@ Status: implemented

 上述每个关键选择都在陈述处记录了被否决的替代方案：**过滤分片的规范日志**（Codex 的 `policy.rs` 形式）破坏连续 seq 契约；**截断崩溃的轮次**会静默销毁长时间自主运行中的真实工作；**日志内 `session/meta` 事件作为第 0 行**——元数据不是可回放状态；**有限的非整数 `createdAt` 值**没有生产方，且与整数 Unix 毫秒存储及查询列不一致；**接受非全新的未版本化 SQLite 文件**可能覆盖无关对象或应用标识；**将 `sessionPersistence` 硬注入循环**会让非持久化的演示永远挂起。

-格式版本控制：header 携带一个 `version`；`load` 拒绝任何非当前版本（不做迁移——预发布阶段的会话格式固定为 `SESSION_FORMAT_VERSION = 0` 并吸收形状变动，遵循 AGENTS.md 的预发布立场）。坦率地说：仅追加 + 刷写对部分尾部写入是健壮的（加载时容忍），但对行写入中途的无 fsync 断电不健壮；数据库/WAL 后端是后续更强的选项。
+格式版本控制：header 携带一个 `version`；`load` 拒绝任何非当前版本。预发布阶段的会话格式仍固定为 `SESSION_FORMAT_VERSION = 0`，不承诺广泛兼容；当持久化用户数据确有需要时，协调器可以负责显式且范围受限的导入升级（[消息标识机制引入前的消息恢复](../bug-fix/2026-07-28-load-pre-identity-session-messages.md)）。坦率地说：仅追加 + 刷写对部分尾部写入是健壮的（加载时容忍），但对行写入中途的无 fsync 断电不健壮；数据库/WAL 后端是后续更强的选项。

 ## 后果

@@ -1,6 +1,6 @@
 # Bilingual-pair consistency record (docs/i18n/README.md): the git blob hash of each
 # side as of the last confirmed-consistent state. Both languages carry equal authority;
 # after editing either side, bring the other along and re-record with:
-#   pnpm run verify-translation-pairing --write
-2026-07-19-gui-layering-and-rpc-protocol.md: 63db4786adcc007d09b7a58824a59f4d1e1e8be1
-2026-07-19-gui-layering-and-rpc-protocol.zh.md: b3037ceb8c172925581d2862ea675e53a7f8c54e
+#   pnpm run verify-translation-pairing --write .agents/notes/implemented/architecture/2026-07-19-gui-layering-and-rpc-protocol.md
+2026-07-19-gui-layering-and-rpc-protocol.md: b9718da4725316c64686adef24827e2984d8723d
+2026-07-19-gui-layering-and-rpc-protocol.zh.md: 2add148054e8f97c65600cd719fb4f8e0283f52d
@@ -165,7 +165,7 @@ One example row (the table structure is the reading key):
 |---|---|---|---|
 | `session.list` | `{ cursor?: string }` (cursor is a reserved seat, unimplemented) | `{ items: SessionSummary[] }` | persisted sessions, updatedAt descending; v1 builds no index |

-The remaining methods (`session.create`/`session.history`/`session.prompt`/`session.cancel`/`host.describe`) are not re-copied here — signatures are the source of truth; see `api/sessions.ts`, `api/host.ts`, and `RpcMethodMap`.
+The remaining methods (`session.create`/`session.history`/`session.rename`/`session.prompt`/`session.cancel`/`host.describe`) are not re-copied here — signatures are the source of truth; see `api/sessions.ts`, `api/host.ts`, and `RpcMethodMap`.

 ### Frames (server→client, named unions)

@@ -187,7 +187,7 @@ The remaining frame types are not re-copied here; the full unions are `MuxFrame`
 - **Cold sessions resume implicitly**: when `history`/`prompt` hits an unattached session the impl auto-resumes, deduplicating concurrent triggers with an in-flight table; attachment status is not exposed to clients (`running` already covers it).
 - **Approvals/questions**: the requested frame mints a stable rpcId on acceptance; first answer wins, and the host's in-memory pending table (keyed by rpcId) is the only referee; after a mux reopen, still-pending requested frames replay after the subscribed frame (rpcId reused verbatim — refresh recovery). The audit events `approval/asked`/`decided` continue through the durable log — frames = the live control plane, events = the durable audit. **Status**: the contract and frame types are shipped; the host-side pending table/wire answerer is unimplemented (`respond` in `api-proxy.ts` is a stub, always `not-pending`); PendingCard v1 is display-only.
 - **No protocol version**: client and host release bound together; `host.describe` has no protocolVersion field; introduce one when an independently released client appears.
- **Reserved-seam discipline**: the map holds only implemented methods; an unknown method fails loud at envelope parse (`bad-request`) — no not-implemented fallback code. The reservation list (implementing = copy the signature into the domain interface + add the map row + add the schema pair): `session.fork`, `prompt.mode` gaining `'inject'`, `task.list`, `host.listModels`, describe gaining `hostInstanceId`.
+- **Reserved-seam discipline**: the map holds only implemented methods; an unknown method fails loud at envelope parse (`bad-request`) — no not-implemented fallback code. The reservation list (implementing = copy the signature into the domain interface + add the map row + add the schema pair): `session.fork`, `prompt.mode` gaining `'inject'`, `task.list`, `host.listModels`, describe gaining `hostInstanceId`. (`session.rename` graduated from this list: it appends a user-source `session/title` event.)

 ## The client carrier: the AbstractApiClient class family (`fetch/client.ts`)

@@ -163,7 +163,7 @@ export type ResponseValue<K> =
 |---|---|---|---|
 | `session.list` | `{ cursor?: string }`（cursor 留座不实现） | `{ items: SessionSummary[] }` | 已持久化 session，updatedAt 倒序；v1 不建索引 |

-其余方法（`session.create`/`session.history`/`session.prompt`/`session.cancel`/`host.describe`）的参数与返回不在此复写——签名即事实源，见 `api/sessions.ts`、`api/host.ts` 与 `RpcMethodMap`。
+其余方法（`session.create`/`session.history`/`session.rename`/`session.prompt`/`session.cancel`/`host.describe`）的参数与返回不在此复写——签名即事实源，见 `api/sessions.ts`、`api/host.ts` 与 `RpcMethodMap`。

 ### 帧（server→client，具名 union）

@@ -185,7 +185,7 @@ export type ResponseValue<K> =
 - **冷 session 隐式 resume**：`history`/`prompt` 命中未 attach 的 session 时 impl 自动 resume，并发触发用在途表去重；attach 与否不对客暴露（`running` 已覆盖）。
 - **审批/问答**：requested 帧受理时 mint 稳定 rpcId；先到先赢，host 内存 pending 表（keyed by rpcId）是唯一裁判；mux 重开后在 subscribed 帧后重放仍 pending 的 requested 帧（rpcId 原样复用，刷新恢复）。审计事件 `approval/asked`/`decided` 照旧走 durable 日志——帧=live 控制面，事件=durable 审计。**现状**：契约与帧类型已 shipped，host 侧 pending 表/wire answerer 未实现（`api-proxy.ts` 的 `respond` 是 stub，恒回 `not-pending`）；PendingCard v1 只展示。
 - **不设协议版本**：client 与 host 绑定发布，`host.describe` 无 protocolVersion 字段；出现独立发布的 client 时再引入。
- **预留接缝纪律**：map 只含已实现方法，未知 method 在信封 parse 即 fail loud（`bad-request`），不设 not-implemented 兜底码。预留清单（实现时把签名抄进域接口+map 加行+schema 加对即升格）：`session.fork`、`prompt.mode` 加 `'inject'`、`task.list`、`host.listModels`、describe 加 `hostInstanceId`。
+- **预留接缝纪律**：map 只含已实现方法，未知 method 在信封 parse 即 fail loud（`bad-request`），不设 not-implemented 兜底码。预留清单（实现时把签名抄进域接口+map 加行+schema 加对即升格）：`session.fork`、`prompt.mode` 加 `'inject'`、`task.list`、`host.listModels`、describe 加 `hostInstanceId`。（`session.rename` 已从本清单毕业：追加 user 来源的 `session/title` 事件。）

 ## 客户端载体：AbstractApiClient 类体系（`fetch/client.ts`）

@@ -2,5 +2,5 @@
 # side as of the last confirmed-consistent state. Both languages carry equal authority;
 # after editing either side, bring the other along and re-record with:
 #   pnpm run verify-translation-pairing --write .agents/notes/implemented/architecture/2026-07-22-unified-send-and-coalesced-user-messages.md
-2026-07-22-unified-send-and-coalesced-user-messages.md: 6936fbfa04c0fdaf1a8786c0465c193e9c285243
-2026-07-22-unified-send-and-coalesced-user-messages.zh.md: 3af14359fa01e92f63ae3b3e51dced9a97f6419f
+2026-07-22-unified-send-and-coalesced-user-messages.md: ed171735cf483938c70291963a6e68dc02d7bde2
+2026-07-22-unified-send-and-coalesced-user-messages.zh.md: 8b2a3ebabb493954e653255e876255b9c0810c19
@@ -12,21 +12,21 @@ Separately, `context/message` and `user/message` had converged: the surface proj

 ## Decision

-**One primitive, three preset aliases.** The `Agent` interface's `send(input, { target, wakeup })` covers the (`target` × `wakeup`) matrix. Its `UserMessageData` input owns the inseparable model-facing `content` and producer `source`; the complete `SendOptions` owns only routing policy. `followup` (`next-turn`/wakeup), `steer` (`next-step`/wakeup), and `inject` (`next-step`/no-wakeup) each accept that one input and fix the policy. `wakeup` means "make the model run": wake a parked driver for a `next-turn` item, or force a continuation for a running `next-step` item. `next-turn`/no-wakeup (queue without waking) is representable with no alias and no current caller.
+**One primitive, three preset aliases.** The `Agent` interface's `send(message, { target, wakeup })` covers the (`target` × `wakeup`) matrix. Its complete `UserMessage` owns identity, role, model-facing `content`, and producer `source`; the complete `SendOptions` owns only routing policy. `followup` (`next-turn`/wakeup), `steer` (`next-step`/wakeup), and `inject` (`next-step`/no-wakeup) each accept that one message and fix the policy. `wakeup` means "make the model run": wake a parked driver for a `next-turn` item, or force a continuation for a running `next-step` item. `next-turn`/no-wakeup (queue without waking) is representable with no alias and no current caller.

-**inject keeps its mechanism.** The `next-step`/no-wakeup path is exactly the old `inject`: durable model-facing context appended at the current log position, deferred while prompt admission or a turn owns the next safe boundary, and appended directly outside that window. It bypasses the FIFOs entirely, while its required `UserMessageData.source` preserves the caller's explicit provenance.
+**inject keeps its mechanism.** The `next-step`/no-wakeup path is exactly the old `inject`: durable model-facing context appended at the current log position, deferred while prompt admission or a turn owns the next safe boundary, and appended directly outside that window. It bypasses the FIFOs entirely, while its required `UserMessage.source` preserves the caller's explicit provenance.

 **context/message is gone.** Injected context is now a `user/message`; context producers supply the appropriate non-user `source` explicitly, and typed source variants carry any domain-specific durable provenance. The surface, derivation, and `SurfaceEventType` drop `context/message`; consumers that need "is this a human prompt?" read `source.kind === 'user'` instead of the event type.

 **Goal replay disambiguates by round, not type.** A goal state change is a round-zero goal-sourced `user/message` whose source carries the complete change; a positive round is an admitted continuation prompt. `decodeGoalEvent` takes a `user/message` and fails loud when goal-state content and its typed source disagree.

-**`send` returns an id.** `send` (and the aliases) return an opaque branded `AgentMessageId` for the accepted message; `send`'s previous return was `void`.
+**`send` does not return identity.** Callers already own the complete message and its opaque `MessageId`; creation and freezing are owned by the [identified immutable message decision](2026-07-28-identified-immutable-message-values.md), not by routing.

-**Three inbox events replace agent/queued.** `agent/inbox/enqueue` (an item entered a FIFO), `agent/inbox/dequeue` (the driver claimed one), and `agent/inbox/discard` (`cancel()` dropped pending items) type their `AgentMessage` payload with only the accepted message's returned `id`, content, and source. Enqueue separately carries the resolved `queued | steering` placement captured by the producer at acceptance time, so observers and reconnect mirrors never reconstruct routing from later status or session history. Injection never touches a FIFO and emits none of these. Every FIFO entry publishes an enqueue, including steering submitted by an `agent/turn-stopping` listener, so the ledger stays balanced with its later dequeue or discard. The `dsh-agent` invariant companion asserts FIFO conservation: a per-agent outstanding count that dequeue and discard can never drive negative.
+**Three inbox events replace agent/queued.** `agent/inbox/enqueue` (an item entered a FIFO), `agent/inbox/dequeue` (the driver claimed one), and `agent/inbox/discard` (`cancel()` dropped pending items) carry the accepted `UserMessage`. Enqueue and dequeue also carry the resolved `queued | steering` placement captured at acceptance, so observers and reconnect mirrors retire repeated message identities from the correct FIFO without reconstructing routing from later status or session history. Injection never touches a FIFO and emits none of these. Every FIFO entry publishes an enqueue, including steering submitted by an `agent/turn-stopping` listener, so the ledger stays balanced with its later dequeue or discard. The `dsh-agent` invariant companion asserts FIFO conservation: a per-agent outstanding count that dequeue and discard can never drive negative.

 **Admission accepts next-step input without becoming a turn.** The loop opens a private next-step acceptance window before `agent/prompt-submit`, keeps it open through the turn, and closes it before `turn/end`. Steering and injection received during admission therefore remain together in the outbox and join an allowed turn. If admission blocks or fails, a context-only caller batch takes idle injection's immediate append, while steering and context staged beside it remain available to retry; neither path writes the rejected prompt. When a later prompt is admitted, retained outbox input enters its turn before that prompt, while input accepted during the current admission remains after the prompt. Closing the window before `turn/end` preserves the rule that reentrant late steering becomes an independent queued turn. `Agent.acceptsNextStep` exposes whether a `next-step` send would currently join this window; `status` remains the broader activity signal rather than a routing predicate.

-**One accepted message keeps one representation.** Durable user-role input and additional model-facing context both use `UserMessageData { content, source }` directly; public `AgentMessage` extends it with the correlation `id`, and the loop-private `PendingMessage` extends that with `wakeup`. The loop clones and freezes `UserMessageData` before publication, queueing, or immediate append, so later caller or observer mutation cannot change the accepted value. A queued message that becomes steering enters the outbox as the same `PendingMessage` object, while injected and tool-produced context enters as plain `UserMessageData`. The outbox therefore stores their union directly instead of wrapping steering beside a duplicate copy of its content and source. Provider-native assistant messages remain adapter-owned output types and do not participate in this input hierarchy.
+**One accepted message keeps one representation.** Durable user-role input and additional model-facing context both use the identified, frozen `UserMessage` directly. The loop stores that value beside private routing state rather than copying its identity, content, or source into another public shape. A queued message that becomes steering keeps the same message value in the outbox, while injected and tool-produced context each carry their own identified message. The [identified immutable message decision](2026-07-28-identified-immutable-message-values.md) supersedes this note's former `UserMessageData`/`AgentMessage` hierarchy and extends the representation to assistant and tool-result messages.

 **Idle wakeup follows acceptance.** Before publishing enqueue, a waking queued send installs quiescence ownership and schedules driver admission for a microtask that runs after the id returns. Every send in one synchronous caller stack therefore resolves placement against the same pre-admission state, while reentrant cancellation or teardown cannot retire before the scheduled admission settles. Two idle `steer()` calls remain two FIFO turns instead of the first opening an admission window that captures the second.

@@ -35,7 +35,7 @@ Separately, `context/message` and `user/message` had converged: the surface proj
 ## Alternatives considered

 - **A dedicated `MessageSource` kind `context`** for injected content. Rejected because `plugin` already means "not a human," so a fourth kind would add a parallel axis the authority checks would have to learn. Plugin-produced injected context supplies its plugin source explicitly.
- **A typed discriminant field on `UserMessageData`** (e.g. `origin: 'prompt' | 'context'`) to replace the event-type split. Rejected in favor of `source`, which every consumer already carries and which the goal system already keyed on; a second discriminant would duplicate that fact.
+- **A typed discriminant field on `UserMessage`** (e.g. `origin: 'prompt' | 'context'`) to replace the event-type split. Rejected in favor of `source`, which every consumer already carries and which the goal system already keyed on; a second discriminant would duplicate that fact.
 - **Keeping `agent/queued` alongside the inbox events.** Rejected as a mirror: `agent/inbox/enqueue` is the same enqueue-time signal with the resolved placement, and the dequeue/discard events complete the FIFO lifecycle the single event could not describe.
 - **Derive inbox placement from agent status or the session log.** Rejected because `running` includes admission and settlement, while reconnect baselines need the original acceptance result even when the earlier turn boundary is absent. The producer already owns the exact routing decision.

@@ -50,3 +50,4 @@ The delivery surface is now one primitive plus three self-documenting presets, a
 - [one-send-one-turn](../simplification/2026-07-17-one-send-one-turn.md) — the one-claimed-message-per-turn rule this builds on.
 - [remove-agent-steering-mirror](../../archived/simplification/2026-07-04-remove-agent-steering-mirror.md) — the precedent for collapsing a mirrored live event.
 - [explicit-turn-cancellation](2026-07-16-explicit-turn-cancellation.md) — the cancel-cause signal `keepInbox` extends.
+- [identified immutable message values](2026-07-28-identified-immutable-message-values.md) — the message identity and representation contract that now underlies this routing decision.
@@ -12,21 +12,21 @@ agent 的对外驱动接口逐渐长出三个近乎平行的动词——`send`

 ## 决策

-**一个原语，三个预设别名。** `Agent` 接口的 `send(input, { target, wakeup })` 覆盖 (`target` × `wakeup`) 矩阵。其 `UserMessageData` 输入持有不可分割的模型可见 `content` 与生产方 `source`；完整的 `SendOptions` 只持有路由策略。`followup`（`next-turn`/wakeup）、`steer`（`next-step`/wakeup）和 `inject`（`next-step`/no-wakeup）都接收这一项输入并固定策略。`wakeup` 意为“让模型运行”：为一个 `next-turn` 队列项唤醒处于停泊状态的驱动器，或为一个运行中的 `next-step` 队列项强制继续执行。`next-turn`/no-wakeup（入队但不唤醒）可以表达，只是没有别名，也没有当前调用方。
+**一个原语，三个预设别名。** `Agent` 接口的 `send(message, { target, wakeup })` 覆盖 (`target` × `wakeup`) 矩阵。完整的 `UserMessage` 持有标识、角色、模型可见 `content` 与生产方 `source`；完整的 `SendOptions` 只持有路由策略。`followup`（`next-turn`/wakeup）、`steer`（`next-step`/wakeup）和 `inject`（`next-step`/no-wakeup）都接收这一条消息并固定策略。`wakeup` 意为“让模型运行”：为一个 `next-turn` 队列项唤醒处于停泊状态的驱动器，或为一个运行中的 `next-step` 队列项强制继续执行。`next-turn`/no-wakeup（入队但不唤醒）可以表达，只是没有别名，也没有当前调用方。

-**inject 保留其机制。** `next-step`/no-wakeup 路径正是旧的 `inject`：持久的面向模型上下文会追加到当前日志位置；提示词准入或一个轮次占有下一个安全边界时，它会延迟处理，而在该窗口之外则直接追加。它完全绕过 FIFO 队列，而必填的 `UserMessageData.source` 会保留调用方显式提供的来源信息。
+**inject 保留其机制。** `next-step`/no-wakeup 路径正是旧的 `inject`：持久的面向模型上下文会追加到当前日志位置；提示词准入或一个轮次占有下一个安全边界时，它会延迟处理，而在该窗口之外则直接追加。它完全绕过 FIFO 队列，而必填的 `UserMessage.source` 会保留调用方显式提供的来源信息。

 **context/message 已移除。** 注入的上下文现在是一条 `user/message`；上下文生产方显式提供合适的非 `user` 类别 `source`，类型化 source 变体携带所有特定于领域的持久来源信息。对外接口、派生逻辑和 `SurfaceEventType` 都不再包含 `context/message`；需要判断“这是不是一条人类提示词？”的消费方改为读取 `source.kind === 'user'`，而不是事件类型。

 **goal 回放靠轮次而非类型来区分。** 一次 goal 状态变更是一条第 0 轮、来源为 goal 的 `user/message`，其 source 携带完整变更；正数轮次则是一条已准入的继续执行提示词。`decodeGoalEvent` 接收一条 `user/message`，并在 goal 状态内容与其类型化 source 不一致时立即报错。

-**`send` 返回一个 id。** `send`（以及其别名）为被接受的消息返回一个不透明的 branded `AgentMessageId`；`send` 此前的返回值是 `void`。
+**`send` 不返回标识。** 调用方已经持有完整消息及其不透明的 `MessageId`；消息的创建与冻结由[带标识的不可变消息值决策](2026-07-28-identified-immutable-message-values.md)负责，而不是由路由负责。

-**三个 inbox 事件取代 agent/queued。** `agent/inbox/enqueue`（一个队列项进入某个 FIFO）、`agent/inbox/dequeue`（驱动器认领了一个）和 `agent/inbox/discard`（`cancel()` 丢弃了待处理项）都将各自的 `AgentMessage` 载荷类型限定为仅包含被接受消息所返回的 `id`、内容和来源。enqueue 还会单独携带生产方在接受消息时捕获的已解析 `queued | steering` 放置方式，因此观察方和重连镜像永远不必根据后续状态或会话历史重建路由。注入从不触及 FIFO，也不发出这些事件中的任何一个。每一次 FIFO 入队都会发布一个 enqueue 事件，包括 `agent/turn-stopping` 监听器提交的 steering，因此账目会与其后的 dequeue 或 discard 保持平衡。`dsh-agent` 的不变量配套断言 FIFO 守恒：一个按 agent 计的未结算计数，dequeue 和 discard 永远无法把它压到负数。
+**三个 inbox 事件取代 agent/queued。** `agent/inbox/enqueue`（一个队列项进入某个 FIFO）、`agent/inbox/dequeue`（驱动器认领了一个）和 `agent/inbox/discard`（`cancel()` 丢弃了待处理项）都会携带已接受的 `UserMessage`。enqueue 和 dequeue 还会携带生产方在接受消息时捕获的已解析 `queued | steering` 放置方式，因此观察方和重连镜像可以从正确的 FIFO 中结算重复出现的消息标识，无需根据后续状态或会话历史重建路由。注入从不触及 FIFO，也不发出这些事件中的任何一个。每一次 FIFO 入队都会发布一个 enqueue 事件，包括 `agent/turn-stopping` 监听器提交的 steering，因此账目会与其后的 dequeue 或 discard 保持平衡。`dsh-agent` 的不变量配套断言 FIFO 守恒：一个按 agent 计的未结算计数，dequeue 和 discard 永远无法把它压到负数。

 **准入接受 next-step 输入，但不会因此成为一个轮次。** 循环会在 `agent/prompt-submit` 前打开一个私有的 next-step 接受窗口，使其贯穿整个轮次，并在 `turn/end` 前关闭。因此，在准入期间收到的 steering 和注入会一起留在 outbox 中并加入获准轮次。如果准入被阻止或失败，仅含调用方上下文的批次会采用空闲注入的立即追加行为，而 steering 及与其一同暂存的上下文仍可重试；两种路径都不会写入被拒绝的提示词。后续提示词获准时，保留在 outbox 中的输入会先于该提示词进入其轮次，而当前准入期间接受的输入则留在提示词之后。在 `turn/end` 前关闭窗口，可以保留这样的规则：可重入的晚到 steering 会成为一个独立的排队轮次。`Agent.acceptsNextStep` 会公开一次 `next-step` 发送当前是否会加入该窗口；`status` 仍是更宽泛的活动信号，而非路由判据。

-**一条已接受消息只保留一种表示。** 持久的用户角色输入和附加的模型可见上下文都直接使用 `UserMessageData { content, source }`；公开的 `AgentMessage` 在此基础上增加用于关联的 `id`，循环私有的 `PendingMessage` 再增加 `wakeup`。循环会在发布、入队或立即追加前克隆并冻结 `UserMessageData`，因此调用方或观察方后续的修改无法改变已接受的值。一条成为 steering 的排队消息会以同一个 `PendingMessage` 对象进入 outbox，而注入和工具产生的上下文则以普通 `UserMessageData` 进入。因此，outbox 直接存储这两种类型的联合，而不再把 steering 与一份重复的内容和来源副本包装在一起。提供方原生的助手消息仍是适配器拥有的输出类型，不参与这套输入层级。
+**一条已接受消息只保留一种表示。** 持久的用户角色输入和附加的模型可见上下文都直接使用带标识且冻结的 `UserMessage`。循环把该值与私有路由状态存放在一起，不会将其标识、内容或来源复制到另一种公开形状中。一条成为 steering 的排队消息会在 outbox 中保留同一个消息值，而注入和工具产生的上下文则各自携带带标识的消息。[带标识的不可变消息值决策](2026-07-28-identified-immutable-message-values.md)取代了本记录此前的 `UserMessageData`/`AgentMessage` 层级，并将这一表示扩展到 assistant 消息和工具结果消息。

 **空闲唤醒在接受之后发生。** 在发布 enqueue 前，一次会唤醒驱动器的排队发送会先取得完全停稳所有权，并把驱动器准入调度到一个会在该次发送返回 id 后运行的微任务中。因此，同一同步调用栈中的每次发送都会基于同一份准入前状态解析放置方式，而可重入的取消或拆除在已调度的准入结算前无法完成退役。空闲时的两次 `steer()` 调用会保留为两个 FIFO 轮次，而不会因第一次调用打开准入窗口而把第二次吸纳进去。

@@ -35,7 +35,7 @@ agent 的对外驱动接口逐渐长出三个近乎平行的动词——`send`
 ## 考虑过的替代方案

 - **为注入内容设立专门的 `MessageSource` 类别 `context`。** 不予采纳，因为 `plugin` 已经表示“不是人类”，因此第四种类别会增加一条平行的轴，让授权检查不得不去学习它。由插件产生的注入上下文会显式提供其 plugin 来源。
- **在 `UserMessageData` 上设一个类型化的判别字段**（例如 `origin: 'prompt' | 'context'`）来取代事件类型的区分。不予采纳，转而采用 `source`——每个消费方都已经携带它，goal 系统也已经以它为键；第二个判别字段会重复这一事实。
+- **在 `UserMessage` 上设一个类型化的判别字段**（例如 `origin: 'prompt' | 'context'`）来取代事件类型的区分。不予采纳，转而采用 `source`——每个消费方都已经携带它，goal 系统也已经以它为键；第二个判别字段会重复这一事实。
 - **在 inbox 事件之外保留 `agent/queued`。** 作为镜像而被否决：`agent/inbox/enqueue` 是同一个入队时刻的信号，只是带有已解析的放置方式，而 dequeue/discard 事件补全了单个事件无法描述的 FIFO 生命周期。
 - **根据 agent 状态或会话日志推导 inbox 放置方式。** 不予采纳，因为 `running` 同时涵盖准入与结算，而重连基线即使缺少此前的轮次边界，也需要最初的接受结果。生产方已经拥有精确的路由决策。

@@ -50,3 +50,4 @@ agent 的对外驱动接口逐渐长出三个近乎平行的动词——`send`
 - [one-send-one-turn](../simplification/2026-07-17-one-send-one-turn.md)——本决策所依托的“每轮次只认领一条消息”规则。
 - [remove-agent-steering-mirror](../../archived/simplification/2026-07-04-remove-agent-steering-mirror.md)——折叠镜像实时事件的先例。
 - [explicit-turn-cancellation](2026-07-16-explicit-turn-cancellation.md)——`keepInbox` 所扩展的取消原因信号。
+- [带标识的不可变消息值](2026-07-28-identified-immutable-message-values.md)——本路由决策现在所依托的消息标识与表示契约。
@@ -2,5 +2,5 @@
 # side as of the last confirmed-consistent state. Both languages carry equal authority;
 # after editing either side, bring the other along and re-record with:
 #   pnpm run verify-translation-pairing --write .agents/notes/implemented/architecture/2026-07-24-separate-context-injection-from-turn-execution.md
-2026-07-24-separate-context-injection-from-turn-execution.md: d44ef5afc8c376790192998bcf3069ceb651ae82
-2026-07-24-separate-context-injection-from-turn-execution.zh.md: ba561b587effb278a67844ce4d876da9cd940e94
+2026-07-24-separate-context-injection-from-turn-execution.md: bf3ae2ecbd2205a4c49e8004ffc694f89a2460a3
+2026-07-24-separate-context-injection-from-turn-execution.zh.md: a805eb651c5c77f3d37c92dacd116bb41f154ed7
@@ -18,7 +18,7 @@ Idle `inject()` exposed a second mismatch. Injection did not request model execu

 `inject()` is the only caller-facing operation for supplementary model-facing input, and a turn means one execution of the model loop.

-`SendOptions` contains only `target` and `wakeup`. A caller that owns context delivers `UserMessageData` through `inject()` and submits the direct message independently with `send()` or `steer()`.
+`SendOptions` contains only `target` and `wakeup`. A caller that owns context delivers an identified, frozen `UserMessage` through `inject()` and submits the direct message independently with `send()` or `steer()`.

 Prompt and tool extension points still return `additionalContexts`. These values are outputs of the extension point, not attachments captured from a caller's inbox item. Prompt admission runs before `run()` opens a turn. An allowed prompt and its returned additional contexts enter the new turn as separate messages; a blocked prompt writes neither and opens no turn. Tool-produced additional contexts enter the outbox after the corresponding tool results.

@@ -59,7 +59,7 @@ This decision preserves the caller-owned framing decision from [unwrapped inject
 ## Verification

 - `SendOptions` and steering inbox records contain no attached contexts; `agent/inbox/enqueue` reports only the message plus its resolved queued-or-steering placement.
- `UserMessageData` is the shared shape across prompt interception, tool execution, hook bridges, guards, and context producers.
+- `UserMessage` is the shared identified, frozen shape across prompt interception, tool execution, hook bridges, guards, and context producers.
 - Prompt-prefix placement, prompt envelopes, and `context/message` are absent from public types, durable events, projection, and UI replay.
 - Idle `inject()` appends one sourced `user/message` without a turn or model call.
 - Admission-time and active-turn injection drain at safe boundaries after complete tool-result batches and before the request that consumes them.
@@ -18,7 +18,7 @@ agent API 曾用三种相互重叠的方式表示面向模型的补充输入：

 `inject()` 是调用方交付补充模型输入的唯一操作，而轮次表示一次模型循环执行。

-`SendOptions` 只包含 `target` 和 `wakeup`。拥有上下文的调用方通过 `inject()` 交付 `UserMessageData`，再独立使用 `send()` 或 `steer()` 提交直接消息。
+`SendOptions` 只包含 `target` 和 `wakeup`。拥有上下文的调用方通过 `inject()` 交付带标识且冻结的 `UserMessage`，再独立使用 `send()` 或 `steer()` 提交直接消息。

 提示词和工具扩展点仍可返回 `additionalContexts`。这些值是扩展点的输出，而不是从调用方收件箱条目捕获的附件。提示词准入在 `run()` 打开轮次之前执行。获准的提示词及其返回的额外上下文会作为独立消息进入新轮次；提示词被阻止时，两者都不写入，也不打开轮次。工具产生的额外上下文则在对应工具结果之后进入 outbox。

@@ -59,7 +59,7 @@ agent API 曾用三种相互重叠的方式表示面向模型的补充输入：
 ## 验证

 - `SendOptions` 与 steering 收件箱记录不包含附加上下文；`agent/inbox/enqueue` 只报告消息及其已解析的 queued 或 steering 放置方式。
- `UserMessageData` 是提示词拦截、工具执行、hook bridge、guard 和上下文生产方共享的形状。
+- `UserMessage` 是提示词拦截、工具执行、hook bridge、guard 和上下文生产方共享的带标识且冻结的形状。
 - 公共类型、持久事件、投影和 UI 回放中均不存在 prompt-prefix 放置方式、提示词封套与 `context/message`。
 - 空闲 `inject()` 在不产生轮次或模型调用的情况下，追加一条带来源的 `user/message`。
 - 准入期间和活跃轮次中的注入会在完整工具结果批次之后的安全边界排空，并在消费它们的请求之前进入日志。
@@ -2,5 +2,5 @@
 # side as of the last confirmed-consistent state. Both languages carry equal authority;
 # after editing either side, bring the other along and re-record with:
 #   pnpm run verify-translation-pairing --write .agents/notes/implemented/architecture/2026-07-25-web-input-machine-and-slash-pipeline.md
-2026-07-25-web-input-machine-and-slash-pipeline.md: 8cf3be7b3b7579d0c37898a58fb0ab4990fd71bf
-2026-07-25-web-input-machine-and-slash-pipeline.zh.md: b1488893558d8b2bf9c104faca968435d46a9640
+2026-07-25-web-input-machine-and-slash-pipeline.md: f446f42c9e202afcb404c7a551a4f715228bb8e5
+2026-07-25-web-input-machine-and-slash-pipeline.zh.md: 8f5e449bb878811b70bc5bc29e4a09bbc1a33bfa
@@ -62,8 +62,8 @@ Calls that stay un-evented (registry registration → explicit call → await):

 A trigger/menu/pick pipeline with zero knowledge of "commands":

- The service holds only the source registry (`SlashSource{trigger: '/'|'@', name, candidates, onPick, matchSpace?, matchEnter?}`; (trigger,name) unique, registration order = group order = polling order) and `sessionOf(sctx)`. Implementing a match hook IS the declaration of participation in space/enter adjudication; the pipeline polls in registration order, the first non-undefined answer wins, and no claimant means the default sink. matchSpace is synchronous (space fires mid-keystroke; hot cache only); matchEnter is asynchronous (it may await the source's own warmup, and a warmup failure rejects).
- The controller holds the single authoritative hit (span included; retained for Space after the menu closes), the per-session menu store, the candidate-fetch generation, keyboard arbitration (combobox mode: focus stays in the textarea, ↑↓/Enter/Escape are intercepted and all pass the IME composition guard, with the single exception Shift+Enter unconditionally going first), and pick orchestration (outcome → self-dispatched bail events); at each session scope's birth it runs `warm(projection)` once over the source roster — within that scope the projection holds only the stable sessionId, with no published/capability transitions; the scope disposer tears down the controller.
+- The service holds only the source registry (`SlashSource{trigger: '/'|'@', name, order?, candidates, onPick, matchSpace?, matchEnter?}`; (trigger,name) unique; the optional `order` sorts the roster — lower first, default 0, ties keep registration order — and that sorted roster is both group order and polling order) and `sessionOf(sctx)`. Implementing a match hook IS the declaration of participation in space/enter adjudication; the pipeline polls in roster order, the first non-undefined answer wins, and no claimant means the default sink. matchSpace is synchronous (space fires mid-keystroke; hot cache only); matchEnter is asynchronous (it may await the source's own warmup, and a warmup failure rejects).
+- The controller holds the single authoritative hit (span included; retained for Space after the menu closes), the per-session menu store, the candidate-fetch generation, keyboard arbitration (combobox mode: focus stays in the textarea, ↑↓/Enter/Escape are intercepted and all pass the IME composition guard, with the single exception Shift+Enter unconditionally going first), and pick orchestration (outcome → self-dispatched bail events); a `dismiss()` verb backs MenuView's injected `onDismiss` (a pointer down outside both the menu and the surrounding composer card closes the menu; MenuView also localizes group titles through the `slash.menu` locale namespace and clamps its height to the viewport space above the composer via ui-primitives' `useAnchoredMaxHeight`); at each session scope's birth it runs `warm(projection)` once over the source roster — within that scope the projection holds only the stable sessionId, with no published/capability transitions; the scope disposer tears down the controller.
 - Trigger-detection word boundaries (`user@host` and URL `/` never trigger) and the guard tiers (plain: `/` everywhere + `@` inline / claimed: `/` suppressed, `@` live / frozen: none) are the frozen pure core.

 ### hub / facade: the resident shell and the strict-session input body
@@ -62,8 +62,8 @@ occurrence 表与 chip 三投影：

 对"命令"零知识的触发/菜单/pick 管线：

- service 只有 source 注册表（`SlashSource{trigger: '/'|'@', name, candidates, onPick, matchSpace?, matchEnter?}`；(trigger,name) 唯一、注册序 = 组序 = 轮询序）与 `sessionOf(sctx)`。实现 match 钩子即参与空格/回车裁决的声明；管线按注册序轮询，首个非 undefined 应答胜出，无人认领落 default sink。matchSpace 同步（空格在击键中触发，只许热缓存）；matchEnter 异步（可 await 源自身预热，预热失败即 reject）。
- controller 持有唯一权威 hit（含 span；菜单关闭后为 Space 保留）、per-session menu store、候选 fetch generation、键盘仲裁（combobox 模式：焦点始终在 textarea，↑↓/Enter/Escape 拦截且全程过 IME composition 守卫，唯一例外 Shift+Enter 无条件先行）、pick 编排（outcome → 自派 bail 事件）；每个 session scope 出生时对 source roster 做一次 `warm(projection)`，projection 在该 scope 内只有稳定的 sessionId，无 published/能力跃迁；scope disposer 拆除 controller。
+- service 只有 source 注册表（`SlashSource{trigger: '/'|'@', name, order?, candidates, onPick, matchSpace?, matchEnter?}`；(trigger,name) 唯一；可选 `order` 对 roster 排序——越小越靠前、默认 0、同值保持注册序——排序后的 roster 同时是组序与轮询序）与 `sessionOf(sctx)`。实现 match 钩子即参与空格/回车裁决的声明；管线按 roster 序轮询，首个非 undefined 应答胜出，无人认领落 default sink。matchSpace 同步（空格在击键中触发，只许热缓存）；matchEnter 异步（可 await 源自身预热，预热失败即 reject）。
+- controller 持有唯一权威 hit（含 span；菜单关闭后为 Space 保留）、per-session menu store、候选 fetch generation、键盘仲裁（combobox 模式：焦点始终在 textarea，↑↓/Enter/Escape 拦截且全程过 IME composition 守卫，唯一例外 Shift+Enter 无条件先行）、pick 编排（outcome → 自派 bail 事件）；`dismiss()` 动词支撑 MenuView 注入的 `onDismiss`（指针落在菜单与所在 composer 卡片之外即关闭菜单；MenuView 还经 `slash.menu` locale 命名空间本地化组标题，并经 ui-primitives 的 `useAnchoredMaxHeight` 把高度收敛到 composer 上方的视口空间）；每个 session scope 出生时对 source roster 做一次 `warm(projection)`，projection 在该 scope 内只有稳定的 sessionId，无 published/能力跃迁；scope disposer 拆除 controller。
 - 触发检测词边界（`user@host`、URL `/` 永不触发）、守卫分档（plain：`/` 到处 + `@` 行内 / claimed：`/` 抑制、`@` 活 / frozen：全无）为冻结纯核。

 ### hub / facade：常驻外壳与严格 session 输入体
@@ -1,6 +1,6 @@
 # Bilingual-pair consistency record (docs/i18n/README.md): the git blob hash of each
 # side as of the last confirmed-consistent state. Both languages carry equal authority;
 # after editing either side, bring the other along and re-record with:
-#   pnpm run verify-translation-pairing --write
-2026-07-26-subprocess-consumer-migration.md: b31f69f1251a7219e168ed7800ba16ff7d6b328c
-2026-07-26-subprocess-consumer-migration.zh.md: 47e7519a72f5482f255d15d87b483e9295f6cd2c
+#   pnpm run verify-translation-pairing --write .agents/notes/implemented/architecture/2026-07-26-subprocess-consumer-migration.md
+2026-07-26-subprocess-consumer-migration.md: 477dffc2271db08b8986b34645067b247ad7ace7
+2026-07-26-subprocess-consumer-migration.zh.md: 5e1035872c6e2101e41d2801cccdfb5ef2c088fc
@@ -6,7 +6,7 @@ English | [中文](2026-07-26-subprocess-consumer-migration.zh.md)

 ## Problem

-The [subprocess seam](2026-07-26-subprocess-seam.md) shipped shaped for exactly one consumer family: batch-collected stdout/stderr, batch stdin, a single escalating `kill()`. That was deliberate scope control, and its own note records "migrate the other spawn sites" as rejected-for-now. Review on the introducing PR reversed that deferral: the stacked follow-up should reshape the interface toward Node's API and move the remaining process-running places onto the service. The remaining spawners each carried a private copy of some slice of the same mechanics — lsp-local had its own detached-tree signalling (POSIX group + Windows taskkill + liveness polling), subagent-subprocess had the dispose ladder and its own scrub, mcp-client and pty-local and the SDK helper each had a third/fourth/fifth copy of the credential scrub — and none of it was swappable or centrally testable.
+The [subprocess seam](2026-07-26-subprocess-seam.md) shipped shaped for exactly one consumer family: batch-collected stdout/stderr, batch stdin, a single escalating `kill()`. That was deliberate scope control, and its own note records "migrate the other spawn sites" as rejected-for-now. Review on the introducing PR reversed that deferral: the stacked follow-up should reshape the interface toward Node's API and move the remaining process-running places onto the service. The remaining spawners each carried a private copy of some slice of the same mechanics — lsp-local had its own detached-tree signalling (POSIX group + Windows taskkill + liveness polling), subagent-subprocess had the dispose ladder and its own scrub, and mcp-client, pty-local, the SDK helper, and the TUI Git probe each carried another credential scrub — and none of it was swappable or centrally testable.

 ## Decision

@@ -15,7 +15,7 @@ The seam's vocabulary is now Node-shaped, and every spawner that can ride the se
 - **Per-stream stdio dispositions** on `SubprocessSpawnSpec`: `'pipe'` (the raw `Readable`/`Writable`, for consumer-owned protocol framing), `'inherit'` (diagnostics to the parent's stream), and collect mode `{ maxBytes, spill? }` — the original bounded tail-keep shape, with the spill file now optional so a diagnostic tail (a language server's stderr) buffers without touching disk. stdin is `'ignore'`, `'pipe'`, or `{ data }` (write-and-close batch).
 - **`SubprocessOutcome` carries exit facts only** (Node's close-event vocabulary); collected output stays readable through `handle.collected` after settlement (spill fds seal at the settle boundary), so batch and streaming callers share one access path and nothing is copied into the outcome.
 - **Tree-scoped termination behind one verb**: `terminate()` owns the SIGTERM→grace→SIGKILL escalation (serves the spec's abort signal too, and is a no-op once the tree is gone) — the handle exposes no single-signal `kill(signal?)`, so a consumer cannot skip the grace window; `waitForExit()` polls tree liveness (POSIX group probe; direct-child boundary on Windows); Windows tree termination (`taskkill /T`, injectable) moved in from lsp-local, so tree semantics are platform-correct for every consumer. (The stdin-EOF-first dispose ladder initially absorbed from `subagent-subprocess` later moved back out to its one consumer — see the [ladder-ownership Agent Note](2026-07-27-dispose-ladder-to-consumer.md).)
- **One scrub definition**: `scrubbedParentEnv()`/`SENSITIVE_ENV_PATTERN` live on the seam. Spawners that cannot route the spawn itself through the service — pty-local (node-pty owns the fork) and mcp-client (the MCP SDK owns the transport spawn) — import the function, so environment policy is single-sourced even where process ownership is not; the SDK helper's `scrubEnvironment()` defaults through it as well.
+- **One scrub definition**: `scrubbedParentEnv()`/`SENSITIVE_ENV_PATTERN` live on the seam, and credential-shaped names contain `KEY`, `PASSWORD`, `SECRET`, or `TOKEN`, case-insensitively. Spawners that cannot route the spawn itself through the service — pty-local (node-pty owns the fork), mcp-client (the MCP SDK owns the transport spawn), and the TUI's synchronous Git branch probe — import the function, so environment policy is single-sourced even where process ownership is not. The SDK helper's `scrubEnvironment()` defaults through the function and applies the exported pattern when its caller supplies an explicit environment; the pattern remains public for this production consumer.

 Migrations landed with the reshape: **bash-local/bash-sandbox** (collect modes + batch stdin; the bash `kill()` maps to `terminate()` so `task_kill` keeps escalation semantics), **lsp-local** (piped protocol streams + a no-spill collected stderr tail; `LspConnection` takes the seam's spawn function; its private tree-op helpers deleted), **subagent-acp** (piped ndjson streams + inherited stderr; spawn failure surfaces through `done` rejection into the same startup race; disposal is the backend-owned `disposeAcpChild` ladder over the seam's verbs, with the plugin's configured graces). **`dsh-subagent-subprocess` is deleted** — the dispose ladder and scrub are the seam's; the unused isolated-config-dir helper died with it (no consumer existed).

@@ -23,16 +23,16 @@ Compositions mounting lsp-local or subagent-acp now load `dsh-subprocess-local`

 ## Alternatives considered

-**Keep the batch-only seam and let stream consumers stay bespoke.** The introducing note's position, rejected by review: it leaves three private copies of tree signalling and five of the scrub, and any future runner (containerized executor, remote process host) would have to pick which private copy to fork. The Node-shaped dispositions cover all three observed stream shapes without widening the outcome type or buffering piped streams.
+**Keep the batch-only seam and let stream consumers stay bespoke.** The introducing note's position, rejected by review: it leaves three private copies of tree signalling and six of the scrub, and any future runner (containerized executor, remote process host) would have to pick which private copy to fork. The Node-shaped dispositions cover all three observed stream shapes without widening the outcome type or buffering piped streams.

 **A single `stdio: 'pipe' | 'inherit' | 'collect'` mode for all three streams at once.** Rejected: real consumers mix modes per stream (lsp: pipe/pipe/collect; acp: pipe/pipe/inherit; bash: data/collect/collect). Per-stream dispositions are exactly Node's shape and avoid a second spawn call for the mixed cases.

 **Migrate pty-local and mcp-client spawns too.** Rejected on ownership grounds, not scope: node-pty's `fork()` allocates the terminal itself, and the MCP SDK's `StdioClientTransport` spawns internally — neither call site is ours to route. They adopt the shared scrub (the part that is policy), and their READMEs say why the spawn stays put.

-**Migrate the test-support launchers (acp-snapshot, loader-smoke) and the SDK package-manager runner.** Rejected: the support packages are deliberately dependency-light test infrastructure that must not depend on product seams, and the SDK wizard's `stdio: 'inherit'`-with-redirect semantics plus its out-of-composition lifecycle (no cordis context at all) make the service a poor fit; it shares the scrub instead.
+**Migrate the test-support launchers (acp-snapshot, loader-smoke), the SDK package-manager runner, and the TUI Git probe.** Rejected: the support packages are deliberately dependency-light test infrastructure that must not depend on product seams; the SDK wizard's `stdio: 'inherit'`-with-redirect semantics plus its out-of-composition lifecycle (no cordis context at all) make the service a poor fit; and the TUI probe is synchronous. The production callers share the scrub instead.

 ## Consequences

 Bought: one implementation of tree signalling, escalation, bounded collection, and the scrub, tested once in `dsh-subprocess-local`'s suites (including injected-platform Windows coverage that lsp-local's private copy never had); lsp-local and subagent-acp shed their process plumbing and their children now survive plugin reloads and die with composition teardown like bash's; a whole package (`dsh-subagent-subprocess`) is gone. The seam README's "one consumer family" limitation is retired.

-Cost: the seam is wider — three stdio modes and the terminate/waitForExit/dispose lifecycle surface instead of one mode and one verb — so a future backend implements more surface; the compositions for lsp-local/subagent-acp each carry the subprocess row now; and `SubprocessOutcome` no longer carries output, a breaking shape change inside the still-unreleased stack (the PR2 layer was updated in place rather than shimmed, per the pre-release stance). pty-local/mcp-client/SDK/test-support spawns remain outside the service by ownership, with the scrub as the shared floor.
+Cost: the seam is wider — three stdio modes and the terminate/waitForExit/dispose lifecycle surface instead of one mode and one verb — so a future backend implements more surface; the compositions for lsp-local/subagent-acp each carry the subprocess row now; and `SubprocessOutcome` no longer carries output, a breaking shape change inside the still-unreleased stack (the PR2 layer was updated in place rather than shimmed, per the pre-release stance). pty-local/mcp-client/SDK/TUI/test-support spawns remain outside the service by ownership or execution shape, with the scrub as the shared floor and its pattern intentionally exported for explicit-environment consumers.
@@ -6,7 +6,7 @@ Status: implemented

 ## 问题

-[进程 seam](2026-07-26-subprocess-seam.md) 交付时恰好只为一个消费方家族塑形：批量收集的 stdout/stderr、批量 stdin、单一的升级式 `kill()`。那是有意的范围控制，其自身的 Agent Note 也把「迁移其余 spawn 调用点」记为暂缓否决项。引入该 seam 的 PR（Pull Request）上的评审推翻了这一暂缓决定：堆叠其上的后续变更应当把接口向 Node 的 API 方向重塑，并把其余运行进程之处迁到该服务上。其余各 spawn 调用点此前各自持有同一套机制中某个切片的私有副本——lsp-local 自带 detached 进程树信号发送（POSIX 进程组 + Windows taskkill + 存活轮询），subagent-subprocess 自带 dispose（资源释放）阶梯和自己的凭据清除，mcp-client、pty-local 与 SDK helper 则各自持有凭据清除的第三、第四、第五份副本——而这一切既不可替换，也无法集中测试。
+[进程 seam](2026-07-26-subprocess-seam.md) 交付时恰好只为一个消费方家族塑形：批量收集的 stdout/stderr、批量 stdin、单一的升级式 `kill()`。那是有意的范围控制，其自身的 Agent Note 也把「迁移其余 spawn 调用点」记为暂缓否决项。引入该 seam 的 PR（Pull Request）上的评审推翻了这一暂缓决定：堆叠其上的后续变更应当把接口向 Node 的 API 方向重塑，并把其余运行进程之处迁到该服务上。其余各 spawn 调用点此前各自持有同一套机制中某个切片的私有副本——lsp-local 自带 detached 进程树信号发送（POSIX 进程组 + Windows taskkill + 存活轮询），subagent-subprocess 自带 dispose（资源释放）阶梯和自己的凭据清除，mcp-client、pty-local、SDK helper 与 TUI Git 探测则各自持有另一份凭据清除——而这一切既不可替换，也无法集中测试。

 ## 决策

@@ -15,7 +15,7 @@ Status: implemented
 - **按流划分的 stdio 处置方式（disposition）**，位于 `SubprocessSpawnSpec` 上：`'pipe'`（原始的 `Readable`/`Writable`，供消费方自有的协议分帧使用）、`'inherit'`（诊断输出直通父进程的流），以及收集模式（collect）`{ maxBytes, spill? }`——即最初的有界尾部保留形状，只是 spill 文件改为可选，使诊断尾部（例如语言服务器的 stderr）无需落盘即可缓冲。stdin 则为 `'ignore'`、`'pipe'` 或 `{ data }`（写完即关闭的批量形式）。
 - **`SubprocessOutcome` 只承载退出事实**（Node close 事件的词汇）；收集到的输出在结算后仍可经 `handle.collected` 读取（spill 文件描述符在结算边界封存），因此批量与流式调用方共用一条访问路径，也没有任何内容被复制进这份结果。
 - **以进程树为范围的终止，集中在一个动词后面**：`terminate()` 拥有 SIGTERM→宽限期→SIGKILL 升级（也承接 spec 的 abort 信号，进程树消亡后为空操作）——句柄不暴露单信号的 `kill(signal?)`，因此消费方无法跳过宽限窗口；`waitForExit()` 轮询进程树存活状态（POSIX 进程组探测；Windows 上以直接子进程为界）。Windows 进程树终止（`taskkill /T`，可注入）自 lsp-local 迁入，因此每个消费方拿到的进程树语义在各平台上都正确。（最初从 `subagent-subprocess` 吸收的以 stdin EOF 打头的 dispose 阶梯，后来又移回其唯一消费方——见[阶梯归属 Agent Note](2026-07-27-dispose-ladder-to-consumer.md)。）
- **凭据清除只有一份定义**：`scrubbedParentEnv()`/`SENSITIVE_ENV_PATTERN` 定义在 seam 上。无法把 spawn 本身路由到该服务的调用点——pty-local（node-pty 拥有 fork）与 mcp-client（MCP SDK 拥有传输层的 spawn）——改为导入该函数，因此即便进程所有权无法统一，环境策略仍是单一来源；SDK helper 的 `scrubEnvironment()` 默认同样委托给它。
+- **凭据清除只有一份定义**：`scrubbedParentEnv()`/`SENSITIVE_ENV_PATTERN` 定义在 seam 上，且凭据形状的名称不区分大小写地包含 `KEY`、`PASSWORD`、`SECRET` 或 `TOKEN`。无法把 spawn 本身路由到该服务的调用点——pty-local（node-pty 拥有 fork）、mcp-client（MCP SDK 拥有传输层的 spawn）与 TUI 的同步 Git 分支探测——改为导入该函数，因此即便进程所有权无法统一，环境策略仍是单一来源。SDK helper 的 `scrubEnvironment()` 默认委托给该函数，并在调用方显式传入环境时应用导出的正则；该正则因此生产消费方而保持公开。

 各项迁移随这次重塑一并落地：**bash-local/bash-sandbox**（收集模式 + 批量 stdin；bash 的 `kill()` 映射到 `terminate()`，因此 `task_kill` 保有升级语义），**lsp-local**（管道化的协议流 + 无 spill 的 stderr 收集尾部；`LspConnection` 改为接收 seam 的 spawn 函数；其私有的进程树操作辅助函数已删除），**subagent-acp**（管道化的 ndjson 流 + inherit 的 stderr；spawn 失败经 `done` 的 reject 汇入同一个启动竞态；dispose 是后端自有的 `disposeAcpChild` 阶梯，经由 seam 的动词运行，携带插件所配置的宽限期）。**`dsh-subagent-subprocess` 已删除**——dispose 阶梯与凭据清除归 seam 所有；无人使用的隔离配置目录辅助函数随之消亡（其消费方本就不存在）。

@@ -23,16 +23,16 @@ Status: implemented

 ## 曾考虑的替代方案

-**保持只支持批量的 seam，让流式消费方继续各自为政。**这正是引入该 seam 的 Agent Note 当初的立场，评审将其否决：这样会留下三份进程树信号发送的私有副本和五份凭据清除的私有副本，而未来任何运行器（容器化执行器、远程进程宿主）都得挑选去 fork 哪一份私有副本。Node 形状的处置方式覆盖已观察到的全部三种流形状，既不拓宽结果类型，也不缓冲管道化的流。
+**保持只支持批量的 seam，让流式消费方继续各自为政。**这正是引入该 seam 的 Agent Note 当初的立场，评审将其否决：这样会留下三份进程树信号发送的私有副本和六份凭据清除的私有副本，而未来任何运行器（容器化执行器、远程进程宿主）都得挑选去 fork 哪一份私有副本。Node 形状的处置方式覆盖已观察到的全部三种流形状，既不拓宽结果类型，也不缓冲管道化的流。

 **用单个 `stdio: 'pipe' | 'inherit' | 'collect'` 模式一次性统辖全部三条流。**否决：真实消费方按流混用模式（lsp：pipe/pipe/collect；acp：pipe/pipe/inherit；bash：data/collect/collect）。按流划分的处置方式恰好就是 Node 的形状，也免去了混用场景的第二个 spawn 调用。

 **把 pty-local 与 mcp-client 的 spawn 也一并迁移。**基于所有权而非范围否决：node-pty 的 `fork()` 自行分配终端，MCP SDK 的 `StdioClientTransport` 在内部完成 spawn——这两处调用点都不归我们路由。它们采纳共享的凭据清除（那正是属于策略的部分），并在各自的 README 中说明 spawn 为何留在原地。

-**迁移 test-support 启动器（acp-snapshot、loader-smoke）与 SDK package-manager 运行器。**否决：support 各包（package）是刻意保持轻依赖的测试基础设施，不得依赖产品 seam；而 SDK 向导那套附带重定向的 `stdio: 'inherit'` 语义，加上其完全脱离组合的生命周期（根本没有 cordis 上下文），使该服务并不合用；它改为共享凭据清除。
+**迁移 test-support 启动器（acp-snapshot、loader-smoke）、SDK package-manager 运行器与 TUI Git 探测。**否决：support 各包（package）是刻意保持轻依赖的测试基础设施，不得依赖产品 seam；SDK 向导那套附带重定向的 `stdio: 'inherit'` 语义，加上其完全脱离组合的生命周期（根本没有 cordis 上下文），使该服务并不合用；TUI 探测则是同步调用。这些生产调用点改为共享凭据清除。

 ## 后果

 换来的是：进程树信号发送、升级、有界收集与凭据清除各自只剩一份实现，且只在 `dsh-subprocess-local` 的测试套件中测试一次（其中包括 lsp-local 的私有副本从未有过的、以注入平台方式实现的 Windows 覆盖）；lsp-local 与 subagent-acp 卸下了自己的进程管道，其子进程如今像 bash 的一样，在插件重载后存活、随组合拆除而终止；一个完整的包（`dsh-subagent-subprocess`）就此消失。seam README 中「只有一个消费方家族」的限制说明也随之退役。

-代价是：这道 seam 变宽了（stdio 模式从一种变为三种、终止动词换成 terminate/waitForExit/dispose 这组生命周期表面），未来的后端因此要实现更宽的表面；lsp-local/subagent-acp 的各组合如今都多出 subprocess 这一行组合配置；`SubprocessOutcome` 也不再承载输出，这是仍未发布的堆叠变更内部的一次破坏性形状变更（依照预发布立场，PR2 那一层被就地更新，而非加 shim）。pty-local/mcp-client/SDK/test-support 的 spawn 因所有权归属留在该服务之外，以凭据清除作为共享底线。
+代价是：这道 seam 变宽了（stdio 模式从一种变为三种、终止动词换成 terminate/waitForExit/dispose 这组生命周期表面），未来的后端因此要实现更宽的表面；lsp-local/subagent-acp 的各组合如今都多出 subprocess 这一行组合配置；`SubprocessOutcome` 也不再承载输出，这是仍未发布的堆叠变更内部的一次破坏性形状变更（依照预发布立场，PR2 那一层被就地更新，而非加 shim）。pty-local/mcp-client/SDK/TUI/test-support 的 spawn 因所有权归属或执行形状留在该服务之外，以凭据清除作为共底线，且为显式环境的生产消费方有意保持该正则导出。
@@ -0,0 +1,6 @@
+# Bilingual-pair consistency record (docs/i18n/README.md): the git blob hash of each
+# side as of the last confirmed-consistent state. Both languages carry equal authority;
+# after editing either side, bring the other along and re-record with:
+#   pnpm run verify-translation-pairing --write .agents/notes/implemented/architecture/2026-07-28-api-browser-trust-boundary.md
+2026-07-28-api-browser-trust-boundary.md: e56d0fc2a7bd551899605491f3a0522b62b961b0
+2026-07-28-api-browser-trust-boundary.zh.md: 2958f7e49bfd4a258c63fc96c2e8aee0f98183ee
@@ -0,0 +1,31 @@
+# Agent Note: One carrier-level browser-trust boundary for the whole /api surface
+
+Status: implemented
+
+English | [中文](2026-07-28-api-browser-trust-boundary.zh.md)
+
+## Problem
+
+The web GUI host serves `/api` over plain HTTP (default `127.0.0.1:3080`, `--host 0.0.0.0` supported), and the surface includes remote-code-execution-grade methods — `session.prompt` drives an agent that runs bash. A browser turns the operator into a confused deputy against such a local API in two classic ways: a malicious page fires a "simple" cross-site POST (`text/plain` — sent without a CORS preflight) whose side effects execute even though the response stays unreadable, and a DNS-rebound origin talks to the socket as if same-origin, making CORS inapplicable entirely, with only the `Host` header betraying the attacker's domain. Before this decision the system's only browser-trust check (`isTrustedNativeDialogRequest`: loopback socket + same-origin + loopback Host) guarded exactly one cosmetic route — `host.pickDirectory`, whose native dialog pops on the host's screen — while every consequential method was unguarded. Guarding per-RPC also could not survive the upcoming in-app directory browser, whose whole point is serving legitimately remote clients that a loopback rule would refuse.
+
+## Decision
+
+Enforce browser trust once, at the carrier, for the entire `/api` prefix — two halves in two stacked PRs:
+
+- **Media-type fence (dsh-host-apiproxy)**: every `/api` POST must declare `application/json`, else 415 before parsing. Cross-site "simple" requests thereby stop existing: any cross-site attempt is forced into a CORS preflight this server never answers.
+- **Authority fence (dsh-client-connection, `src/api-request-trust.ts`)**: every request must present a `Host` that is loopback or matches a `trustedHosts` entry (exact on `host:port`, any port on port-less entries, WHATWG-normalized; rebinding defense). Deliberately no shortcut for unmarked requests: over plain HTTP a browser attaches neither `Origin` nor Fetch-Metadata to reads (EventSource, images, navigations — those headers go only to trustworthy destinations), so an unmarked request may be a rebound browser read whose response the page can read, and Host is the one header rebinding cannot forge; non-browser clients pass via loopback, the derived LAN IP literals, or a declared authority. An attached `Origin` must equal the Host authority; `sec-fetch-site: cross-site` is refused outright. A `trustedHosts` entry that is not a bare, canonical authority fails the plugin load — WHATWG parsing would otherwise quietly authorize the hostname inside a typo or broaden an exact-port grant. `host.pickDirectory` loses its bespoke guard and rides the same fence.
+
+Two boundaries stay deliberately out of scope: reachability is the webserver binding's policy (`host: 127.0.0.1 | 0.0.0.0`), and authentication for genuinely remote deployments is deferred work recorded in the connection README — the fence is a confused-deputy defense, not an auth layer. The old guard's loopback-socket check was dropped rather than generalized: with binding expressing reachability and `trustedHosts` naming remote authorities, the socket address adds nothing a header fence does not already cover.
+
+## Alternatives considered
+
+- **Per-RPC guards (status quo extended).** Rejected: the guard list trails the method list forever, the highest-value methods were already unguarded, and a loopback rule on browse RPCs would break the remote deployments they exist for.
+- **CORS headers + credential omission.** Rejected: we never want cross-origin reads at all, so answering preflights only widens the surface; refusing them is strictly stronger and simpler.
+- **Auth tokens now.** Rejected for this change: token minting/storage/rotation is real product surface; the fence closes the browser-deputy holes today without pre-deciding the auth design.
+
+## Consequences
+
+- Any future `/api` method is covered by construction; there is no per-route trust decision left to forget.
+- Non-loopback deployments must have their serving authorities trusted or requests are refused. The dsh CLI keeps its advertised `--host 0.0.0.0` LAN URL working by deriving the machine's LAN IP literals into the connection row (port-less entries — an IP-literal Host cannot be a rebound name, and the bound port may be OS-assigned) and offers `dsh web --trusted-host` for named authorities; compositions the CLI does not boot declare `trustedHosts` themselves. Non-browser automation rides the same fence: loopback, a derived LAN IP, or a declared authority passes; an undeclared DNS alias is refused.
+- Clients must label POST bodies `application/json` (ours always did; raw-fetch tests gained the header).
+- The trusted-network assumption of an unauthenticated `0.0.0.0` deployment is now documented instead of implicit.
@@ -0,0 +1,31 @@
+# Agent Note：整个 /api 面共用一道载体级浏览器信任边界
+
+状态：已实现
+
+[English](2026-07-28-api-browser-trust-boundary.md) | 中文
+
+## 问题
+
+Web GUI 宿主以纯 HTTP 提供 `/api`（默认 `127.0.0.1:3080`，支持 `--host 0.0.0.0`），而这个面上有远程代码执行级别的方法——`session.prompt` 驱动的 agent 可以运行 bash。浏览器会用两种经典方式把操作者变成攻击此类本地 API 的"混淆代理人"：恶意页面发出跨站"简单请求" POST（`text/plain`——不经 CORS 预检即发出），其副作用照常执行、只是响应不可读；以及 DNS rebinding 后的源以"同源"身份直连 socket，CORS 整体失效，只有 `Host` 头会暴露攻击者的域名。在本决策之前，系统里唯一的浏览器信任检查（`isTrustedNativeDialogRequest`：回环 socket + 同源 + 回环 Host）只守着一个装饰性的路由——`host.pickDirectory`，其原生对话框弹在宿主屏幕上——而所有真正要命的方法都在裸奔。按 RPC 逐个设防也活不过即将到来的应用内目录浏览器：它存在的意义就是服务合法的远程客户端，回环规则恰恰会拒绝它们。
+
+## 决策
+
+在载体层对整个 `/api` 前缀一次性执行浏览器信任检查——两半各占一个栈式 PR：
+
+- **媒体类型栅栏（dsh-host-apiproxy）**：每个 `/api` POST 必须声明 `application/json`，否则在解析前以 415 拒绝。跨站"简单请求"由此不复存在：任何跨站尝试都被逼进一次本服务器从不应答的 CORS 预检。
+- **权威栅栏（dsh-client-connection，`src/api-request-trust.ts`）**：每个请求的 `Host` 都必须是回环地址，或与某个 `trustedHosts` 条目匹配（带端口的 `host:port` 条目精确匹配，不带端口的条目匹配任意端口，均经 WHATWG 归一化；rebinding 防御）。刻意不为无标记请求开捷径：明文 HTTP 下浏览器的读取（EventSource、图片、导航——这些头只发给可信目标）既不带 `Origin` 也不带 Fetch-Metadata，因此无标记请求可能是被重绑页面发起且响应可被读走的读取，而 Host 是重绑唯一伪造不了的请求头；非浏览器客户端经由回环地址、推导的 LAN IP 字面量或已声明的权威通过。若带 `Origin` 则必须与 Host 权威完全一致；`sec-fetch-site: cross-site` 一律拒绝。不是纯的、规范形权威的 `trustedHosts` 条目会让插件加载失败——否则 WHATWG 解析会悄悄授权笔误里的 hostname，或放大精确端口授权。`host.pickDirectory` 失去专属守卫，与其他请求同栅而行。
+
+两条边界刻意留在范围之外：可达性归 webserver 绑定配置（`host: 127.0.0.1 | 0.0.0.0`）管辖；真正远程部署的认证是延期工作，记录在 connection README——这道栅栏是混淆代理人防御，不是认证层。旧守卫的回环 socket 检查被放弃而非泛化：绑定表达可达性、`trustedHosts` 点名远程权威之后，socket 地址提供不了头部栅栏覆盖不到的任何东西。
+
+## 曾考虑的替代方案
+
+- **按 RPC 设防（延续现状）。** 否决：守卫清单永远追着方法清单跑，价值最高的方法本来就没被守住，而 browse RPC 上的回环规则会破坏它们为之存在的远程部署。
+- **CORS 头 + 省略凭据。** 否决：我们根本不想要任何跨源读取，应答预检只会扩大暴露面；拒绝预检严格更强也更简单。
+- **现在就上认证令牌。** 在本变更中否决：令牌的签发/存储/轮换是真实的产品面；栅栏今天就能封死浏览器代理人漏洞，无需预先决定认证设计。
+
+## 后果
+
+- 未来任何 `/api` 方法天然在覆盖范围内；不存在会被遗忘的按路由信任决定。
+- 非回环部署的服务权威必须获得信任，否则请求会被拒绝。dsh CLI 通过把本机 LAN IP 字面量推导进 connection 行（不带端口的条目——IP 字面量 Host 不可能是被重绑的域名，且绑定端口可能由操作系统分配）来保住它广告出的 `--host 0.0.0.0` LAN URL，并提供 `dsh web --trusted-host` 声明具名权威；CLI 不参与引导的组合自行声明 `trustedHosts`。非浏览器自动化走同一道栅栏：回环地址、推导的 LAN IP 或已声明的权威可通过；未声明的 DNS 别名会被拒绝。
+- 客户端必须给 POST 体标注 `application/json`（我们自己的客户端一向如此；裸 fetch 测试补上了该头）。
+- 无认证 `0.0.0.0` 部署的"信任网络"假设从隐含变为成文。
@@ -0,0 +1,6 @@
+# Bilingual-pair consistency record (docs/i18n/README.md): the git blob hash of each
+# side as of the last confirmed-consistent state. Both languages carry equal authority;
+# after editing either side, bring the other along and re-record with:
+#   pnpm run verify-translation-pairing --write .agents/notes/implemented/architecture/2026-07-28-consolidated-tui-presentation.md
+2026-07-28-consolidated-tui-presentation.md: f87d543a698d6e77abf9120c6579100df4b60b64
+2026-07-28-consolidated-tui-presentation.zh.md: 005e408f0e75207027315546942f9eab57d595d1
@@ -0,0 +1,63 @@
+# Agent Note: Consolidated TUI presentation and navigation
+
+Status: implemented
+
+English | [中文](2026-07-28-consolidated-tui-presentation.zh.md)
+
+## Problem
+
+The terminal UI accumulated independent presentation rules that interacted poorly: palette roles aliased one another or inverted emphasis on light terminals; tool-card framing, output, and exit markers repeated or competed; injected context was parsed as XML and could not fold reliably; and `/resume` excluded sessions outside the current workspace even when the launcher could reach them. Each symptom appeared local, but the durable decision is one terminal-reading model: a small inspectable palette, status-led cards with recessed bodies, content-independent transcript folding, and workspace-aware navigation.
+
+## Decision
+
+### Palette
+
+`paletteSpec(scheme)` is the single table of SGR codes, close codes, and purposes. `createPalette` derives every wrapper from it and `/palette` prints the same table in the running terminal. Components do not emit their own SGR sequences except for the fixed startup brand gradient. Every close resets every SGR group its open sets.
+
+Duplicate roles are merged: `muted` into `dim`, `added` into `success`, `removed` into `error`, and the unused second accent is removed. `dim` uses `2;39` and closes with `22;39` on both schemes so recessed text stays relative to the terminal foreground rather than becoming a fixed heavy gray on light backgrounds. Colors and attributes are branded separately in TypeScript, allowing attribute/color composition while rejecting nested colors whose reset would discard the outer color.
+
+### Tool cards
+
+A tool card has one colored `Tool / <name>` status header over one dim body. Presenter titles, terminal commands and cwd rows, output, XML text, and fold markers use that body tone. Diff colors remain because red and green carry meaning, and signal markers remain errors.
+
+`renderUnknownXml` receives an explicit body styler for unknown tool results. Terminal presenters parse and remove the model-facing final exit or signal marker before returning `TerminalResultView.output`; the TUI renders the structured status once as its own pill. Truncation, timeout, and sandbox lines remain in the body because the pill does not represent them.
+
+### Injected context and folding
+
+Injected context renders as prose in `ContextCardComponent`, not through the XML tree renderer. Exact matched outer `<system-reminder>` lines are stripped, but mismatched, unpaired, or inline tag-like text remains verbatim. Model-facing content is unchanged. Folding uses the shared `preview` helper after body assembly, so it depends only on row count, never parser success or payload characters.
+
+`Ctrl+O` cycles collapsed, expanded, and hidden. Tool cards disappear in the hidden state together with their card-owned leading gap. Context cards participate in collapsed and expanded states but fall back to collapsed while tools are hidden, because injected instructions are not disposable tool traffic.
+
+### Cross-workspace resume
+
+The resume picker summarizes all records and owns a current-workspace/all-workspaces scope toggled with Tab. It defaults to the current workspace, adds workspace labels only in the broader scope, and refuses records without a cwd because there is no directory to enter.
+
+`TuiResumeHost.handoff` receives the selected `SessionId` and the cwd re-read during preflight. The CLI changes directory before disposing the current app, so an unreachable directory fails while the terminal can still recover; `execve` then inherits the selected workspace. The launcher also supplies the exit message rather than asking the TUI to reconstruct launcher syntax.
+
+## Alternatives considered
+
+**Keep separate notes and local fixes for each visual symptom.** Rejected: the decisions share one reading hierarchy and repeatedly superseded each other. One owner makes the final palette, card, context, and navigation rules clear without requiring readers to reconstruct chronology.
+
+**Keep aliases and enforce presentation by convention.** Rejected: aliases imply distinctions that do not exist, and nested color resets or incomplete SGR closes fail silently. A single table plus types makes the contract inspectable and mechanically checked.
+
+**Retain framing/output color splits inside tool cards.** Rejected: real cards mixed default foreground, cyan commands, dim cwd, unstyled XML, and dim output. The status header already provides the scan anchor; one recessed body removes noise. Diff colors are the narrow semantic exception.
+
+**Parse or repair injected context as XML.** Rejected: reminder frames are prompting conventions around arbitrary prose containing raw ampersands, comparisons, and placeholder angle brackets. Repairing or escaping it would either guess structure or alter model-visible text.
+
+**Hide context cards with tool cards.** Rejected: context carries injected instructions, not recoverable execution detail. The hidden phase therefore removes only tool traffic.
+
+**Keep resume restricted to one workspace or infer cwd after boot.** Rejected: the restriction forces manual relaunch, while restored header cwd does not control filesystem and shell resolution. The target directory must cross the host seam before process replacement.
+
+**Drop the TUI exit pill or remove model-facing exit markers.** Rejected: the pill is the scannable UI status, while the text marker is the model's status signal. The presenter consumes the marker when constructing the structured view so both audiences receive one representation.
+
+## Consequences
+
+The transcript reads as colored status headers over recessed detail, context presentation is stable for arbitrary prose, and one shortcut controls transcript density. The public `TuiTheme.muted` role is removed; extensions use `dim`. The palette and `renderUnknownXml` contracts are stricter, adding small compile-time friction in exchange for preventing silent style loss.
+
+Cross-workspace resume can move every path-resolving tool to another directory. A missing or inaccessible cwd prevents handoff. The broader picker also makes concurrent access to a shared session store easier to reach; cross-process session locking remains separate work.
+
+The terminal presenter still treats a final output line exactly matching its exit-marker grammar as structured status, so a command that intentionally prints such a line can lose it from the card body. This residual is documented by `dsh-tool-bash`.
+
+## Testing
+
+TUI unit and keyless terminal snapshots cover palette enumeration, light/dark roles, legal and illegal style composition, uniformly dim card bodies, semantic diff colors, marker-free terminal output with one exit pill, prose-preserving context frames, content-independent folding, the three-state Ctrl+O cycle, model filtering, and both resume scopes. CLI handoff tests cover passing the re-read cwd and rejecting directory-entry failure before teardown. Tool-bash tests pin result-marker emission, parse, and stripping as one round trip.
@@ -0,0 +1,63 @@
+# Agent Note: 统一的 TUI 呈现与导航
+
+Status: implemented
+
+[English](2026-07-28-consolidated-tui-presentation.md) | 中文
+
+## Problem
+
+终端 UI 逐步积累了多套彼此干扰的呈现规则：调色板角色互为别名，或在浅色终端中颠倒强调层级；工具卡片的框架、输出和退出标记重复或争夺注意力；注入上下文被当作 XML 解析，无法可靠折叠；`/resume` 即使能通过启动器访问其他工作区，也会排除不属于当前工作区的会话。每个症状看似局部，但持久决策只有一个终端阅读模型：精简且可检查的调色板、以状态为首且正文内收的卡片、与内容无关的记录折叠，以及感知工作区的导航。
+
+## Decision
+
+### 调色板
+
+`paletteSpec(scheme)` 是 SGR 开始码、结束码和用途的唯一表。`createPalette` 从该表派生所有包装器，`/palette` 在运行中的终端打印同一张表。除固定的启动品牌渐变外，组件不自行发出 SGR 序列。每个结束码都重置对应开始码设置的所有 SGR 组。
+
+重复角色被合并：`muted` 并入 `dim`，`added` 并入 `success`，`removed` 并入 `error`，未使用的第二强调色被移除。`dim` 在两种配色方案中都使用 `2;39`，并以 `22;39` 结束，使内收文本相对于终端前景色变暗，而不会在浅色背景上变成固定的深灰色。TypeScript 分别标记颜色和属性，允许属性与颜色组合，同时拒绝会因重置而丢失外层颜色的嵌套颜色。
+
+### 工具卡片
+
+工具卡片由一行带颜色的 `Tool / <name>` 状态标题和一块统一的 dim 正文组成。呈现器标题、终端命令及 cwd 行、输出、XML 文本和折叠标记都使用正文色调。差异颜色继续保留，因为红绿承载语义；信号标记也继续作为错误显示。
+
+`renderUnknownXml` 对未知工具结果显式接收正文样式器。终端呈现器在返回 `TerminalResultView.output` 前解析并移除面向模型的末尾退出或信号标记；TUI 只把结构化状态呈现一次。截断、超时和沙箱信息继续留在正文中，因为状态标记不表达这些事实。
+
+### 注入上下文与折叠
+
+注入上下文由 `ContextCardComponent` 按普通文本呈现，不经过 XML 树渲染器。仅移除精确配对的外层 `<system-reminder>` 行；不匹配、单边或正文内类似标签的文本都原样保留。面向模型的内容不变。折叠在正文组装完成后使用共享 `preview` 辅助函数，因此只取决于行数，不依赖解析是否成功或载荷包含哪些字符。
+
+`Ctrl+O` 在折叠、展开和隐藏之间循环。隐藏状态会连同卡片自有的前导间距一起移除工具卡片。上下文卡片参与折叠和展开状态，但工具隐藏时回到折叠状态，因为注入指令不是可丢弃的工具流量。
+
+### 跨工作区恢复
+
+恢复选择器汇总所有记录，并维护可用 Tab 切换的当前工作区/所有工作区范围。默认范围是当前工作区；只有更宽范围才显示工作区标签。没有 cwd 的记录会被拒绝，因为没有可进入的目录。
+
+`TuiResumeHost.handoff` 接收选中的 `SessionId` 和预检时重新读取的 cwd。CLI 在释放当前应用前切换目录，因此无法访问的目录会在终端仍可恢复时失败；随后 `execve` 继承所选工作区。退出提示也由启动器提供，而不是让 TUI 反推启动器命令语法。
+
+## Alternatives considered
+
+**为每个视觉症状保留独立 Agent Note 和局部修复。** 否决：这些决策共享同一阅读层级，而且彼此多次取代。由一份记录统一拥有最终的调色板、卡片、上下文和导航规则，读者无需重建变更顺序。
+
+**保留别名，并依靠约定执行呈现规则。** 否决：别名暗示并不存在的差异；嵌套颜色重置或不完整的 SGR 结束会静默失败。单一表格加类型约束使契约可检查且可机械验证。
+
+**保留工具卡片内部的框架/输出颜色分层。** 否决：真实卡片会混用默认前景、青色命令、dim cwd、无样式 XML 和 dim 输出。状态标题已经提供扫描锚点；统一内收正文能消除噪声。差异颜色是狭窄的语义例外。
+
+**把注入上下文继续解析或修复成 XML。** 否决：提醒框架只是包裹任意普通文本的提示约定，其中会包含原始 `&`、比较表达式和尖括号占位符。修复或转义要么猜测结构，要么改变模型可见文本。
+
+**随工具卡片一起隐藏上下文卡片。** 否决：上下文承载注入指令，不是可恢复的执行细节。因此隐藏阶段只移除工具流量。
+
+**把恢复限制在一个工作区，或在启动后推断 cwd。** 否决：前者迫使用户手动重启；后者恢复的会话头 cwd 并不控制文件系统和 shell 的路径解析。目标目录必须在进程替换前跨过主机接口。
+
+**移除 TUI 退出状态标记，或移除面向模型的退出标记。** 否决：前者是便于扫描的 UI 状态，后者是模型的状态信号。呈现器在构造结构化视图时消费文本标记，使两类受众各看到一种表示。
+
+## Consequences
+
+记录现在表现为带颜色的状态标题和内收细节；上下文对任意普通文本都稳定呈现；一个快捷键控制记录密度。公共 `TuiTheme.muted` 角色被移除，扩展改用 `dim`。调色板和 `renderUnknownXml` 契约更严格，以少量编译期摩擦换取对静默样式丢失的防护。
+
+跨工作区恢复会把所有依赖路径解析的工具移动到另一个目录。cwd 缺失或不可访问时不能交接。更宽的选择范围也使共享会话存储的并发访问更容易触达；跨进程会话锁仍是独立后续工作。
+
+终端呈现器仍会把与退出标记语法完全一致的最后一行输出视为结构化状态，因此命令有意打印这种行时，卡片正文可能丢失该行。`dsh-tool-bash` 已记录这一残余限制。
+
+## Testing
+
+TUI 单元测试和无密钥终端快照覆盖调色板枚举、浅色/深色角色、合法与非法样式组合、统一 dim 卡片正文、保留语义的差异颜色、仅有一个退出状态且正文无标记、普通文本上下文框架、与内容无关的折叠、Ctrl+O 三态循环、模型过滤和两种恢复范围。CLI 交接测试覆盖传递重新读取的 cwd，并在释放前拒绝目录切换失败。tool-bash 测试把结果标记的生成、解析和移除固定为同一轮往返契约。
@@ -0,0 +1,6 @@
+# Bilingual-pair consistency record (docs/i18n/README.md): the git blob hash of each
+# side as of the last confirmed-consistent state. Both languages carry equal authority;
+# after editing either side, bring the other along and re-record with:
+#   pnpm run verify-translation-pairing --write .agents/notes/implemented/architecture/2026-07-28-directory-picker-capability-seam.md
+2026-07-28-directory-picker-capability-seam.md: 7c8f8cb67690cb4c5858cefb52b8cd79e649ec38
+2026-07-28-directory-picker-capability-seam.zh.md: 05545fc3cd758523814b31afa705249972d86464
@@ -0,0 +1,39 @@
+# Agent Note: A capability-discriminated directory-picker seam for the web-GUI host
+
+Status: implemented
+
+English | [中文](2026-07-28-directory-picker-capability-seam.zh.md)
+
+## Problem
+
+The web GUI's "Open local folder" flow was hardwired to one interaction: `host.pickDirectory` invoked a native OS chooser compiled into `dsh-host-apiproxy` (private module, test-only injection seam). That shape cannot serve remote deployments — no OS dialog reaches a browser on another machine — and the planned in-app directory browser (Figma `Harness` 802-56979) needs listing/creation primitives, which are a different interaction contract, not a different implementation of the same one. Swapping interactions required editing gateway source, against the repo's everything-is-a-plugin stance.
+
+## Decision
+
+A three-package capability seam in `packages/host/` — `directory-picker` (interface), `directory-picker-native`, `directory-picker-browse` (backends) — with one contract method: `capability()` returns a **discriminated union**, `{ kind: 'native', pick(signal) }` or `{ kind: 'browse', list(path?), createDirectory(path, name) }`. The gateway (`dsh-host-apiproxy`) injects `directoryPicker`, serves the matching RPCs, and answers `directory-picker-unavailable` for the other kind. The union is discriminated because the backends differ in *interaction shape* — flattening them into one method set would force every backend to fake the other's shape.
+
+**The client side is slot-composed, not advertisement-branched.** ui-workspace's two trigger surfaces each declare a `single` directory-flow hole (`conversation.hero.workspace.directoryFlow` / `sidebar.workspaces.directoryFlow`; two keys because a hole has exactly one declaring slot entry — same owner contract, same occupant). Backend packages are **dual-face**: the browser half registers the matching interaction into both holes — `-native` a renderless occupant driving `host.pickDirectory`, `-browse` the in-app Select Workspace Directory dialog. The hole's owner conversation (`open`/`busy`/`onPicked`/`onCancel`/`onError`) carries the whole exchange: ui-workspace keeps the trigger (menu entry rendered only while the hole is occupied) and the adoption (`createWorkspace({path})`, conflict/error dialog, Choose again), the occupant owns everything between `open` and the picked path. One `cordis.yml` row therefore swaps the host capability and the client flow together; a mismatch is impossible by construction, and mounting two flow packages fails at client load (`single` hole). The earlier `host.describe.directoryPicker` advertisement and the client's kind branching are deleted — with composition wiring both sides, a wire fact for the client to branch on had no remaining consumer. The hole registry (`ctx.slots.entries`) replaces it as the per-menu-open occupancy read.
+
+Placement and policy rulings folded into this decision:
+
+- **Not the `ctx.fs` seam.** `packages/fs/` is the model/session-facing storage stack (policy events, sandbox-swappable backends). Riding it would couple GUI browsing to the model's confinement backend — swapping `fs-sandbox` for the model must never change GUI behavior — and OS facts (home anchoring, hidden conventions) are not storage primitives. The picker seam stays presentation-free and model-free; `packages/host/` is its consumer-domain home.
+- **Dependency survey (hand-roll vs adopt).** Node's stdlib *is* the maintained cross-platform OS layer (`readdir(withFileTypes)`, `homedir`, path semantics); surveyed alternatives fail the dependency bar — file-manager packages (`node-file-manager`, `files-and-folders`, Syncfusion's provider) are whole HTTP apps (fit), drive-letter helpers (`drivelist` native addon, `windows-drive-letters` ~7y stale) fail health/proportionality. The browse backend is a thin adapter over stdlib.
+- **Hidden entries: return-and-flag.** The host stamps `hidden` (POSIX dot convention) and returns everything; the client filters. Display policy stays client-side, and the planned show-hidden toggle becomes a client-only change. Windows' `FILE_ATTRIBUTE_HIDDEN` is not exposed by dirents — documented limitation until a native probe pays for itself.
+- **Symlinks: follow for enterability.** `stat` probes symlinks (broken/cyclic → skipped); crumbs keep the logical path the operator navigated, and `workspace.create` already canonicalizes via realpath at adoption.
+- **Listing levels are bounded, and streamed.** One `list` call returns at most `maxEntries` rows (config, default 1000 — GitHub's web-UI directory-listing bound). The level streams via `opendir` into a name-sorted window of `maxEntries + 1` candidates, so memory stays O(maxEntries) and enterability probing touches only windowed candidates; the wire `DirectoryListing` carries a required `truncated` flag so the client states incompleteness instead of silently missing tail entries. A windowed broken symlink is not backfilled from beyond the window — the eviction already marks the level truncated. Window insertion is binary with an O(1) full-window tail rejection (an oversized level must not pay a window scan per dirent), and `list(path, signal)` threads the carrier's request signal so a scan of a stalled network directory cannot outlive a disconnected caller — every await in the scan (open, each read, each symlink probe) races the signal, an aborted exit abandons rather than awaits the close (Node queues close behind in-flight reads), and abandoned settlements are swallowed so cleanup can never surface as an unhandled rejection. An unbounded level is a memory/responsiveness hole for large or adversarial directories.
+- **Whole-filesystem scope, no roots config.** `workspace.create` accepts arbitrary paths and the API serves bash-driving methods, so a browse root would be UX scoping, not a boundary; configurability without a consumer fails the evidence bar. Deferred until a deployment needs it.
+- **The native backend stays.** Plugin-form was the point: multiple providers can serve the seam (an Electron shell would provide the `native` interaction through its own dialog API). Kind naming: `dialog` was the first pick and was dropped — the browse interaction also presents a dialog (the in-app modal), so the word failed to discriminate; `native` names where the chooser runs.
+
+## Alternatives considered
+
+- **Extend `ctx.fs` with browse methods.** Rejected: authority-domain coupling above; also a listing-for-display contract (hidden flags, crumbs, home anchor) does not belong on a storage seam.
+- **One uniform seam method set (`pick(): path`).** Rejected: an in-app browser cannot be served behind a single host-side call — the browsing loop lives in the client and needs primitives on the wire; the native chooser cannot implement primitives. The interaction difference is irreducible, hence the discriminant.
+- **Direct stdlib calls inside apiproxy (no seam).** Rejected: keeps the gateway the only swap point (source edits), loses fixture/test backends, and contradicts the plugin doctrine that motivated the work.
+- **Adopting a file-manager/drive-enumeration dependency.** Rejected per the survey above; recorded here as the dependency policy requires.
+
+## Consequences
+
+- `cordis.yml` chooses the interaction; `apps/cli` mounts `-browse` (the shipped default — remote-capable picking out of the box), one row having swapped backend and UI together; `-native` remains the host-display alternative.
+- The wire gains `host.listDirectory`/`host.createDirectory` and four error codes; the connection fixture serves a deterministic browse tree and a deterministic `pickDirectory` path for keyless assembled tests.
+- A future interaction (or an Electron provider of the `native` interaction) is one dual-face backend package — no gateway surgery, no ui-workspace edits.
+- `ApiProxyDefaults.pickDirectory` (test-only injection) is gone; tests provide a stub `ctx.directoryPicker` like any other service.
@@ -0,0 +1,39 @@
+# Agent Note：web GUI 宿主的能力可辨识目录选择 seam
+
+状态：已实现
+
+[English](2026-07-28-directory-picker-capability-seam.md) | 中文
+
+## 问题
+
+web GUI 的"打开本地文件夹"流程被焊死在一种交互上：`host.pickDirectory` 调用编译进 `dsh-host-apiproxy` 的原生 OS 选择器（私有模块，仅测试注入缝）。这个形态服务不了远程部署——没有任何 OS 对话框能弹到另一台机器的浏览器里——而计划中的应用内目录浏览器（Figma `Harness` 802-56979）需要列举／创建原语，那是**另一种交互契约**，不是同一契约的另一种实现。想换交互只能改网关源码，违背仓库"一切皆插件"的立场。
+
+## 决策
+
+在 `packages/host/` 落一个三包能力 seam——`directory-picker`（接口）、`directory-picker-native`、`directory-picker-browse`（后端）——唯一契约方法 `capability()` 返回**可辨识联合**：`{ kind: 'native', pick(signal) }` 或 `{ kind: 'browse', list(path?), createDirectory(path, name) }`。网关（`dsh-host-apiproxy`）注入 `directoryPicker`，提供对应的 RPC，另一种 kind 的调用以 `directory-picker-unavailable` 应答。联合之所以可辨识，是因为后端差异在**交互形态**——压平成统一方法集会逼每个后端伪装另一方的形态。
+
+**client 侧靠 slot 组合，而非按广播分支。** ui-workspace 的两个触发表层各自声明一个 `single` 目录流洞（`conversation.hero.workspace.directoryFlow`／`sidebar.workspaces.directoryFlow`；之所以是两个 key，是因为一个洞只有一个声明它的 slot entry——owner 契约相同、占用者相同）。后端包是**双面包**：browser half 把匹配的交互注册进两个洞——`-native` 是驱动 `host.pickDirectory` 的无渲染占用者，`-browse` 是应用内的选择工作区目录对话框。洞的 owner 会话（`open`/`busy`/`onPicked`/`onCancel`/`onError`）承载整个交换：ui-workspace 保留触发（菜单入口仅在洞被占用时渲染）与接纳（`createWorkspace({path})`、冲突／错误对话框、重新选择），占用者持有从 `open` 到所选路径之间的一切。因此一行 `cordis.yml` 同时切换宿主能力与 client 流程；错配在构造上不可能，同时挂两个流程包会在 client 加载期失败（`single` 洞）。早先的 `host.describe.directoryPicker` 广播与客户端 kind 分支被删除——组合已经接好两侧后，供客户端分支用的 wire 事实不再有任何消费者。洞注册表（`ctx.slots.entries`）取而代之，成为每次打开菜单的占用读取。
+
+并入本决策的位置与策略裁决：
+
+- **不用 `ctx.fs` seam。** `packages/fs/` 是面向模型／会话的存储栈（policy 事件、sandbox 可换后端）。骑上去会把 GUI 浏览耦合进模型的限制后端——为模型换 `fs-sandbox` 绝不能改变 GUI 行为——而 OS 事实（home 锚定、隐藏约定）也不是存储原语。picker seam 保持无展示、无模型；`packages/host/` 是它消费方域的家。
+- **依赖调研（手写 vs 引入）。** Node 标准库本身就是维护中的跨平台 OS 层（`readdir(withFileTypes)`、`homedir`、路径语义）；调研过的替代品都过不了依赖门槛——文件管理器包（`node-file-manager`、`files-and-folders`、Syncfusion 的 provider）是整套 HTTP 应用（契合度不过），盘符工具（原生插件 `drivelist`、约七年未更的 `windows-drive-letters`）健康度／比例失当。browse 后端是标准库上的薄适配。
+- **隐藏条目：返回并打标。** 宿主标注 `hidden`（POSIX 点前缀约定）并返回全部条目；客户端过滤。展示策略留在客户端，计划中的"显示隐藏"开关变成纯客户端改动。Windows 的 `FILE_ATTRIBUTE_HIDDEN` 不被 dirent 暴露——记为限制，直到原生探测值回其成本。
+- **符号链接：为可进入性而跟随。** 用 `stat` 探测符号链接（断链／循环→跳过）；面包屑保留操作者导航的逻辑路径，`workspace.create` 在接纳时本就做 realpath 规范化。
+- **列举层级有上限，且流式处理。** 单次 `list` 至多返回 `maxEntries` 行（配置项，默认 1000——GitHub 网页端目录列举的同一上限）。层级经 `opendir` 流入一个按名排序、容量 `maxEntries + 1` 的候选窗口，内存保持 O(maxEntries)，可进入性探测只触及窗口内候选；线上 `DirectoryListing` 携带必填的 `truncated` 标志，让客户端明示不完整而不是静默缺尾。窗口内的断链符号链接不从窗口外回填——发生过驱逐本身已把层级标记为截断。窗口插入为二分查找、满窗尾部单次比较即拒绝（超大层级不能为每个 dirent 付出一次全窗扫描），且 `list(path, signal)` 透传载体的请求信号，滞塞网络目录的扫描不会在调用方断连后继续存活——扫描中的每个 await（打开、每次读取、每次符号链接探测）都与信号赛跑，中止路径放弃而非等待 close（Node 会把 close 排在在飞读取之后），被放弃的 settlement 全部吞掉，清理不会以未处理拒绝的形式冒出。无上限的层级对超大或恶意构造的目录就是内存／响应性漏洞。
+- **全盘可浏览，不做 roots 配置。** `workspace.create` 接受任意路径且 API 本就提供驱动 bash 的方法，浏览根只会是 UX 范围而非边界；没有消费方的可配置性过不了证据门槛。等到有部署需要再做。
+- **native 后端保留。** 插件化正是目的：多方都能提供该 seam（Electron 壳可以经自己的对话框 API 提供 `native` 交互）。kind 命名：最初选了 `dialog` 后被放弃——browse 交互同样以对话框呈现（应用内弹窗），这个词起不到判别作用；`native` 命名的是选择器运行的位置。
+
+## 曾考虑的替代方案
+
+- **给 `ctx.fs` 增加浏览方法。** 否决：上述权限域耦合；且面向展示的列举契约（hidden 标志、面包屑、home 锚点）不属于存储 seam。
+- **统一方法集的 seam（`pick(): path`）。** 否决：应用内浏览器无法藏在一次宿主侧调用后面——浏览循环在客户端，需要协议上的原语；而对话框实现不了原语。交互差异不可约，故用判别标签。
+- **apiproxy 里直接调标准库（不建 seam）。** 否决：换装点仍是改网关源码，失去 fixture／测试后端，与促成这项工作的插件教义相悖。
+- **引入文件管理器／盘符枚举依赖。** 按上文调研否决；依赖政策要求记录于此。
+
+## 后果
+
+- `cordis.yml` 决定交互形态；`apps/cli` 挂 `-browse`（随附默认——开箱即得可远程的选取），一行同时切换了后端与 UI；`-native` 仍是宿主屏幕方案。
+- 协议新增 `host.listDirectory`／`host.createDirectory` 与四个错误码；connection fixture 提供确定性浏览树与确定性 `pickDirectory` 路径供无密钥组装测试使用。
+- 未来的新交互（或提供 `native` 交互的 Electron 实现）只是一个双面后端包——无需网关手术，也不动 ui-workspace。
+- `ApiProxyDefaults.pickDirectory`（仅测试注入）删除；测试像提供其他服务一样提供 stub `ctx.directoryPicker`。
@@ -0,0 +1,6 @@
+# Bilingual-pair consistency record (docs/i18n/README.md): the git blob hash of each
+# side as of the last confirmed-consistent state. Both languages carry equal authority;
+# after editing either side, bring the other along and re-record with:
+#   pnpm run verify-translation-pairing --write .agents/notes/implemented/architecture/2026-07-28-dsh-native-typescript-source-launch.md
+2026-07-28-dsh-native-typescript-source-launch.md: 1ba1dd2663038ad7c49af71f8b428245f7fa3e2b
+2026-07-28-dsh-native-typescript-source-launch.zh.md: 02f84f34820469ad9e810ae17d79d3fe12b0cd4c
@@ -0,0 +1,47 @@
+# Agent Note: Native TypeScript source launch for dsh
+
+Status: implemented
+
+English | [中文](2026-07-28-dsh-native-typescript-source-launch.zh.md)
+
+> The Node-native launch vector is superseded by [dsh source launch through the tsx ESM hook](2026-07-29-dsh-source-launch-tsx-esm.md): Node 26.0.0 removed `--experimental-transform-types`, and the paths loader described here is deleted. The Cordis-config declaration gate (`verify-cordis-config`), the app-boot fail-loud plugin diagnostic, and the vendored `import type` marks remain current.
+
+## Problem
+
+The `dsh` source entry point originally used `tsx` to run `apps/cli/src/bin.ts`, with the same third-party loader implicitly handling both TypeScript transformation and the root tsconfig's `paths` resolution. With Node handling TypeScript natively, it does not apply tsconfig path mappings; resolving through package exports would instead mix potentially stale or nonexistent `lib/` artifacts into the source launch.
+
+Node's transform also does not perform type analysis. A type imported through an ordinary value import remains a runtime ESM request, and TypeScript's `export =` becomes a CommonJS assignment rather than an ESM default export. The source graph therefore has to use explicit type-only imports and native ESM exports; a resolve hook cannot repair incompatible source syntax.
+
+Cordis configuration introduces a separate resolution boundary. Bare plugins in `cordis.yml` do not pass through TypeScript import analysis, so their resolver manifest may omit the required dependencies. The Cordis Loader logs plugin import errors and leaves an entry without a fiber, but does not fail startup itself; a typo in the configuration can therefore produce an incomplete application with exit code 0.
+
+## Decision
+
+The `dsh` TUI, Web, and headless source launches use `node --experimental-transform-types`; Node performs TypeScript transformation without loading `tsx` or esbuild. `bin/dsh`, the root-level `dsh`/TUI/Web demos, and Code Mode TUI enter the same `apps/cli/src/bin.ts` launch chain. Test and E2E launchers retain their existing strategies, and the built `lib/bin.js` continues to run under ordinary Node.
+
+`scripts/tspath-loader.ts` registers only a module resolve hook. It uses `TSX_TSCONFIG_PATH` when set (resolving relative values from the invoking cwd) and otherwise reads the root `tsconfig.json`; `TsconfigPathsResolver` follows that config's `extends` chain through the repository's existing TypeScript development tool, selects exact or wildcard `paths` entries according to tsconfig rules, and maps matching workspace bare specifiers to `.ts`/`.mts`/`.cts` source files or directory index files. Node remains solely responsible for code transformation. The source-only loader is not part of the built CLI and `apps/cli` does not declare `typescript` as a runtime dependency.
+
+Source imports are redirected only when the target package is either the nearest package manifest's own name or one of that manifest's declared runtime dependencies. The Cordis Loader uses the configuration directory URL as the import parent; the resolver then searches upward for the workspace manifest that declares the plugin, so dependency ownership for `examples/tui-agent/cordis.yml` lies with `examples/package.json`, and dependency ownership for `apps/cli/cordis.yml` lies with `apps/cli/package.json`. Specifiers that do not match tsconfig paths, refer to undeclared dependencies, or are not bare all fall back to Node's default resolution.
+
+`verify-cordis-config` performs a one-way completeness check on both resolver manifests: every bare plugin package in a configuration must appear in the corresponding manifest's `dependencies`, while the manifest may contain extra dependencies not referenced by that configuration. The root `AGENTS.md` makes updating the configuration and dependencies together a standing rule.
+
+After the Loader settles, the shared `dsh-app-boot` checks every enabled entry that has no fiber and rejects startup with `plugin(s) failed to load: ...; Cordis startup failed because these plugin(s) could not be resolved`, listing all failed plugins. This diagnostic lives at the app layer and does not change the vendored Loader's startup behavior.
+
+Node-compatible TypeScript is part of this source-launch contract. Vendored Cordis, Loader, Include, HMR, and Schemastery mark erased imports with `import type`. Schemastery uses a native ESM default export and declares `type: module`; its `.mjs` and `.cjs` build outputs retain the existing ESM-default and callable-`require()` behavior. These divergences are recorded in `vendor/README.md`; no runtime behavior is added to the vendored frameworks.
+
+## Alternatives considered
+
+**Continue using `tsx`.** Rejected because `tsx`/esbuild would continue to own TypeScript transformation, so this launch chain could not prove that Node's native transformation works.
+
+**Load the built `lib/` through package exports from the source entry point.** Rejected because this would mix the source plane with the artifact plane; a zero-build development launch could read stale artifacts or fail outright.
+
+**Apply the root tsconfig `paths` unconditionally.** Rejected because this would allow undeclared cross-package imports and Cordis plugins to keep resolving, hiding mismatches between the manifest and the actual runtime graph.
+
+**Transform imports inside the custom loader.** Rejected because type-aware source rewriting would reintroduce a compiler-style transform and make the loader, rather than Node, responsible for TypeScript execution. Making the checked-in source Node-compatible keeps the launch boundary explicit.
+
+## Consequences
+
+- TUI/headless retain a zero-build source loop, while Web still builds its frontend artifacts before starting the CLI source entry. TypeScript syntax passes only through Node's native transform; the URL-only loader uses the checkout's root development dependencies and adds no CLI runtime dependency.
+- Workspace package imports and Cordis configuration dependencies must both be declared explicitly in the resolver manifest; the static gate prevents configuration from landing before its dependencies, while extra dependencies are not errors.
+- Plugin import failures no longer leave an incomplete application with exit code 0; the final error identifies both the Cordis startup failure and the specific plugin names, while the Loader's original error remains earlier in the logs.
+- Vendored source in the CLI graph must remain compatible with Node's transform-types module semantics; the local-modification log makes the upstream sync obligation explicit.
+- CI's `lib` mode, test/E2E launchers, and other example launchers retain their existing strategies; this native source loader covers only the `dsh` CLI application chain.
@@ -0,0 +1,47 @@
+# Agent Note: dsh 原生 TypeScript 源码启动
+
+Status: implemented
+
+[English](2026-07-28-dsh-native-typescript-source-launch.md) | 中文
+
+> Node 原生启动向量已被 [dsh 通过 tsx ESM hook 源码启动](2026-07-29-dsh-source-launch-tsx-esm.md) 取代：Node 26.0.0 移除了 `--experimental-transform-types`，本文描述的 paths loader 已删除。Cordis 配置声明门禁（`verify-cordis-config`）、app-boot 的 fail-loud 插件诊断以及 vendor 中的 `import type` 标注仍然有效。
+
+## 问题
+
+`dsh` 源码入口原本使用 `tsx` 运行 `apps/cli/src/bin.ts`，TypeScript 转换和根 tsconfig 的 `paths` 解析都由同一个第三方 loader 隐式处理。改由 Node 原生处理 TypeScript 后，Node 不会应用 tsconfig 路径映射；如果改为通过包导出解析，源码启动会混入可能陈旧或不存在的 `lib/` 产物。
+
+Node 的转换也不执行类型分析。通过普通值 import 导入的类型会保留为运行时 ESM 请求，而 TypeScript 的 `export =` 会变成 CommonJS 赋值，而不是 ESM default export。因此，源码图必须显式使用仅类型导入和原生 ESM 导出；resolve hook 无法修复不兼容的源码语法。
+
+Cordis 配置还引入了另一条解析边界。`cordis.yml` 中的 bare plugin 不经过 TypeScript import 分析，其解析方 manifest 可能漏掉所需依赖。Cordis Loader 会记录插件 import 错误，并留下没有 fiber 的 entry，但不会让启动本身失败；配置中的拼写错误因此可能得到退出码为 0 的残缺应用。
+
+## 决策
+
+`dsh` 的 TUI、Web 和无头源码启动使用 `node --experimental-transform-types`，由 Node 完成 TypeScript 转换，不加载 `tsx` 或 esbuild。`bin/dsh`、根级 `dsh`/TUI/Web demo 以及 Code Mode TUI 都进入同一条 `apps/cli/src/bin.ts` 启动链路。测试与 e2e 启动器保留各自现有策略，构建后的 `lib/bin.js` 继续由普通 Node 运行。
+
+`scripts/tspath-loader.ts` 只注册一个模块 resolve hook。设置 `TSX_TSCONFIG_PATH` 时，它会使用该路径（相对路径从调用方的 cwd 解析），否则读取根 `tsconfig.json`；`TsconfigPathsResolver` 使用仓库已有的 TypeScript 开发工具沿该配置的 `extends` 链解析，按 tsconfig 规则选择精确或 wildcard `paths` 条目，并将命中的 workspace bare specifier 映射到 `.ts`/`.mts`/`.cts` 源文件或目录 index 文件。代码转换始终只由 Node 负责。该源码专用 loader 不属于构建后的 CLI，`apps/cli` 也不会把 `typescript` 声明为运行时依赖。
+
+只有当目标包是最近 package manifest 的自身名称或其已声明的运行时依赖时，源码 import 才会重定向。Cordis Loader 使用配置目录 URL 作为 import parent；此时 resolver 会向上查找声明该插件的 workspace manifest。因此，`examples/tui-agent/cordis.yml` 的依赖由 `examples/package.json` 持有，`apps/cli/cordis.yml` 的依赖由 `apps/cli/package.json` 持有。未命中 tsconfig paths、引用未声明依赖或不是 bare specifier 的说明符全部交回 Node 默认解析。
+
+`verify-cordis-config` 对这两个解析方 manifest 执行单向完整性检查：配置中的每个 bare plugin package 都必须出现在对应 manifest 的 `dependencies` 中，manifest 可以包含该配置未引用的额外依赖。根 `AGENTS.md` 将同步更新配置和依赖定为常驻规则。
+
+Loader 完成结算后，共享的 `dsh-app-boot` 会检查每个已启用但没有 fiber 的 entry，并以 `plugin(s) failed to load: ...; Cordis startup failed because these plugin(s) could not be resolved` 拒绝启动，同时列出全部加载失败的插件。该诊断位于应用层，不改变 vendor 中 Loader 的启动行为。
+
+Node-compatible TypeScript 是这项源码启动契约的一部分。vendor 中的 Cordis、Loader、Include、HMR（热模块替换）和 Schemastery 使用 `import type` 标记会被擦除的导入。Schemastery 使用原生 ESM default export 并声明 `type: module`；其 `.mjs` 和 `.cjs` 构建产物分别保留现有的 ESM default export 行为和 `require()` 返回可调用值的行为。这些差异记录在 `vendor/README.md` 中；没有为 vendor 中的框架新增运行时行为。
+
+## 曾考虑的替代方案
+
+**继续使用 `tsx`。** 不采用，因为 `tsx`/esbuild 会继续负责 TypeScript 转换，本启动链路无法因此证明 Node 原生转换可用。
+
+**让源码入口通过包导出加载构建后的 `lib/`。** 不采用，因为这会混合 source plane 与 artifact plane；零构建开发启动可能读取陈旧产物或直接失败。
+
+**无条件应用根 tsconfig `paths`。** 不采用，因为这会让未声明的跨包 import 和 Cordis 插件继续成功解析，从而掩盖 manifest 与实际运行图之间的不一致。
+
+**在自定义 loader 内转换 import。** 不采用，因为感知类型的源码改写会重新引入编译器式转换，并让 loader 而非 Node 负责执行 TypeScript。使签入仓库的源码兼容 Node，可以让启动边界保持显式。
+
+## 后果
+
+- TUI／无头界面保留零构建源码回路，Web 仍会在启动 CLI 源码入口前构建前端产物。TypeScript 语法只经过 Node 原生转换；仅处理 URL 的 loader 使用 checkout 根目录的开发依赖，不增加 CLI 运行时依赖。
+- workspace package import 和 Cordis 配置依赖都必须在解析方 manifest 中明确声明；静态门禁防止配置先于依赖落地，额外依赖不构成错误。
+- 插件 import 失败不再留下退出码为 0 的残缺应用；最终错误同时说明 Cordis 启动失败及具体插件名，Loader 的原始错误仍会保留在更早的日志中。
+- CLI 源码图中的 vendor 源码必须与 Node 的 transform-types 模块语义兼容；本地修改记录明确了上游同步义务。
+- CI 的 `lib` 模式、测试／e2e 启动器和其他示例启动器保留各自现有策略；该原生源码 loader 只覆盖 `dsh` CLI 应用链路。
@@ -0,0 +1,6 @@
+# Bilingual-pair consistency record (docs/i18n/README.md): the git blob hash of each
+# side as of the last confirmed-consistent state. Both languages carry equal authority;
+# after editing either side, bring the other along and re-record with:
+#   pnpm run verify-translation-pairing --write .agents/notes/implemented/architecture/2026-07-28-identified-immutable-message-values.md
+2026-07-28-identified-immutable-message-values.md: cdb0f1aadc4796b5aa0642a3994d3e3e4ab67bd9
+2026-07-28-identified-immutable-message-values.zh.md: 3e1732cb5b7f49fb9349b2e1790cf5b3ec1474be
@@ -0,0 +1,50 @@
+# Agent Note: Create every message as an identified immutable value
+
+Status: implemented
+
+English | [中文](2026-07-28-identified-immutable-message-values.zh.md)
+
+## Problem
+
+The harness had several message-shaped representations with different identity rules. Agent input acquired an inbox correlation id only when the loop accepted it, while durable user messages, assistant messages, tool results, and model-request messages could have no identity. Prompt admission therefore sat between creation and identity, and equivalent content was copied across live events, durable events, and model requests without one value that named the message throughout its lifetime.
+
+This made identity a routing side effect rather than a message invariant. Producers could not refer to a message before calling the agent, prompt hooks received content and source separately, and later projections had to reconstruct a message while deciding whether an id existed. Immutability also began at different boundaries: some inputs were frozen by the loop, some only by session append, and provider-produced assistant output used a separate provenance-bearing shape.
+
+## Decision
+
+`@deepseek-ai/dsh-llm` owns one `Message` value with required `id`, `role`, `content`, and `source`. `MessageId` is opaque and shared by user, assistant, and tool-result messages. A message receives its id at creation, before routing, prompt admission, durable append, or request projection. The same id survives every representation boundary.
+
+`createMessage(input)` is the canonical role-generic creation boundary. It mints a `MessageId`, detaches the supplied role, content, and source, and deep-freezes the complete value before returning it. `createUserMessage({ content, source })` fixes the user role for prompt and context producers. `createAssistantMessage({ content, source })` fixes both the assistant role and the model source kind, so model-output producers supply only content and model provenance. All creation helpers exclude an input id so callers cannot accidentally present creation as import. `freezeMessage(message)` is the separate import or transformation boundary: it detaches and deep-freezes a message whose identity already exists, without minting a replacement.
+
+The helpers live in `dsh-llm` beside the base message vocabulary because their complete contracts depend only on that vocabulary. `createToolResultMessage()` belongs with the other creation helpers: it couples a tool call id to the exact user-role tool-result block and source without depending on session state or events. `dsh-session` consumes complete messages rather than owning their construction.
+
+The `Agent` interface accepts a complete `UserMessage`. `send`, `followup`, `steer`, and `inject` never allocate or return identity; they freeze an imported value whose id the caller already holds. Prompt admission receives that message directly. A content rewrite creates a frozen replacement with the same id, while an additional context is a separately created `UserMessage` with its own id.
+
+Durable message-producing events store complete messages. `user/message` stores its `UserMessage` directly; `assistant/message`, `tool/result`, and `steering/message` wrap their role-specialized message beside event-local position, usage, failure, or presentation facts. Session derivation returns those frozen values instead of reconstructing anonymous messages. Assistant assembly creates a model-sourced message when a response completes, and tool execution creates a tool-sourced message when a result is committed.
+
+Any operation that changes only the representation of an existing semantic message preserves its id and returns another frozen value. An operation that creates a new semantic message mints a new id. Compaction content rewrites therefore preserve the rewritten tool-result identity, while a summary checkpoint is a new message.
+
+## Alternatives considered
+
+**Keep ids optional on the base message.** This would minimize fixture migration and allow provider or persistence shapes to remain anonymous. It would also preserve the original ambiguity: every consumer would need to branch on whether identity exists, and no type would prove that admission, logging, or projection retained it.
+
+**Let `Agent.send()` allocate the id.** This keeps identity scoped to inbox correlation but makes the agent call the earliest point at which a producer can name its own message. Prompt construction, UI attachments, and synchronous enqueue/discard coordination then need content matching or an out-of-band token before `send()` returns.
+
+**Let each durable event allocate a new id.** This gives persisted messages identities but deliberately breaks correlation with the live input and makes replayed requests appear to contain different messages. Identity belongs to the semantic value, not to each envelope that carries it.
+
+**Freeze only at agent or session admission.** This avoids a creation helper but leaves an identified mutable interval in which caller code can change the meaning associated with an id. The decision makes “has an id” and “is an immutable snapshot” coincide.
+
+## Consequences
+
+Every message producer must choose creation or import explicitly, and tests construct complete values rather than partial content/source records. UUID generation moves outward to the first semantic creation point, so deterministic fixtures that provide an existing id use `freezeMessage()` instead of `createMessage()`.
+
+Live inbox events, durable events, derived history, and model requests can correlate one message without content equality or envelope-specific ids. Prompt admission and UI attachment cleanup can compare `MessageId` before a turn exists. Deep freezing prevents a producer, hook, or observer from changing the value after identity is established.
+
+The shared representation removes the old `UserMessageData`/`AgentMessage` split and folds provider provenance into typed message sources. Event envelopes still own facts that are not message semantics, such as turn and step position, token usage, internal tool failure identity, and presentation metadata.
+
+The message and helper unit tests pin immediate identity, detachment, deep immutability, and preservation of an imported id. Agent-loop tests pin identity across admission, inbox lifecycle, durable append, content rewriting, and cancellation; session tests pin frozen derivation and identity-preserving replacement.
+
+## Related
+
+- [Unify agent delivery on send(target × wakeup) and coalesce injected context into user/message](2026-07-22-unified-send-and-coalesced-user-messages.md) — this note supersedes its input-representation and agent-assigned-id details while retaining its routing decision.
+- [Reconstructable requests](2026-07-05-reconstructable-requests.md) — the session log remains the authority for every model-visible input.
@@ -0,0 +1,50 @@
+# Agent Note: 将每条消息创建为带标识的不可变值
+
+Status: implemented
+
+[English](2026-07-28-identified-immutable-message-values.md) | 中文
+
+## 问题
+
+harness 曾存在多种形似消息的表示，各自采用不同的标识规则。agent（智能体）输入只有在 loop 接受后才会取得 inbox 关联 id，而持久用户消息、assistant 消息、工具结果和模型请求消息都可能没有标识。因此，提示词准入介于创建消息与建立标识之间；等价内容会在实时事件、持久事件和模型请求之间复制，却没有一个值能在消息的整个生命周期中标识它。
+
+这使标识成为路由的副作用，而不是消息不变量。生产方无法在调用 agent 前引用一条消息，提示词钩子会分别接收内容和来源，后续投影则必须一边重建消息，一边决定 id 是否存在。不可变性也从不同边界开始：部分输入由 loop 冻结，部分直到会话追加时才冻结，提供方产生的 assistant 输出则使用另一种携带溯源信息的形状。
+
+## 决策
+
+`@deepseek-ai/dsh-llm` 持有唯一一种 `Message` 值，其 `id`、`role`、`content` 和 `source` 均为必填。`MessageId` 是不透明标识，由用户消息、assistant 消息和工具结果消息共享。消息在创建时就会获得 id，早于路由、提示词准入、持久追加或请求投影。同一个 id 会跨越每个表示边界。
+
+`createMessage(input)` 是角色通用的规范创建边界。它会生成 `MessageId`，将输入的角色、内容和来源与输入分离，并在返回完整值前将其深度冻结。`createUserMessage({ content, source })` 为提示词和上下文生产方固定 user 角色。`createAssistantMessage({ content, source })` 同时固定 assistant 角色与模型来源类别，因此模型输出生产方只需提供内容和模型溯源信息。所有创建辅助函数的输入都不包含 id，因此调用方不会意外地把创建表达为导入。`freezeMessage(message)` 是独立的导入或转换边界：它会将已有标识的消息与输入分离并深度冻结，不会生成替代标识。
+
+这些辅助函数位于基础消息词汇旁的 `dsh-llm` 中，因为它们的完整契约只依赖该词汇。`createToolResultMessage()` 与其他创建辅助函数同属此处：它使用同一个工具调用 id，将工具来源与确切的 user-role 工具结果块耦合起来，不依赖会话状态或事件。`dsh-session` 只消费完整消息，不负责构造它们。
+
+`Agent` 接口接收完整的 `UserMessage`。`send`、`followup`、`steer` 和 `inject` 绝不会分配或返回标识；它们会冻结导入的值，而调用方已经持有该值的 id。提示词准入会直接接收该消息。改写内容时会创建具有相同 id 的冻结替代值，而每个附加上下文都是单独创建的 `UserMessage`，拥有自己的 id。
+
+产生持久消息的事件会存储完整消息。`user/message` 直接存储其 `UserMessage`；`assistant/message`、`tool/result` 和 `steering/message` 则将各自角色专用的消息与事件本地的位置、用量、失败或呈现事实包装在一起。会话派生会返回这些冻结值，而不是重建匿名消息。assistant 组装会在响应完成时创建模型来源的消息，工具执行会在提交结果时创建工具来源的消息。
+
+仅改变已有语义消息表示的操作会保留其 id，并返回另一个冻结值。创建新语义消息的操作则会生成新 id。因此，压缩（compaction）内容改写会保留被改写工具结果的标识，而摘要检查点是一条新消息。
+
+## 考虑过的替代方案
+
+**让基础消息的 id 保持可选。** 这能减少 fixture（测试前置数据）迁移，并允许提供方或持久化形状继续保持匿名，但也会保留原有歧义：每个消费方都必须根据标识是否存在执行分支，且没有任何类型能证明准入、记录或投影保留了标识。
+
+**让 `Agent.send()` 分配 id。** 这会将标识限定在 inbox 关联范围内，却也会让 agent 调用成为生产方可以标识自身消息的最早时机。这样一来，在 `send()` 返回前，提示词构造、UI 附件和同步入队／丢弃协调都需要进行内容匹配，或使用带外 token。
+
+**让每个持久事件分配新 id。** 这能为持久消息提供标识，却会有意切断它与实时输入的关联，并让回放请求表现得像包含了不同消息。标识属于语义值，而不是承载它的每个封装。
+
+**只在 agent 或会话准入时冻结。** 这能省去创建辅助函数，却会留下一个带标识但可变的时间区间，调用方代码可以在这段时间内改变该 id 所关联的含义。本决策让「拥有 id」与「是不可变快照」同时成立。
+
+## 后果
+
+每个消息生产方都必须显式选择创建或导入，测试也会构造完整值，而不是不完整的内容／来源记录。UUID 的生成会前移至最初的语义创建点，因此提供已有 id 的确定性 fixture 会使用 `freezeMessage()`，而不是 `createMessage()`。
+
+实时 inbox 事件、持久事件、派生历史和模型请求可以关联同一条消息，无需比较内容或使用封装专用 id。提示词准入和 UI 附件清理可以在轮次存在之前比较 `MessageId`。深度冻结可以防止生产方、钩子或观察方在标识建立后更改消息值。
+
+共享表示移除了旧的 `UserMessageData`/`AgentMessage` 划分，并将提供方溯源信息纳入带类型的消息来源。事件封装仍持有不属于消息语义的事实，例如轮次与步骤位置、token 用量、内部工具失败标识和呈现元数据。
+
+消息和辅助函数的单元测试会固定即时标识、输入分离、深度不可变性，以及导入 id 的保留。agent loop 测试会固定标识跨越准入、inbox 生命周期、持久追加、内容改写和取消的行为；会话测试会固定冻结派生和保留标识的替换行为。
+
+## 相关
+
+- [统一通过 send(target × wakeup) 交付 agent 消息，并将注入上下文合并到 user/message](2026-07-22-unified-send-and-coalesced-user-messages.md)——本记录取代其中的输入表示和由 agent 分配 id 的细节，同时保留其路由决策。
+- [可重建的请求](2026-07-05-reconstructable-requests.md)——会话日志仍是每项模型可见输入的权威来源。
@@ -0,0 +1,6 @@
+# Bilingual-pair consistency record (docs/i18n/README.md): the git blob hash of each
+# side as of the last confirmed-consistent state. Both languages carry equal authority;
+# after editing either side, bring the other along and re-record with:
+#   pnpm run verify-translation-pairing --write .agents/notes/implemented/architecture/2026-07-29-dsh-source-launch-tsx-esm.md
+2026-07-29-dsh-source-launch-tsx-esm.md: 93fbb248b45efde37d5fbdb1ec4b812ab3332088
+2026-07-29-dsh-source-launch-tsx-esm.zh.md: 48f410bd846e5808cc95180279348a0ac5ba1c95
@@ -0,0 +1,38 @@
+# Agent Note: dsh source launch through the tsx ESM hook
+
+Status: implemented
+
+English | [中文](2026-07-29-dsh-source-launch-tsx-esm.zh.md)
+
+> Supersedes [native TypeScript source launch](2026-07-28-dsh-native-typescript-source-launch.md): Node removed the capability that decision was built on.
+
+## Problem
+
+The [native source-launch decision](2026-07-28-dsh-native-typescript-source-launch.md) ran `apps/cli/src/bin.ts` under `node --experimental-transform-types` with a resolve-only paths loader, so Node owned TypeScript transformation. Node 26.0.0 removed `--experimental-transform-types` (the process rejects the flag with `bad option`), keeping only strip mode, and strip mode rejects syntax this source graph requires: vendored Cordis parameter properties (`constructor(private ctx: Context)`), the `@Inject` decorators in `vendor/hmr`, and runtime enums/namespaces throughout `vendor/` and `packages/workflow`. The repository's engines range (`^22.19.0 || >=24.0.0`) includes Node 26, so the native launch chain could not start at all there — and no CI job executed the real launch vector, so the incompatibility shipped silently.
+
+Startup latency also mattered: the off-thread `module.register()` hooks worker serialized every resolution across threads (~440ms of `makeSyncRequest` wait during TUI boot), and the full tsx default (`--import tsx`) pays ~0.4s in its CJS hook's resolution amplification.
+
+## Decision
+
+The `dsh` TUI, Web, and headless source launches run `node --import tsx/esm`: tsx's ESM-only hook owns both TypeScript transformation and tsconfig `paths` projection. `bin/dsh`, the root `dsh`/`demo:tui`/`demo:web` scripts, and the Code Mode TUI overlay use the same vector; `bin/dsh` references the hook and tsconfig by absolute checkout paths (bare `tsx/esm` does not resolve from an arbitrary cwd) and pins `TSX_TSCONFIG_PATH` to the root tsconfig. The CJS hook stays off because the CLI source graph is ESM-only; measured TUI time-to-banner is ~0.7s versus ~1.1s under the full tsx default and ~0.75s under the removed native chain.
+
+`scripts/tspath-loader.ts` and `apps/cli/src/tsconfig-paths-loader.ts` are deleted. With them went the loader's runtime rule of mapping a workspace import only for declared runtime dependencies — tsx applies the `paths` map unconditionally. Declaration completeness now rests on the static gates alone: `verify-cordis-config` for configured bare plugins, and workspace constraints for manifests. (That runtime rule found real bugs: `dsh-plan-mode` and `dsh-tool-tasks` imported `@deepseek-ai/dsh-llm` while declaring it only in devDependencies; fixed alongside this change.)
+
+The node-compat CI matrix (Node 22.19 and 26) gains `dsh-source-launch-smoke` (`apps/cli/tests/source-launch.compat.spec.ts`): a keyless piped-stdio launch of the exact production vector asserting the non-zero-exit TTY refusal. Any future Node change to module hooks or TypeScript handling turns this gate red instead of breaking developers' `pnpm dsh`.
+
+## Alternatives considered
+
+**Keep the native chain on Node ≤25 and branch by version.** Rejected: two transformation semantics (amaro versus esbuild) diverge on edge syntax, the launcher grows version probing, and the node-compat matrix must cover both paths — heavy maintenance for an experimental flag that already changed under us. amaro also rejects the `@Inject` decorators `vendor/hmr` uses, so the native path could not boot the shipped default TUI config anyway.
+
+**Make the source graph erasable-only so Node 26 strip mode accepts it.** Rejected: parameter properties and value namespaces pervade vendored Cordis/cosmokit/loader/schemastery; rewriting them is unbounded churn re-applied on every vendor sync.
+
+**A repo-owned in-thread loader (`module.registerHooks()` + esbuild or `@swc/core` transform).** Rejected for now: prototypes measured ~0.45s (esbuild path untested end-to-end; SWC breaks on `vendor/hmr`'s decorator + namespace merge in both decorator modes), but it means owning transform correctness and a resolve hook that tsx already provides. Revisit only if the ~0.3s gap becomes a real cost; the profiling evidence lives in the PR discussion.
+
+**Run built `lib/` for Node 26 and keep native for 24.** Rejected: loses the zero-build development loop on the newest Node line and mixes source and artifact planes.
+
+## Consequences
+
+- One launch vector across the whole engines range, including future Node lines that change native TypeScript support; the smoke gate enforces it per matrix line.
+- TypeScript transformation is delegated to tsx/esbuild again, reversing the prior note's goal of proving Node-native transformation; that goal is unreachable while vendored sources use non-erasable syntax and Node ships no transform mode.
+- The runtime declared-dependency enforcement in source launches is gone; undeclared workspace imports now surface only through static gates or built-mode resolution failures.
+- Startup improves ~0.4s over the full tsx default (`demo:headless` and ACP keep `--import tsx`; their graphs were not audited for CJS-hook dependence and their launch latency is not on the interactive path).
@@ -0,0 +1,38 @@
+# Agent Note: dsh 通过 tsx ESM hook 源码启动
+
+Status: implemented
+
+[English](2026-07-29-dsh-source-launch-tsx-esm.md) | 中文
+
+> 取代[原生 TypeScript 源码启动](2026-07-28-dsh-native-typescript-source-launch.md)：Node 移除了该决策所依赖的能力。
+
+## 问题
+
+[原生源码启动决策](2026-07-28-dsh-native-typescript-source-launch.md)让 `apps/cli/src/bin.ts` 在 `node --experimental-transform-types` 下运行，配合一个只做解析的 paths loader，由 Node 负责 TypeScript 转换。Node 26.0.0 移除了 `--experimental-transform-types`（进程以 `bad option` 拒绝该 flag），只保留 strip 模式，而 strip 模式无法接受这个源码图必需的语法：vendor Cordis 中的参数属性（`constructor(private ctx: Context)`）、`vendor/hmr` 中的 `@Inject` 装饰器，以及遍布 `vendor/` 与 `packages/workflow` 的运行时 enum/namespace。仓库的 engines 范围（`^22.19.0 || >=24.0.0`）包含 Node 26，因此原生启动链在其上完全无法启动——且没有任何 CI 任务执行过真实启动向量，这一不兼容悄然发布。
+
+启动延迟同样是问题：off-thread 的 `module.register()` hooks worker 把每次解析都跨线程序列化（TUI 启动期间约 440ms 的 `makeSyncRequest` 等待），而完整 tsx 默认形态（`--import tsx`）的 CJS hook 解析放大要多付约 0.4s。
+
+## 决策
+
+`dsh` 的 TUI、Web 与无头源码启动运行 `node --import tsx/esm`：由 tsx 的 ESM-only hook 同时负责 TypeScript 转换与 tsconfig `paths` 投影。`bin/dsh`、根目录的 `dsh`/`demo:tui`/`demo:web` 脚本以及 Code Mode TUI overlay 使用同一向量；`bin/dsh` 以 checkout 的绝对路径引用 hook 与 tsconfig（裸的 `tsx/esm` 无法从任意 cwd 解析），并将 `TSX_TSCONFIG_PATH` 固定到根 tsconfig。CJS hook 保持关闭，因为 CLI 源码图是纯 ESM；实测 TUI 到 banner 约 0.7s，对比完整 tsx 默认形态约 1.1s、已移除的原生链约 0.75s。
+
+`scripts/tspath-loader.ts` 与 `apps/cli/src/tsconfig-paths-loader.ts` 已删除。随之消失的还有该 loader "仅为已声明运行时依赖映射 workspace import" 的运行时规则——tsx 无条件应用 `paths` 映射。声明完整性现在仅由静态门禁保障：配置的裸插件走 `verify-cordis-config`，manifest 走 workspace constraints。（该运行时规则确实发现过真实缺陷：`dsh-plan-mode` 与 `dsh-tool-tasks` 导入 `@deepseek-ai/dsh-llm` 却只声明在 devDependencies；已随本变更修复。）
+
+node-compat CI 矩阵（Node 22.19 与 26）新增 `dsh-source-launch-smoke`（`apps/cli/tests/source-launch.compat.spec.ts`）：以精确的生产启动向量做 keyless 管道 stdio 启动，断言非零退出的 TTY 拒绝。未来 Node 对模块 hook 或 TypeScript 处理的任何改动都会让该门禁变红，而不是破坏开发者的 `pnpm dsh`。
+
+## 备选方案
+
+**在 Node ≤25 保留原生链并按版本分叉。** 拒绝：两套转换语义（amaro 与 esbuild）在边缘语法上会分歧，启动器要加版本探测，node-compat 矩阵要覆盖两条路径——为一个已经变动过的 experimental flag 付出沉重维护。而且 amaro 也不支持 `vendor/hmr` 使用的 `@Inject` 装饰器，原生路径本来就无法启动随附的默认 TUI 配置。
+
+**把源码图改成 erasable-only 以适配 Node 26 strip 模式。** 拒绝：参数属性与值 namespace 遍布 vendor 的 Cordis/cosmokit/loader/schemastery；改写是无界 churn，且每次 vendor sync 都要重做。
+
+**仓库自有的同线程 loader（`module.registerHooks()` + esbuild 或 `@swc/core` 转换）。** 暂拒：原型实测约 0.45s（esbuild 路径未端到端验证；SWC 在 `vendor/hmr` 的装饰器 + namespace 合并上两种装饰器模式都会崩），但意味着自行负责转换正确性和一个 tsx 已经提供的 resolve hook。仅当约 0.3s 的差距成为真实成本时再重启；profiling 证据在 PR 讨论中。
+
+**Node 26 运行构建产物 `lib/`，24 保留原生。** 拒绝：在最新 Node 版本线上失去零构建开发循环，且混淆源码面与产物面。
+
+## 结果
+
+- 整个 engines 范围（包括未来改变原生 TypeScript 支持的 Node 版本线）只有一个启动向量；冒烟门禁按矩阵行强制执行。
+- TypeScript 转换重新委托给 tsx/esbuild，逆转了前一篇 note "证明 Node 原生转换可用" 的目标；在 vendor 源码使用不可擦除语法且 Node 不再提供 transform 模式的情况下，该目标不可达。
+- 源码启动中的运行时依赖声明强制不复存在；未声明的 workspace import 现在只能通过静态门禁或构建模式的解析失败暴露。
+- 启动相比完整 tsx 默认形态快约 0.4s（`demo:headless` 与 ACP 保持 `--import tsx`：其依赖图未就 CJS hook 依赖性做审计，且其启动延迟不在交互路径上）。
@@ -0,0 +1,6 @@
+# Bilingual-pair consistency record (docs/i18n/README.md): the git blob hash of each
+# side as of the last confirmed-consistent state. Both languages carry equal authority;
+# after editing either side, bring the other along and re-record with:
+#   pnpm run verify-translation-pairing --write .agents/notes/implemented/bug-fix/2026-07-28-load-pre-identity-session-messages.md
+2026-07-28-load-pre-identity-session-messages.md: 2901527658421b37576bdf5b49e66829104a3b41
+2026-07-28-load-pre-identity-session-messages.zh.md: 61d57ac9f3318299b63faa659b6d155e8e89fae3
@@ -0,0 +1,38 @@
+# Agent Note: Load sessions persisted before message identity
+
+Status: implemented
+
+English | [中文](2026-07-28-load-pre-identity-session-messages.zh.md)
+
+## Problem
+
+The identified immutable message change replaced four durable event payloads with complete message values. Existing v0 JSONL and SQLite sessions still held the immediately preceding shapes: direct `content`/`source` on user and steering events, `content`/`provenance` on assistant events, and `callId`/`content`/`isError` on tool results. Their headers still matched `SESSION_FORMAT_VERSION`, but current-shape validation rejected them before resume could construct a live `Session`.
+
+Changing the message representation without a version bump made those logs indistinguishable at the header level from current v0 logs. The runtime needs a narrow import rule that restores data created by the supported first-party backends without weakening validation for unrelated obsolete or malformed events.
+
+## Decision
+
+`PersistenceCoordinator` normalizes the four exact pre-identity message payloads after backend decoding and before current message validation. It wraps their existing semantic fields in the current role-specific message shape and assigns `legacy-message:<session-id>:<event-seq>` as the deterministic imported `MessageId`. A legacy `tool/result` content replacement inherits the imported id of its replacement target, preserving the current content-only rewrite invariant.
+
+The same normalization runs for `load`, `inspect`, an ownerless loaded state claiming its live session, and HMR prefix adoption. Prefix comparisons therefore compare the live current-shape seed with the same normalized stored view. Current-looking wrappers with missing or invalid fields are not repaired, and unsupported event vocabulary, request headers, versions, and surface relations retain their existing rejection paths.
+
+The upgrade is read-only. Stored legacy records remain unchanged; a resumed session appends only current-shape events after them. Deterministic identities make repeated loads and a mixed legacy/current log reproduce the same message ids without a backend-specific rewrite transaction.
+
+## Alternatives considered
+
+**Reject the logs under the pre-release compatibility stance.** This is the default for unrelated v0 churn, but it strands real first-party sessions even though every old field maps unambiguously to the current message representation.
+
+**Rewrite the complete stored log in place.** This would canonicalize the artifact but violate the append-only storage contract, require separate atomic replacement mechanisms for JSONL and SQLite, and expand a read compatibility fix into a migration system.
+
+**Mint random ids on each load.** The messages would satisfy the type shape but lose stable identity across inspect, resume, restart, and mixed legacy/current appends.
+
+## Consequences
+
+Pre-identity JSONL and SQLite sessions resume with their original message content, sources, provider provenance, tool correlation, errors, metadata, and surface replacements. The returned events are otherwise indistinguishable from current imported message snapshots and remain deeply frozen.
+
+This is one explicit same-version import exception, not a general v0 compatibility layer. Adding another exception requires another complete, unambiguous mapping at the persistence boundary; malformed current data continues to fail rather than being guessed into validity. The shared coordinator contract exercises the upgrade against the in-memory reference, JSONL, and SQLite backends, including deterministic reload and tool-result replacement identity.
+
+## Related
+
+- [Create every message as an identified immutable value](../architecture/2026-07-28-identified-immutable-message-values.md) — owns the current message identity and immutability contract.
+- [Session persistence as an abstract service](../architecture/2026-06-14-session-persistence.md) — owns the append-only backend and resume boundary.
@@ -0,0 +1,38 @@
+# Agent Note: 加载消息标识机制引入前持久化的会话
+
+Status: implemented
+
+[English](2026-07-28-load-pre-identity-session-messages.md) | 中文
+
+## 问题
+
+带标识的不可变消息变更将四种持久事件载荷替换为完整消息值。现有的 v0 JSONL 和 SQLite 会话仍保留紧邻该变更之前的形状：用户事件和 steering（中途引导）事件直接携带 `content`/`source`，assistant 事件携带 `content`/`provenance`，工具结果则携带 `callId`/`content`/`isError`。这些会话的 header 仍与 `SESSION_FORMAT_VERSION` 匹配，但当前形状验证会拒绝它们，导致恢复流程无法构造实时 `Session`。
+
+消息表示改变时没有提升版本，导致这些日志无法仅凭 header 与当前的 v0 日志区分。运行时需要一条范围受限的导入规则，既能恢复受支持的第一方后端所创建的数据，又不削弱对无关过时事件或格式错误事件的验证。
+
+## 决策
+
+`PersistenceCoordinator` 会在后端解码之后、当前消息验证之前，规范化消息标识机制引入前的四种特定消息载荷。它将载荷现有的语义字段包装进当前按角色区分的消息形状，并为其分配确定性的导入 `MessageId`：`legacy-message:<session-id>:<event-seq>`。旧版 `tool/result` 的内容替换会继承替换目标导入后的 id，从而保持当前仅改写内容的不变量。
+
+同一项规范化也用于 `load`、`inspect`、无 owner 的已加载状态认领其实时会话，以及 HMR（热模块替换）前缀接管。因此，前缀比较会将实时的当前形状 seed 与同一份规范化存储视图进行比较。看似当前形状、但字段缺失或无效的包装层不会被修复；不受支持的事件词汇、请求 header、版本和 surface 关系仍沿用现有拒绝路径。
+
+这项升级只发生在读取时。存储中的旧版记录保持不变；会话恢复后，只会在其后追加当前形状的事件。确定性标识使重复加载以及新旧形状混合的日志无需执行后端专用的重写事务，也能复现相同的消息 id。
+
+## 考虑过的替代方案
+
+**按照预发布兼容性立场拒绝这些日志。** 这是处理其他 v0 形状变动的默认方式，但即使每个旧字段都能明确映射到当前消息表示，它仍会导致真实的第一方会话无法恢复。
+
+**就地重写完整的存储日志。** 这会使产物规范化，但违反仅追加存储契约，还需要为 JSONL 和 SQLite 分别实现原子替换机制，并将一次读取兼容性修复扩大为迁移系统。
+
+**每次加载时随机生成 id。** 这些消息会满足类型形状，却无法在检查、恢复、重启以及新旧形状混合追加之间保持稳定标识。
+
+## 后果
+
+消息标识机制引入前的 JSONL 和 SQLite 会话可以恢复，并保留原始的消息内容、来源、提供方溯源信息、工具关联、错误、元数据和 surface 替换。除此之外，返回事件与当前导入的消息快照无法区分，并且仍然经过深度冻结。
+
+这是一个显式的同版本导入例外，而非通用的 v0 兼容层。若要增加另一个例外，必须在持久化边界提供另一套完整且无歧义的映射；当前数据若格式错误，系统仍会拒绝，而不会猜测如何将其变成有效数据。共享协调器契约会针对内存参考实现、JSONL 和 SQLite 后端验证这项升级，包括重新加载时的确定性，以及工具结果替换时的标识继承。
+
+## 相关
+
+- [将每条消息创建为带标识的不可变值](../architecture/2026-07-28-identified-immutable-message-values.md)：该记录负责当前的消息标识与不可变性契约。
+- [会话持久化作为抽象服务](../architecture/2026-06-14-session-persistence.md)：该记录负责仅追加后端与恢复边界。
@@ -0,0 +1,6 @@
+# Bilingual-pair consistency record (docs/i18n/README.md): the git blob hash of each
+# side as of the last confirmed-consistent state. Both languages carry equal authority;
+# after editing either side, bring the other along and re-record with:
+#   pnpm run verify-translation-pairing --write .agents/notes/implemented/bug-fix/2026-07-28-themed-scrollbars-and-reserved-gutter.md
+2026-07-28-themed-scrollbars-and-reserved-gutter.md: 38228c868bb00210118e8110feb722fb81d0d56c
+2026-07-28-themed-scrollbars-and-reserved-gutter.zh.md: 9fafe1faa9303b5e2e23a1b3064904f71494026d
@@ -0,0 +1,84 @@
+# Agent Note: The scrollbar tokens get their consumer, and the workspace list reserves its gutter
+
+Status: implemented
+
+English | [中文](2026-07-28-themed-scrollbars-and-reserved-gutter.zh.md)
+
+## Problem
+
+`design-platform.css` declares four `--dsw-alias-scrollbar-*` tokens (`bg-l1`, `bg-l2`, `hover-l1`, `hover-l2`) in both palettes, and no rule anywhere in the client read them. A defined token with no consumer is not a theme: every scrolling region rendered the user agent's own scrollbar, which knows nothing about the palette, so the dark theme showed a light native bar against dark surfaces.
+
+The visible symptom that surfaced the gap was elsewhere. The workspace browser's session list (`.list` in `WorkspaceBrowser.module.css`) is the sidebar's only scrolling region, and each row's trailing content sits flush against the row's 8px right padding — `.time` in `rows/Rows.module.css` is `flex: none`, as are the action buttons that replace it on hover. An overlaid scrollbar therefore painted on top of the relative timestamp. Reserving space in that one list would have left the bar itself unthemed, so the two halves are one change.
+
+## Decision
+
+`packages/client/ui-theme/src/styles/scrollbar.css` is the sole consumer of the four tokens, and the fifth ui-theme sheet in the shell's import chain (`packages/client/web/src/base.css`). It follows `design-platform.css` there because it reads that sheet's tokens.
+
+The rules sit on `body`, not `html`. `design-platform.css` declares the `--dsw-alias-*` tokens on `body`, with the dark overrides on `body[data-ds-dark-theme]`, and custom properties inherit only downward; an `html` rule resolves them to the guaranteed-invalid value, at which point `scrollbar-color` computes to `auto` and no theming happens at all.
+
+`scrollbar-width` and `scrollbar-color` are declared on `body, body *` rather than once at the top. Inheritance would pass down the color already substituted at `body`, so a descendant rebinding the indirection could not change its own scrollbar; re-declaring makes each element substitute the variable as it sees it. `scrollbar-width` is not an inherited property in the first place, so it needs the per-element declaration regardless. The `::-webkit-scrollbar*` pseudo-elements are likewise not inherited and are matched unscoped.
+
+The two renderings are mutually exclusive, and the exclusion is enforced rather than assumed. A non-`auto` `scrollbar-width` or `scrollbar-color` makes Chromium and Safari discard every `::-webkit-scrollbar*` rule for that element, `::-webkit-scrollbar-thumb:hover` included. Declaring both unconditionally therefore leaves the hover token rendering nowhere at all: the engines that implement the hover pseudo-element are exactly the ones the standard properties silence, and Firefox has no hover pseudo-element to fall back on. The standard properties consequently sit inside `@supports not selector(::-webkit-scrollbar)`, which is true only where the pseudo-element is unimplemented, so Firefox takes the standard path and WebKit-based engines take the pseudo-element path. The WebKit rules are not gated in turn: an engine without those pseudo-elements drops them as unknown selectors, so a gate would only restate what selector matching already does. An engine too old for the `selector()` function makes the condition invalid, which evaluates false and selects the pseudo-element path — the correct side for the pre-16.4 Safari that is the realistic case for that reading.
+
+Both paths read one indirection pair, `--dsh-scrollbar-thumb` and `--dsh-scrollbar-thumb-hover`, bound on `body` to the l1 (base-surface) tokens. **This is the rebinding contract, and it is the part the CSS alone does not state**: an elevated surface sets `--dsh-scrollbar-thumb: var(--dsw-alias-scrollbar-bg-l2)` and `--dsh-scrollbar-thumb-hover: var(--dsw-alias-scrollbar-hover-l2)` on its own container, and that one rebind reaches the standard properties and the WebKit pseudo-elements together. The pair is rebound as a pair; rebinding the resting thumb alone leaves the hover state on the base-surface token. Eight surfaces rebind today: the command popup, the slash menu, the model-select panel, the settings panel, the shared `ui-primitives` menu card, the composer input card, the question composer card, and the todo panel. Most declare it on the elevated card rather than on the scrolling descendant, because the elevation is a property of the surface and custom properties inherit down to whichever child actually scrolls.
+
+The last four were missed in the first implementation and found in review, which is why the rebinding contract is now checked mechanically rather than by inspection: a sheet that scrolls somewhere and paints an elevated surface somewhere must rebind.
+
+The elevated set is resolved from the palette's own dark elevation ladder — the surface tokens whose dark value lands on `bg-layer-2` or `bg-layer-3`, which is the step the l1/l2 split encodes. Deriving it instead from the sheets that already rebind was the first attempt and is unsound: such a set can only confirm what someone already remembered, and a surface nobody has rebound yet — exactly the case the check exists for — defines itself as unelevated. `--dsw-specific-tip` proved it, resolving to the menu surface's rung while the todo panel scrolled on it unrebound and the derived check stayed green.
+
+Scope is by token family, not by geometry: only `--dsw-alias-bg-*` and `--dsw-specific-*` name a surface. `--dsw-alias-button-*`, `--dsw-alias-interactive-*`, and `--dsw-alias-markdown-*` reach the same rungs while naming a control or an inline span that no scroll container renders its bar against. Shape cannot make that call, since a floating button legitimately carries a radius, a shadow, and a fixed size. The check is per sheet rather than per rule because the card and the descendant that scrolls are separate rules, and CSS text does not express which contains which.
+
+The track and the corner stay transparent, so the thumb reads against whatever surface scrolls under it; only the thumb and its hover state carry a token color.
+
+`.list` declares `scrollbar-gutter: stable`, which keeps the bar beside the rows instead of on top of them. `stable` rather than `auto` because `auto` reserves the gutter only while the list actually overflows: expanding a workspace group would then shift every row horizontally at the moment it starts scrolling. The reservation is unconditional and the rows never move.
+
+The gutter and the sheet's `::-webkit-scrollbar` width are jointly necessary against an overlay scrollbar, which is the configuration where the symptom exists at all. Measured on the running app by deleting each from the live cascade with the other left in force: either deletion alone takes the list's band from 8 to 0. The gutter states that space be reserved, and the pseudo-element width is what makes chromium treat the bar as occupying layout space rather than floating over the content. Neither half of this change is therefore optional for the reported bug, which is a second reason the two halves ship together.
+
+## Alternatives considered
+
+**Per-module `::-webkit-scrollbar` rules in each scrolling component sheet.** Rejected: the client has thirteen scrolling containers across nine packages, every one would carry the same block, and the fourteenth would ship unthemed with nothing failing. A skin driven by design tokens belongs in the package that owns the tokens.
+
+**An opt-in utility class that each scroll container adds.** Same duplication removed, but the failure mode stays: a new scroll container is themed only if its author remembers the class, and the omission is invisible in review. The `body, body *` form has no opt-in step to forget; a container that genuinely wants a different bar overrides the indirection, which is the same mechanism elevated surfaces use.
+
+**Bind the properties on `html`.** The natural place for a document-wide skin, and it fails measurably: with the rule on `html` a scroll container computes `scrollbar-color: auto` in chromium, because the alias tokens are not in scope there.
+
+**Declare the properties once and let them inherit.** Fewer matched elements, and it breaks the rebinding contract — inheritance carries the substituted color, not the variable reference, so an elevated surface could not retint its own scrollbar. It is also incomplete on its own terms, since `scrollbar-width` does not inherit.
+
+**Declare the standard properties and the pseudo-elements unconditionally, without the `@supports` gate.** This is what the change originally shipped, and review caught it. Measured in chromium on probe elements with `scrollbar-gutter: stable` so the band is observable: an 8px `::-webkit-scrollbar` alone reserved a 30px band (the sheet's width plus the UA's buttons), and adding `scrollbar-width: thin` to the same element dropped it to the 10px `thin` reserves — the pseudo-element rules were being discarded, not merged. Every `::-webkit-scrollbar-thumb:hover` rule went with them, so both hover tokens and all four elevated surfaces' hover rebinds were dead code on the engine most users run.
+
+**Gate the WebKit rules too, behind `@supports selector(::-webkit-scrollbar)`.** Symmetrical to read, and wrong in one direction: it would hide the rules from an engine that implements the pseudo-elements but not `selector()`, which is the pre-16.4 Safari the ungated form serves correctly. Unknown selectors are already dropped, so the gate adds no protection to pay for that.
+
+**Pad the rows instead of reserving the gutter (extra right padding on `.list`, or moving `.time` inward).** Rejected: padding applies whether or not a bar is present, so it costs horizontal room in the common short-list case, and it fixes exactly one container while leaving every other scrolling region's content under its bar.
+
+**`scrollbar-gutter: auto` on `.list`.** The reservation appears when the list overflows, which is when the bar exists. Rejected because the sidebar's lists grow and shrink as groups expand, so the reservation would appear and disappear under the user's cursor and shift the rows with it.
+
+## Consequences
+
+- Every scroll container in the client draws the themed thumb: `rgb(229, 229, 229)` on a light base surface, `rgb(60, 60, 61)` on a dark one, and `rgb(84, 85, 87)` for a dark elevated surface that rebinds to the l2 pair.
+- The two renderings are separately specified, so a change to the thumb's geometry or hover behavior has to be made twice — once in `scrollbar-width`/`scrollbar-color`, once in the pseudo-elements. Routing both through the indirection pair confines that duplication to the properties Firefox and WebKit do not share.
+- The hover tokens (`--dsw-alias-scrollbar-hover-l1`/`-l2`) render only on the pseudo-element path. Firefox states one thumb color through `scrollbar-color` and derives its own hover treatment, so a design change to the hover colors is visible in Chromium and Safari and not in Firefox. This is a limit of `scrollbar-color`, not of the sheet.
+- `body *` matches every element, for two properties whose effect the user agent already limits to elements that actually scroll. The cost is a broad selector; the alternative was a rebinding contract that does not work.
+- The workspace list is permanently narrower by the reserved band, at every list length. That is the trade the fix buys: stable row geometry instead of a timestamp that is legible only while the list is short.
+- There is no track token in the palette, so a design that later wants an opaque track needs a new alias token rather than a literal color in this sheet.
+
+## Testing
+
+Three unit specs read the CSS text on disk. `ui-theme/tests/scrollbar-styles.spec.ts` scans the scrollbar token set out of `design-platform.css` rather than hardcoding it, so adding, renaming, or dropping a token moves the assertions with it, and checks that every token has a consumer and that each elevated surface rebinds a complete pair. It also pins the path split by source offset: the standard properties inside the gate block, the `::-webkit-scrollbar*` rules and every read of the hover indirection outside it. That split needs an offset assertion because the spec's rule parser flattens through at-rules, so a gate deleted or a declaration moved across it leaves every other assertion in the file green.
+
+`apps/web/tests/sidebar-scrollbar.e2e.ts` covers the facts only a real engine reports: the reserved band width, and which rendering path the engine took. It needs no model calls — the list only has to overflow — so it seeds cold sessions from an existing committed fixture read-only.
+
+That scenario also commits a golden, `snapshots/sidebar-scrollbar/geometry.expected.md`, holding the resolved scrollbar style and geometry in both palettes. The aria goldens the other web scenarios commit cannot carry a CSS-only change: it alters no DOM and no accessible name, so their normalized trees are byte-identical with and without it. Recording the resolved values instead makes an unintended shift in thumb colour, band width, or rendering path a reviewable diff rather than a threshold someone has to reason about. Absolute coordinates are deliberately excluded — `timeRight` and the two edges depend on the sidebar's laid-out width and on font metrics, so committing them would produce a fixture that has to be re-recorded per platform and would document the platform rather than the change. What is recorded is the band, the overlap, and two orderings, each a difference or a comparison that survives any layout preserving the reservation.
+
+Confirmed in headless chromium on the built client by reading computed values, which is what distinguishes a working token chain from a syntactically valid one: a scroll container computes the l1 thumb color in each palette, and a container that rebinds the indirection computes the l2 color, proving the rebind reaches the computed value rather than only the custom property. Firefox was verified the same way for the standard path, including the l1-to-l2 rebind on `scrollbar-color`; headless Firefox reports `scrollbar-width: none` on every element, styled or not, which is a headless artifact rather than an effect of the sheet.
+
+Two chromium measurement limits shape what the e2e can assert. The gate makes chromium report `scrollbar-width` and `scrollbar-color` as `auto`, so the substituted `scrollbar-color` is no longer the observable — the e2e asserts the `auto` reading deliberately, since a concrete value there would mean the gate leaked and silenced the pseudo-elements. And `getComputedStyle(el, '::-webkit-scrollbar-thumb')` folds in the `::-webkit-scrollbar-thumb:hover` rule, so it reports the hover color at rest and pins neither state; proven by deleting the hover rule through `CSSStyleSheet.deleteRule` in the live page, which flipped that same query from the hover color to the resting one. The e2e therefore reads the resting and hover colors as the indirection variables resolve on the list — one throwaway probe element per variable, because `getComputedStyle` returns a live declaration and a reused probe reports only the last value read — and reads the hover declaration out of the cascade as rule text.
+
+The gate itself has a negative control at the level it operates on: removing the `@supports` wrapper from the sheet, rebuilding `build:web`, and rerunning the e2e turns the `scrollbar-width: auto` assertion red with `thin`, which is the suppression the gate exists to prevent.
+
+Headless chromium draws overlay scrollbars, and that is the configuration in which the reported symptom exists, so the e2e reproduces the bug rather than approximating it: against clean master the list's band is 0 and the bar covers 7px of the relative time. A reserved gutter there does not shrink `clientWidth`, so an assertion comparing the time element's right edge against the client-area edge holds with and without the reservation and would pass or fail on the platform's scrollbar style rather than on the declaration under test. The two signals that do separate the states are the `offsetWidth - clientWidth` band and `timeCoveredBy`, the overlap measured against the bar's own width.
+
+Both are asserted because each catches a different regression, established by mutating one declaration at a time with the other assertions in that test silenced. Removing only the gutter leaves `timeCoveredBy` at 0 — the bar is then 8px and the row's right padding is also 8px, so it abuts the timestamp without covering it — and the band assertion is what fails. Removing the pseudo-element width as well, which is the actual master state, produces the overlap, and `timeCoveredBy` fails at 7. A headed run under xvfb cannot show the symptom in either state, because chromium paints a classic space-consuming bar there and `clientWidth` already excludes it.
+
+Verifying browser-visible plugin CSS needs a rebuild `pnpm run build:web` does not perform. `WorkspaceBrowser.module.css` never reaches `apps/web/dist`: ui-workspace loads as a runtime plugin and its CSS is inlined into `packages/client/ui-workspace/lib/client.js`, built by that package's own `bundle` script. A negative control that reruns only `build:web` therefore exercises a stale bundle and passes with the declaration removed, which reads as a vacuous test rather than as an invalid control. Rebuild with `pnpm --filter @deepseek-ai/dsh-client-ui-workspace run bundle`, confirm the artifact by grepping `lib/client.js` for the declaration, then `build:web`.
+
+`test:web` ran `build:web` alone, so every scroll-region or plugin-CSS change hit that trap; it now runs `build` first, which covers `packages/*/*` and so rebuilds the plugin bundles. `check-all` already ordered `build` before `build:web`, so CI was never exposed — only the local script was, which is exactly where a stale-bundle pass is most likely to be believed.
@@ -0,0 +1,84 @@
+# Agent Note: 滚动条 token 有了消费方，工作区列表预留出滚动条空位
+
+Status: implemented
+
+[English](2026-07-28-themed-scrollbars-and-reserved-gutter.md) | 中文
+
+## 问题
+
+`design-platform.css` 在亮色与暗色两套调色板中都声明了四个 `--dsw-alias-scrollbar-*` token（`bg-l1`、`bg-l2`、`hover-l1`、`hover-l2`），而客户端里没有任何一条规则读取它们。定义了却无人消费的 token 构不成主题：所有滚动区域渲染的都是浏览器自带的滚动条，它对调色板一无所知，因此暗色主题下暗色表面上出现的是一条亮色的原生滚动条。
+
+暴露这一缺口的可见症状出在别处。工作区浏览器的会话列表（`WorkspaceBrowser.module.css` 中的 `.list`）是侧边栏里唯一的滚动区域，而每一行的尾部内容都紧贴该行 8px 的右内边距——`rows/Rows.module.css` 中的 `.time` 取 `flex: none`，hover 时取代它的操作按钮也是如此。于是覆盖式滚动条会画在相对时间戳之上。只在这一个列表里预留空间，滚动条本身仍然没有主题，因此两部分合为一次变更。
+
+## 决策
+
+`packages/client/ui-theme/src/styles/scrollbar.css` 是这四个 token 的唯一消费方，也是壳的导入链（`packages/client/web/src/base.css`）中第五张 ui-theme 样式表。它排在 `design-platform.css` 之后，因为它读取那张样式表的 token。
+
+规则挂在 `body` 上，而非 `html`。`design-platform.css` 在 `body` 上声明 `--dsw-alias-*` token，暗色覆盖挂在 `body[data-ds-dark-theme]` 上，而自定义属性只向下继承；挂在 `html` 上的规则会把它们解析为 guaranteed-invalid 值，此时 `scrollbar-color` 计算为 `auto`，主题完全不起作用。
+
+`scrollbar-width` 与 `scrollbar-color` 声明在 `body, body *` 上，而不是只在顶层声明一次。继承传下去的是已经在 `body` 处代入完成的颜色值，因此后代元素重新绑定这层间接变量也无法改变自己的滚动条；逐元素重新声明使每个元素按它自己看到的取值代入变量。`scrollbar-width` 本身就不是可继承属性，无论如何都需要逐元素声明。`::-webkit-scrollbar*` 伪元素同样不继承，因此以不加限定的选择器匹配。
+
+两种渲染互斥，而这种互斥是被强制的，不是假定的。`scrollbar-width` 或 `scrollbar-color` 只要取非 `auto` 值，Chromium 与 Safari 就会丢弃该元素上的全部 `::-webkit-scrollbar*` 规则，`::-webkit-scrollbar-thumb:hover` 也在其中。因此无条件地同时声明会让 hover token 在任何地方都得不到渲染：实现了 hover 伪元素的引擎，恰恰就是被标准属性静音的那些，而 Firefox 没有 hover 伪元素可作退路。于是标准属性写在 `@supports not selector(::-webkit-scrollbar)` 之内，该条件只在伪元素未被实现处为真，因此 Firefox 走标准属性路径，WebKit 系引擎走伪元素路径。WebKit 规则不再反向加门禁：不实现这些伪元素的引擎会把它们当作未知选择器丢弃，因此加门禁只是重述选择器匹配本身已经做的事。对于旧到不支持 `selector()` 函数的引擎，该条件无效，从而求值为假并选中伪元素路径——对于这条判断下现实存在的 16.4 之前的 Safari，这正是正确的一侧。
+
+两条路径都读取同一组间接变量 `--dsh-scrollbar-thumb` 与 `--dsh-scrollbar-thumb-hover`，它们在 `body` 上绑定到 l1（基础表面）token。**这就是重新绑定契约，也是单看 CSS 无法得知的部分**：抬升表面在自己的容器上设置 `--dsh-scrollbar-thumb: var(--dsw-alias-scrollbar-bg-l2)` 与 `--dsh-scrollbar-thumb-hover: var(--dsw-alias-scrollbar-hover-l2)`，这一次重新绑定同时作用于标准属性和 WebKit 伪元素。这组变量必须成对重新绑定；只改静止态滑块会让 hover 状态仍留在基础表面的 token 上。目前有八处抬升表面做了重新绑定：命令浮层、斜杠菜单、模型选择面板、设置面板、`ui-primitives` 共用菜单卡片、输入条卡片、提问组件卡片与待办面板。多数把声明写在抬升卡片上而非滚动的后代元素上，因为抬升层级是这个表面的属性，而自定义属性会继承到真正滚动的那个子元素。
+
+后四处在最初的实现里被漏掉、由评审发现，因此重新绑定契约现在由机械检查把关，而不再依赖人工审阅：一张样式表只要在某处滚动、又在某处绘制抬升表面，就必须重新绑定。
+
+抬升表面集合是从调色板自身的暗色抬升阶梯解析出来的——暗色取值落在 `bg-layer-2` 或 `bg-layer-3` 上的那些表面 token，而这一档正是 l1/l2 之分所编码的层级差。最初的做法是从已经做了重新绑定的样式表反向推导，那是不成立的：这样得到的集合只能确认别人已经记得的部分，而尚无人重新绑定的表面——恰恰就是这项检查存在的理由——会把自己定义成「非抬升」。`--dsw-specific-tip` 证明了这一点：它解析到与菜单表面相同的那一档，待办面板在它上面滚动却没有重新绑定，而推导式的检查依然是绿的。
+
+判定范围依据 token 家族而非几何形状：只有 `--dsw-alias-bg-*` 与 `--dsw-specific-*` 表述的是表面。`--dsw-alias-button-*`、`--dsw-alias-interactive-*` 与 `--dsw-alias-markdown-*` 会落到相同档位，但它们表述的是控件或行内片段，没有任何滚动容器会把滚动条画在它们之上。形状无法做这个判断，因为悬浮按钮本来就会带圆角、阴影和固定尺寸。这项检查以样式表为粒度而非以规则为粒度，因为卡片与真正滚动的后代元素是两条不同的规则，而 CSS 文本无法表达谁包含谁。
+
+轨道与两条滚动条相交的角落保持透明，因此滑块是以其下滚动的任何表面为背景被看到；只有滑块及其 hover 状态带 token 颜色。
+
+`.list` 声明 `scrollbar-gutter: stable`，使滚动条位于行的旁边而非行的上方。取 `stable` 而非 `auto`，因为 `auto` 只在列表确实溢出时才预留空位：那样展开一个工作区分组时，所有行会在列表开始滚动的那一刻发生水平位移。`stable` 的预留是无条件的，行不会移动。
+
+面对覆盖式滚动条——也就是这个症状唯一存在的那种形态——空位声明与样式表里的 `::-webkit-scrollbar` 宽度是共同必要的。在运行中的应用上实测：保留其中一条、从活的层叠中删掉另一条，任意一次单独删除都会让列表的条带从 8 降到 0。空位声明表述的是「要预留空间」，而伪元素宽度才是让 chromium 把滚动条视为占据布局空间、而不是浮在内容之上的原因。因此对这个 bug 而言，本次变更的两半都不是可选项，这也是两半必须一起交付的第二个理由。
+
+## 曾考虑的替代方案
+
+**在每个滚动组件的样式表里各写一份 `::-webkit-scrollbar` 规则。** 之所以否决：客户端共有分布在九个包中的十三个滚动容器，每一个都要带上同一段规则，而第十四个会在没有任何门禁报错的情况下漏掉主题。由设计 token 驱动的皮肤应当归属于拥有这些 token 的包。
+
+**提供一个工具类，由各滚动容器自行加上。** 重复同样被消除，但失败方式依旧存在：新的滚动容器只有在作者记得加类名时才有主题，而遗漏在评审中看不出来。`body, body *` 这种写法没有需要记住的启用步骤；确实想要不同滚动条的容器可以覆盖间接变量，这与抬升表面使用的机制相同。
+
+**把这两个属性绑定在 `html` 上。** 这是文档级皮肤最自然的落点，而它的失败是可测量的：规则挂在 `html` 上时，chromium 中滚动容器计算出的 `scrollbar-color` 为 `auto`，因为别名 token 在那个作用域内不存在。
+
+**只声明一次，靠继承下传。** 匹配的元素更少，但它破坏重新绑定契约——继承携带的是代入后的颜色，而不是变量引用，因此抬升表面无法给自己的滚动条换色。它本身也不完整，因为 `scrollbar-width` 不继承。
+
+**不加 `@supports` 门禁，无条件同时声明标准属性与伪元素。** 这正是本次变更最初提交的形态，被评审发现。在 chromium 中于带 `scrollbar-gutter: stable`（使条带可观测）的探针元素上实测：单独一条 8px 的 `::-webkit-scrollbar` 预留出 30px 条带（样式表指定的宽度加上浏览器自带的按钮），而给同一元素加上 `scrollbar-width: thin` 后降到 `thin` 所预留的 10px——说明伪元素规则是被丢弃，而不是被合并。全部 `::-webkit-scrollbar-thumb:hover` 规则随之失效，因此两个 hover token 与四处抬升表面的 hover 重新绑定，在多数用户实际使用的引擎上都是死代码。
+
+**给 WebKit 规则也加门禁，写成 `@supports selector(::-webkit-scrollbar)`。** 读起来对称，但在一个方向上是错的：它会对「实现了伪元素但不支持 `selector()`」的引擎隐藏这些规则，而那正是不加门禁时能被正确服务的 16.4 之前的 Safari。未知选择器本就会被丢弃，因此这道门禁不提供任何能抵偿该代价的保护。
+
+**改用内边距而不是预留空位（给 `.list` 加右内边距，或把 `.time` 向内移）。** 之所以否决：内边距无论滚动条是否存在都生效，因此在常见的短列表情形下白白占用横向空间；而且它只修好一个容器，其余每个滚动区域的内容仍然压在滚动条之下。
+
+**给 `.list` 用 `scrollbar-gutter: auto`。** 空位在列表溢出时出现，也就是滚动条存在的时候。之所以否决：侧边栏的列表会随分组展开与收起而伸缩，因此空位会在用户光标之下出现又消失，并带动行一起位移。
+
+## 后果
+
+- 客户端的每个滚动容器都绘制带主题的滑块：亮色基础表面为 `rgb(229, 229, 229)`，暗色基础表面为 `rgb(60, 60, 61)`，重新绑定到 l2 的暗色抬升表面为 `rgb(84, 85, 87)`。
+- 两种渲染分别指定，因此改动滑块的几何或 hover 行为需要改两处：一处在 `scrollbar-width`／`scrollbar-color`，一处在伪元素。让两者都经由这组间接变量，把这份重复限制在 Firefox 与 WebKit 不共用的那些属性上。
+- hover token（`--dsw-alias-scrollbar-hover-l1`／`-l2`）只在伪元素路径上渲染。Firefox 通过 `scrollbar-color` 只表述一个滑块颜色，其 hover 表现由引擎自行推导，因此对 hover 颜色的设计改动在 Chromium 与 Safari 上可见，在 Firefox 上不可见。这是 `scrollbar-color` 本身的限制，不是这张样式表的限制。
+- `body *` 匹配所有元素，涉及的两个属性其效果本就被浏览器限制在实际会滚动的元素上。代价是一个覆盖面很宽的选择器；另一种选择是一个不生效的重新绑定契约。
+- 工作区列表在任何列表长度下都永久少了预留空位那一条宽度。这正是该修复换来的代价：以稳定的行几何，换掉只在列表较短时才可读的时间戳。
+- 调色板中没有轨道 token，因此日后若设计需要不透明轨道，要新增一个别名 token，而不是在这张样式表里写字面颜色。
+
+## 测试
+
+三份单元测试读取磁盘上的 CSS 文本。`ui-theme/tests/scrollbar-styles.spec.ts` 从 `design-platform.css` 中扫描出滚动条 token 集合，而不是把它写死，因此新增、重命名或删除 token 时断言会随之变化；它检查每个 token 都有消费方，且每处抬升表面重新绑定的都是完整的一对。它还以源码偏移量锁定两条路径的划分：标准属性在门禁块之内，`::-webkit-scrollbar*` 规则与每一处对 hover 间接变量的读取都在门禁块之外。这个划分必须用偏移量断言，因为该测试文件的规则解析器会把 at-rule 拉平，所以删掉门禁或把某条声明移到门禁另一侧，文件里其余全部断言仍然是绿的。
+
+`apps/web/tests/sidebar-scrollbar.e2e.ts` 覆盖只有真实渲染引擎才能报告的事实：预留条带的宽度，以及引擎实际走的是哪条渲染路径。它不需要任何模型调用——列表只要溢出即可——因此以只读方式复用一份既有的已提交 fixture（测试前置数据）来铺入冷会话。
+
+这个场景还提交了一份 golden（期望产物）`snapshots/sidebar-scrollbar/geometry.expected.md`，记录两套调色板下解析后的滚动条样式与几何。其余 web 场景提交的 aria golden 承载不了纯 CSS 的改动：它不改变任何 DOM、也不改变任何无障碍名称，因此有无这次改动，它们规范化后的树都是逐字节相同的。改为记录解析后的取值，就让滑块颜色、条带宽度或渲染路径的意外变化成为可评审的 diff，而不是一条需要人去推敲的阈值断言。绝对坐标被特意排除——`timeRight` 与两条边缘取决于侧边栏排版后的宽度和字体度量，把它们提交进去会得到一份需要按平台重新录制的 fixture，那记录的是平台而不是这次改动。真正记录下来的是条带、重叠量与两个先后关系，每一项都是差值或比较，因此只要预留仍然成立，任何排版下都不变。
+
+在构建产物客户端上于 headless chromium 中读取计算值确认，这正是区分「token 链真正生效」与「语法合法」的手段：滚动容器在两套调色板下分别计算出 l1 的滑块颜色，而重新绑定间接变量的容器计算出 l2 的颜色，证明重新绑定作用到了计算值，而不只是作用到自定义属性上。Firefox 的标准属性路径以同样方式做了验证，包含 `scrollbar-color` 上从 l1 到 l2 的重新绑定；headless Firefox 对任何元素（无论是否被样式命中）都报告 `scrollbar-width: none`，这是 headless 的产物，不是这张样式表造成的。
+
+chromium 上有两处测量限制决定了 e2e 能断言什么。门禁使 chromium 报告的 `scrollbar-width` 与 `scrollbar-color` 都是 `auto`，因此代入后的 `scrollbar-color` 不再是可观测量——e2e 特意断言这个 `auto` 读数，因为此处出现具体值就意味着门禁泄漏、伪元素被静音。另外，`getComputedStyle(el, '::-webkit-scrollbar-thumb')` 会把 `::-webkit-scrollbar-thumb:hover` 规则一并折算进去，因此它在静止态就报告 hover 颜色，两种状态都锁不住；这一点由在运行中的页面里用 `CSSStyleSheet.deleteRule` 删掉 hover 规则得证——同一查询随之从 hover 颜色翻转为静止态颜色。因此 e2e 改为读取那组间接变量在列表上代入后的静止态与 hover 颜色（每个变量用一个一次性探针元素，因为 `getComputedStyle` 返回的是活的声明对象，复用探针只会报告最后一次读到的值），并把 hover 声明当作规则文本从层叠中读出。
+
+门禁本身在它起作用的层面有反向对照：把样式表中的 `@supports` 包裹去掉、重新 `build:web`、再跑 e2e，`scrollbar-width: auto` 那条断言会以 `thin` 变红，而这正是门禁存在所要阻止的那种静音。
+
+headless chromium 绘制的是覆盖式滚动条，而这恰好就是被报告症状存在的那种形态，因此这个 e2e 复现的是这个 bug 本身，而不是它的近似：在干净的 master 上，列表条带为 0，滚动条盖住相对时间 7px。其中预留空位不会缩小 `clientWidth`，因此把时间元素右边缘与内容区右边缘做比较的断言在有无预留的两种状态下都成立，它的通过或失败取决于平台的滚动条样式，而不是取决于被测的那条声明。真正能区分两种状态的两个量是 `offsetWidth - clientWidth` 条带，以及以滚动条自身宽度为基准量出的重叠量 `timeCoveredBy`。
+
+两者都要断言，因为各自捕捉的是不同的回归；这一点通过每次只改动一条声明、并把同一个测试里的其余断言静音来确定。只删掉空位声明时 `timeCoveredBy` 仍为 0——此时滚动条是 8px，而行的右内边距也是 8px，于是它紧贴时间戳但并未盖住——失败的是条带那条断言。再把伪元素宽度也删掉（这才是 master 的真实状态）才会产生重叠，此时 `timeCoveredBy` 以 7 变红。在 xvfb 下的有头运行无论哪种状态都看不到这个症状，因为 chromium 在那里画的是经典占位滚动条，`clientWidth` 本来就已经把它排除了。
+
+验证浏览器可见的插件 CSS 需要一次 `pnpm run build:web` 并不执行的重建。`WorkspaceBrowser.module.css` 从不进入 `apps/web/dist`：ui-workspace 以运行时插件方式加载，其 CSS 内联进 `packages/client/ui-workspace/lib/client.js`，由该包自己的 `bundle` 脚本构建。因此只重跑 `build:web` 的反向对照实际测的是旧产物，去掉声明后仍会通过，看起来像测试无效，实际是对照无效。正确做法是先 `pnpm --filter @deepseek-ai/dsh-client-ui-workspace run bundle`，用 grep 在 `lib/client.js` 中确认该声明确实存在或消失，然后再 `build:web`。
+
+`test:web` 原先只运行 `build:web`，因此任何滚动区域或插件 CSS 的改动都会碰到这个陷阱；现在它先运行 `build`，而 `build` 覆盖 `packages/*/*`，从而会重建各插件产物。`check-all` 本来就把 `build` 排在 `build:web` 之前，所以 CI 从未受影响——受影响的只有本地脚本，而这恰恰是「产物过期却通过」最容易被当真的地方。
@@ -0,0 +1,6 @@
+# Bilingual-pair consistency record (docs/i18n/README.md): the git blob hash of each
+# side as of the last confirmed-consistent state. Both languages carry equal authority;
+# after editing either side, bring the other along and re-record with:
+#   pnpm run verify-translation-pairing --write .agents/notes/implemented/bug-fix/2026-07-29-pnpm-setup-runner-isolation.md
+2026-07-29-pnpm-setup-runner-isolation.md: 743535d0394cbea0374c412ba6968910ce858de4
+2026-07-29-pnpm-setup-runner-isolation.zh.md: 1e51070f88dead17b9d3f5625e337c558786aba2
@@ -0,0 +1,27 @@
+# Agent Note: Isolate pnpm setup per GitHub Actions runner
+
+Status: implemented
+
+English | [中文](2026-07-29-pnpm-setup-runner-isolation.zh.md)
+
+## Problem
+
+`pnpm/action-setup@v4` defaults its install destination to `~/setup-pnpm` and replaces that directory during setup. The self-hosted CI failover runs six GitHub Actions runner services under one VM user, so concurrent jobs shared the same destination. In [run 30375670773](https://github.com/deepseek-harness/deepseek-harness/actions/runs/30375670773), three jobs entered pnpm setup within 73 milliseconds; one setup removed another process's current working directory and two jobs failed in Node's `uv_cwd` initialization. A retry on another runner passed, making the failure timing-dependent rather than a repository-test regression.
+
+## Decision
+
+Every `pnpm/action-setup` step in [the primary CI workflow](../../../../.github/workflows/ci.yml) sets `dest: ${{ runner.temp }}/setup-pnpm`. Each runner service owns its temporary directory, so one setup cannot replace another runner's install directory. Persistent store reuse remains separate through `PNPM_CONFIG_STORE_DIR`, as established by the [pnpm provisioning decision](../process/2026-07-26-pnpm-action-setup-for-symmetric-ci-caching.md).
+
+[The workflow regression test](../../../../scripts/ci-workflow.spec.ts) discovers every `pnpm/action-setup` step in `ci.yml` and rejects one without the runner-private destination. This keeps newly added jobs inside the same isolation boundary.
+
+## Alternatives considered
+
+**Serialize failover jobs.** Rejected because it discards the six-runner pool's intended parallelism and turns an action-local directory collision into queueing across otherwise independent jobs.
+
+**Assign a separate Unix user to every runner service.** This would also separate `HOME`, but it moves the invariant into external VM provisioning and complicates ownership of the deliberately shared persistent pnpm store. The workflow already receives a runner-private temporary directory.
+
+**Retry failed setup steps.** Rejected because retries only reduce the observed collision rate; another concurrent setup can remove the same shared directory again.
+
+## Consequences
+
+pnpm's executable installation is ephemeral and isolated per runner, while package downloads still use the configured persistent or cached store. Hosted jobs use the same explicit destination without changing cache policy. The workflow carries three extra configuration lines per setup step, and the regression test must be updated only if pnpm provisioning intentionally moves to a different isolation mechanism.
@@ -0,0 +1,27 @@
+# Agent Note: 按 GitHub Actions runner 隔离 pnpm 设置
+
+Status: implemented
+
+[English](2026-07-29-pnpm-setup-runner-isolation.md) | 中文
+
+## 问题
+
+`pnpm/action-setup@v4` 的安装目标目录默认为 `~/setup-pnpm`，并会在设置期间替换该目录。自托管 CI 故障切换在同一个 VM 用户下运行六个 GitHub Actions runner 服务，因此并发作业会共用同一目标目录。在 [run 30375670773](https://github.com/deepseek-harness/deepseek-harness/actions/runs/30375670773) 中，三个作业在 73 毫秒内进入 pnpm 设置；其中一个设置过程删除了另一个进程的当前工作目录，导致两个作业在 Node 的 `uv_cwd` 初始化阶段失败。换到另一台 runner 重试后通过，说明该故障取决于时序，并非仓库测试回归。
+
+## 决策
+
+[主 CI 工作流](../../../../.github/workflows/ci.yml)中的每个 `pnpm/action-setup` 步骤都设置 `dest: ${{ runner.temp }}/setup-pnpm`。每个 runner 服务独占自己的临时目录，因此一个设置过程无法替换另一个 runner 的安装目录。持久 store 的复用仍由 `PNPM_CONFIG_STORE_DIR` 独立处理，遵循 [pnpm 提供机制决策](../process/2026-07-26-pnpm-action-setup-for-symmetric-ci-caching.md)。
+
+[工作流回归测试](../../../../scripts/ci-workflow.spec.ts)会找出 `ci.yml` 中的每个 `pnpm/action-setup` 步骤，并拒绝缺少 runner 专属目标目录的步骤。这可确保后续新增的作业也处于同一隔离边界内。
+
+## 曾考虑的替代方案
+
+**串行执行故障切换作业。** 否决：这会牺牲由六个 runner 组成的池所具备的预期并行能力，并把 action 内部的目录冲突变成原本相互独立作业之间的队列等待。
+
+**为每个 runner 服务分配独立的 Unix 用户。** 这同样能够隔离 `HOME`，但会把该不变量转移到外部 VM 配置中，并使刻意共享的持久 pnpm store 的所有权变得复杂。工作流已经获得 runner 专属临时目录。
+
+**重试失败的设置步骤。** 否决：重试只能降低观测到的冲突发生率；另一个并发设置过程仍可能再次删除同一个共享目录。
+
+## 后果
+
+pnpm 可执行文件采用临时安装，并按 runner 隔离；包下载仍使用已配置的持久或缓存 store。托管作业使用相同的显式目标目录，不改变缓存政策。工作流中的每个设置步骤因此增加三行配置；只有在 pnpm 提供机制有意迁移到另一种隔离机制时，才需要更新回归测试。
@@ -0,0 +1,6 @@
+# Bilingual-pair consistency record (docs/i18n/README.md): the git blob hash of each
+# side as of the last confirmed-consistent state. Both languages carry equal authority;
+# after editing either side, bring the other along and re-record with:
+#   pnpm run verify-translation-pairing --write .agents/notes/implemented/bug-fix/2026-07-29-sticky-composer-conversation-scroll.md
+2026-07-29-sticky-composer-conversation-scroll.md: 7ceae95dafffdb756ef49bb5612cd4e711eb59ca
+2026-07-29-sticky-composer-conversation-scroll.zh.md: d925d82f94635b5fe67b0be119c041d003def393
@@ -0,0 +1,29 @@
+# Agent Note: Fixed header, sticky composer inside the transcript scrollport
+
+Status: implemented
+
+English | [中文](2026-07-29-sticky-composer-conversation-scroll.zh.md)
+
+## Problem
+
+The active conversation column split scrolling: the chat (and trajectory) view owned `overflow-y: auto`, while the composer stack sat as a sibling below that scrollport. A wheel gesture over the stats line or input therefore hit a non-scrolling region and did nothing — the transcript only moved when the pointer was over the message list. Long drafts made it worse: the textarea is itself a scrollport, so wheel over the composer could be trapped there. The session header must occupy the top of the column as ordinary chrome (not `position: sticky` inside the scrollport), while the composer must stick to the bottom of the same scrollport as the transcript so wheel over the footer moves the flow.
+
+## Decision
+
+While a session exists, `ConversationRoot` always supplies a `wrapActiveBody` owner callback that wraps the view ring in a `data-conversation-scroll` body and places a `data-composer-seat` around the whole `'conversation.composer'` chain output (fallback + elected overlay siblings from `overlay: true`). Active CSS sticks that seat with `position: sticky; bottom: 0` so Question/Approval takeovers stay visible when the user is not pinned to the floor; hero CSS centers the fallback stack inside the scroll body. `ConversationSession` keeps a chrome-hidden header + body shell while blank so that tree seat does not change on the first send. The session header remains `flex: none` column chrome above the scrollport when visible. ChatView and Trajectory/Waterfall keep a local scroller only when mounted outside that host (unit tests); under the host they set `overflow: visible` and resolve bottom-follow / prepend anchoring through `closest('[data-conversation-scroll]')`.
+
+Session stats live on `'conversation.composer.dock'` (above `'conversation.input.dock'`). The InputBar textarea, when inside the host, chains `wheel` with `{ passive: false }`: while the capped textarea can still scroll in that direction it keeps the native gesture; only at its own edge does it `preventDefault` and apply `deltaY` to the host.
+
+## Alternatives considered
+
+**Sticky header and sticky composer inside one column scrollport.** Rejected for the header: it must occupy the top as fixed layout chrome, not participate in the scrollport's sticky layer.
+
+**Fixed flex-none composer below the scrollport with wheel forwarding.** Rejected: the product requires the composer to stick inside the transcript scrollport so the footer is part of that scroll hit-testing surface, not a sibling that only forwards deltas.
+
+**Portal the composer into ChatView's scroller.** Rejected: the composer is shared across view tabs; the wrap target is the Session body owned by the resident shell.
+
+**Keep StatsLine inside ChatView below the message column.** Rejected: outside the sticky composer it would scroll away while the input stayed pinned.
+
+## Consequences
+
+Wheel over the footer scrolls the transcript; the visible layout is a fixed header, scrolling transcript, and sticky bottom composer. Stats appear on every active view tab. Nested view scrollers under the host are suppressed so sticky Turn headers in Trajectory stick to the column host. Hero → active keeps the same textarea DOM node (assembled slash-flow snapshot) and the InputHub draft.
@@ -0,0 +1,29 @@
+# Agent Note: 固定标题栏，sticky 编辑器位于 transcript 滚动容器内
+
+Status: implemented
+
+[English](2026-07-29-sticky-composer-conversation-scroll.md) | 中文
+
+## Problem
+
+活跃会话列把滚动拆成两段：聊天（以及 trajectory）视图自有 `overflow-y: auto`，编辑器栈则作为该滚动容器的兄弟节点坐在下方。指针落在统计行或输入区上时，滚轮打在不可滚动区域上因而毫无效果——只有指针在消息列表上时 transcript 才会移动。草稿变长时更糟：textarea 本身也是滚动容器，编辑器上的滚轮可能被截在那里。会话标题栏必须以普通 chrome 占据列顶（不能在滚动容器内 `position: sticky`），而编辑器必须与 transcript 贴在同一滚动容器底部，使页脚上的滚轮能带动内容流动。
+
+## Decision
+
+只要存在会话，`ConversationRoot` 就会始终提供 `wrapActiveBody` owner 回调，将视图环包进 `data-conversation-scroll` 主体，并用 `data-composer-seat` 包住整条 `'conversation.composer'` chain 输出（`overlay: true` 下的 fallback 与选举出的 overlay 兄弟节点）。活跃阶段 CSS 以 `position: sticky; bottom: 0` 钉住该 seat，使用户未贴底时 Question／Approval 接管仍可见；hero CSS 在滚动主体内居中 fallback 栈。`ConversationSession` 在 blank 时保留隐藏 chrome 的 header + body 壳，使首次发送时树座位不变。可见时会话标题栏仍是滚动容器之上的 `flex: none` 列 chrome。ChatView 与 Trajectory/Waterfall 仅在宿主之外挂载时（单元测试）保留本地 scroller；位于宿主下时设为 `overflow: visible`，并通过 `closest('[data-conversation-scroll]')` 解析贴底跟随与前置锚定。
+
+会话统计挂在 `'conversation.composer.dock'`（位于 `'conversation.input.dock'` 之上）。InputBar 的 textarea 在宿主内以 `{ passive: false }` 链式处理 `wheel`：在限高 textarea 仍能沿该方向滚动时保留原生手势；仅在自身边缘才 `preventDefault` 并将 `deltaY` 施加到宿主。
+
+## Alternatives considered
+
+**标题栏与编辑器都在同一列滚动容器内 sticky。** 标题栏否决：它必须作为固定布局 chrome 占据顶部，而不是参与滚动容器的 sticky 层。
+
+**滚动容器下方 flex-none 固定编辑器并转发滚轮。** 否决：产品要求编辑器 sticky 在 transcript 滚动容器内，使页脚成为该滚动命中面的一部分，而不是仅转发增量的兄弟节点。
+
+**把编辑器 portal 进 ChatView 的 scroller。** 否决：编辑器跨视图标签共享；包装目标是常驻壳拥有的 Session 主体。
+
+**把 StatsLine 留在 ChatView 消息列下方。** 否决：落在 sticky 编辑器之外会随内容滚走，而输入区仍钉在底部。
+
+## Consequences
+
+在页脚上滚轮会滚动 transcript；可见布局是固定标题栏、可滚动 transcript 与 sticky 底部编辑器。统计出现在每一个活跃视图标签上。宿主下的嵌套视图 scroller 被抑制，因而 Trajectory 的 sticky Turn 标题贴在列宿主上。hero → active 保持同一 textarea DOM 节点（assembled slash-flow 快照）以及 InputHub 草稿。
@@ -0,0 +1,6 @@
+# Bilingual-pair consistency record (docs/i18n/README.md): the git blob hash of each
+# side as of the last confirmed-consistent state. Both languages carry equal authority;
+# after editing either side, bring the other along and re-record with:
+#   pnpm run verify-translation-pairing --write .agents/notes/implemented/bug-fix/2026-07-29-web-details-session-lifecycle.md
+2026-07-29-web-details-session-lifecycle.md: d9e0255768f165bed0631b9324e971b57ec7dcae
+2026-07-29-web-details-session-lifecycle.zh.md: 09452ba80ff240ddca76df239b40ea661566f8e2
@@ -0,0 +1,29 @@
+# Agent Note: Web details follow the current Session lifecycle
+
+Status: implemented
+
+English | [中文](2026-07-29-web-details-session-lifecycle.zh.md)
+
+## Problem
+
+The details entry is Session-scoped, but its preferred grid width is root-scoped. Selecting a different Session replaced the details content without closing that root preference, so the new owner inherited stale viewing geometry. Hero and other unselected states render no Session-scoped details; they need a derived zero track without becoming false owners in the comparison.
+
+## Decision
+
+`AppFrame` reads the current Session id and its `blank` summary flag from the authoritative Session projection. It records the last non-blank selected id only when that Session can own details, so hero and other unselected states neither trigger closure nor replace the last Session owner; their rendered details track derives as zero without changing the stored preference. The first Session keeps the default details width; returning to the same Session restores its current width; selecting a different Session closes the root-scoped details preference through the layout store before paint. The per-Session chat selection remains owned by the session-scoped store described by the [slot system standard](../architecture/2026-07-22-slot-type-chain-implementation.md).
+
+The layout store is transient and starts details at its default width. It neither reads nor writes `localStorage`, so reload resets both panel widths and needs no Session-baseline exception. Manual close and reopen inside one unchanged Session retain their existing behavior. The lifecycle effect changes neither the [Workspace-owned New Session flow](../feature/2026-07-25-workspace-ui-product-flow.md), composer drafts, Session navigation, nor concession-chain resizing.
+
+## Alternatives considered
+
+**Close details in the New Session click handler.** Rejected because an unselected surface has no Session-scoped details and must not mutate geometry. Closure belongs to the later comparison between two defined Session owners.
+
+**Persist panel geometry per Session.** Rejected because the product contract needs stale context removed, not a new map of remembered widths. Per-Session geometry would also reopen details when users return, contrary to the chosen close-on-leave behavior.
+
+**Preserve persisted layout after the Session baseline is ready.** Rejected because it duplicates startup lifecycle in a presentation component solely to validate stale viewing state. Transient defaults make reload deterministic without a readiness flag.
+
+**Treat every current-projection change as a Session switch.** Rejected because startup materialization, hero, clearing selection, and invalidation are not transitions between two Session owners.
+
+## Consequences
+
+Details is open by default, including when the first Session materializes. Switching to a different Session forgets the dragged details width because close writes zero and reopen uses the contract default. Unselected states derive a zero rendered track while leaving the preferred geometry unchanged; returning to the same Session through one of those states restores its width. Reload forgets sidebar and details geometry. The layout behavior test covers initial defaults, first materialization, direct and hero-mediated Session switches, same-Session return, and the absence of layout storage; the keyless browser e2e drives the same owner transitions through the shipped composition while checking the full grid track and browser errors.
@@ -0,0 +1,29 @@
+# Agent Note: Web 详情栏遵循当前会话生命周期
+
+Status: implemented
+
+[English](2026-07-29-web-details-session-lifecycle.md) | 中文
+
+## 问题
+
+详情入口由会话作用域拥有，而其首选网格宽度由根作用域拥有。选择不同会话时，系统会替换详情内容，却不会关闭根作用域的该首选宽度，因此新 owner 会继承陈旧的查看几何信息。hero 和其他未选中状态不会渲染会话作用域的详情；其轨道需派生为零宽度，但不能因此在比较中成为伪 owner。
+
+## 决策
+
+`AppFrame` 从权威会话投影读取当前会话 id 及其摘要中的 `blank` 标志。它只在该会话能够拥有详情时记录最后一个选中的非 blank 会话 id，因此 hero 和其他未选中状态既不会触发关闭，也不会替换最后一个会话 owner；这些状态下，详情栏轨道的渲染宽度派生为零，但存储的首选宽度不变。首个会话保留详情栏的默认宽度；返回同一会话时恢复其当前宽度；选择不同会话时，系统会先通过布局 store 关闭根作用域存储的详情栏首选宽度，再进行绘制。逐会话的聊天选中项继续由 [slot 体系标准](../architecture/2026-07-22-slot-type-chain-implementation.md)所述的会话作用域 store 拥有。
+
+布局 store 是瞬时状态，详情栏以默认宽度启动。它既不读取也不写入 `localStorage`，因此重新加载会重置两个面板的宽度，无需会话基线例外。在同一个未变化的会话内手动关闭和重新打开详情栏，仍保持原有行为。该生命周期 effect 不改变 [Workspace 拥有的 New Session 动线](../feature/2026-07-25-workspace-ui-product-flow.md)、composer 草稿、会话导航或让步链缩放。
+
+## 考虑过的替代方案
+
+**在 New Session 点击处理器中关闭详情栏。** 之所以否决：未选中表面没有会话作用域的详情，不得修改几何信息。详情栏是否关闭，应由随后对两个已定义会话 owner 的比较决定。
+
+**按会话持久化面板几何信息。** 之所以否决：产品契约需要移除陈旧上下文，而不是新增一张保存各宽度的映射。按会话保存几何信息还会在用户返回时重新打开详情栏，与选定的离开即关闭行为相悖。
+
+**在会话基线就绪后保留持久化布局。** 之所以否决：这会仅为验证陈旧的查看状态，在呈现组件中重复实现启动生命周期。瞬时默认值无需就绪标志即可使重新加载具有确定性。
+
+**将当前投影的每次变化都视为会话切换。** 之所以否决：启动时的物化、hero、清除选中项和选中状态失效都不是两个会话 owner 之间的过渡。
+
+## 后果
+
+详情栏默认打开，首次会话物化时亦然。切换到不同会话会忘记拖动后的详情宽度，因为关闭操作会写入零值，重新打开时则使用契约默认值。未选中状态会将轨道的渲染宽度派生为零，同时保持首选几何信息不变；经由这些状态返回同一会话时，会恢复其宽度。重新加载会忘记侧边栏与详情栏的几何信息。布局行为测试覆盖初始默认值、首次物化、直接及经 hero 中转的会话切换、返回同一会话，以及不存在布局存储的情况；无密钥浏览器 e2e 则通过已交付的组合驱动相同的 owner 过渡，同时检查完整网格轨道和浏览器错误。
@@ -1,6 +1,6 @@
 # Bilingual-pair consistency record (docs/i18n/README.md): the git blob hash of each
 # side as of the last confirmed-consistent state. Both languages carry equal authority;
 # after editing either side, bring the other along and re-record with:
-#   pnpm run verify-translation-pairing --write
-2026-06-22-acp-subagent-backend.md: a45ce5e34873249969bbe4dabb87a89d10246b3d
-2026-06-22-acp-subagent-backend.zh.md: 3b6a11efc1ae0427eb5ce3b30029b77b00d6801a
+#   pnpm run verify-translation-pairing --write .agents/notes/implemented/feature/2026-06-22-acp-subagent-backend.md
+2026-06-22-acp-subagent-backend.md: 5f12aa1c08d4f4cfaa35f2f7f4b09ad341c3eae8
+2026-06-22-acp-subagent-backend.zh.md: 61359246b0c552f6126cd82ae843d489de809a38
@@ -34,7 +34,7 @@ ACP `StopReason` → harness `SubagentStopReason`: `end_turn`→`completed`, `ma

 ### Security: scrubbed child environment

-The child is a separate process, so it inherits an environment. Credential-shaped ambient vars (`/KEY|SECRET|TOKEN/i`) are NOT forwarded by default — the parent harness's own secrets must not leak into a spawned process implicitly (the same policy the bash executor applies). The child's OWN credentials (it needs a model key) are supplied EXPLICITLY via `config.env`, layered AFTER the scrub, so an intended `DEEPSEEK_API_KEY` survives while an incidental `AWS_SECRET_ACCESS_KEY` does not. Child stderr is inherited to the parent's stderr (diagnostics surface naturally); a spawn-level `error` event (e.g. ENOENT for a bad command) is captured and raced against the ACP drive, so a bad command settles `error` instead of crashing the parent with an unhandled error.
+The child is a separate process, so it inherits an environment. Credential-shaped ambient vars (`/KEY|PASSWORD|SECRET|TOKEN/i`) are NOT forwarded by default — the parent harness's own secrets must not leak into a spawned process implicitly (the same policy the bash executor applies). The child's OWN credentials (it needs a model key) are supplied EXPLICITLY via `config.env`, layered AFTER the scrub, so an intended `DEEPSEEK_API_KEY` survives while an incidental `AWS_SECRET_ACCESS_KEY` does not. Child stderr is inherited to the parent's stderr (diagnostics surface naturally); a spawn-level `error` event (e.g. ENOENT for a bad command) is captured and raced against the ACP drive, so a bad command settles `error` instead of crashing the parent with an unhandled error.

 ## Testing

@@ -34,7 +34,7 @@ ACP `StopReason` → harness `SubagentStopReason`：`end_turn`→`completed`、`

 ### 安全：清洗子进程环境

-子 agent 是独立进程，因此会继承环境变量。形如凭证的环境变量（`/KEY|SECRET|TOKEN/i`）默认不转发——父 harness 自身的密钥不得隐式泄露到派生进程中（与 bash 执行器采用的策略相同）。子 agent 自己的凭证（它需要模型密钥）通过 `config.env` 显式提供，在清洗之后叠加，因此有意传入的 `DEEPSEEK_API_KEY` 得以保留，而偶然存在的 `AWS_SECRET_ACCESS_KEY` 则不会。子进程的 stderr 继承到父进程的 stderr（诊断信息自然浮现）；spawn 级别的 `error` 事件（如命令不存在时的 ENOENT）被捕获并与 ACP 驱动竞速，因此错误命令解析为 `error` 而非以未处理错误崩溃父进程。
+子 agent 是独立进程，因此会继承环境变量。形如凭证的环境变量（`/KEY|PASSWORD|SECRET|TOKEN/i`）默认不转发——父 harness 自身的密钥不得隐式泄露到派生进程中（与 bash 执行器采用的策略相同）。子 agent 自己的凭证（它需要模型密钥）通过 `config.env` 显式提供，在清洗之后叠加，因此有意传入的 `DEEPSEEK_API_KEY` 得以保留，而偶然存在的 `AWS_SECRET_ACCESS_KEY` 则不会。子进程的 stderr 继承到父进程的 stderr（诊断信息自然浮现）；spawn 级别的 `error` 事件（如命令不存在时的 ENOENT）被捕获并与 ACP 驱动竞速，因此错误命令解析为 `error` 而非以未处理错误崩溃父进程。

 ## 测试

@@ -1,6 +1,6 @@
 # Bilingual-pair consistency record (docs/i18n/README.md): the git blob hash of each
 # side as of the last confirmed-consistent state. Both languages carry equal authority;
 # after editing either side, bring the other along and re-record with:
-#   pnpm run verify-translation-pairing --write
-2026-06-29-todo-write-tool.md: 760373d64f462e3717a174d5793e6d47ab76b0b4
-2026-06-29-todo-write-tool.zh.md: fd29e6049e6a729c2c7e78860ad7c065422da60e
+#   pnpm run verify-translation-pairing --write .agents/notes/implemented/feature/2026-06-29-todo-write-tool.md
+2026-06-29-todo-write-tool.md: 288932f641a37c13ea6beeb069ac360c4a8447c1
+2026-06-29-todo-write-tool.zh.md: 03dae6328e5c7f4379694ab01db3434b4469826c
@@ -18,7 +18,7 @@ The model sends the entire list every call; the new list replaces the old (last-

 ### State on the session log, not a service

-The list is appended as a `todo/write` event carrying the full `{ todos }` snapshot. The harness is event-sourced — the LLM history, tool calls, and turn structure all live on the log — so the todo list lives there too. This buys durability, replay, and resume reconstruction for free: a reopened session re-derives the current list from the latest `todo/write`, with no separate persistence backend, in-memory service to rehydrate, or extra wiring. An in-memory `ctx.todos` service would have to reinvent all of that. (Full-log consumers get this reconstruction outright; the web client's paged window gets it from the tail history page's host-computed projection — see the [web todo display note](2026-07-23-web-todo-display.md).)
+The list is appended as a `todo/write` event carrying the full `{ todos }` snapshot. The harness is event-sourced — the LLM history, tool calls, and turn structure all live on the log — so the todo list lives there too. This buys durability, replay, and resume reconstruction for free: a reopened session re-derives the standing plan from the latest `todo/write` that is not followed by a later `turn/start` ([plan strip lifetime](2026-07-28-todo-plan-clears-on-next-turn.md)), with no separate persistence backend, in-memory service to rehydrate, or extra wiring. An in-memory `ctx.todos` service would have to reinvent all of that. (Full-log consumers get this reconstruction outright; the web client's paged window gets it from the tail history page's host-computed projection — see the [web todo display note](2026-07-23-web-todo-display.md).)

 ### NOT a surface event

@@ -18,7 +18,7 @@ harness 为模型提供了 bash 和 subagent 工具，却没有办法记录结

 ### 状态在会话日志上，而非服务

-列表作为 `todo/write` 事件追加到日志，携带完整的 `{ todos }` 快照。harness 是事件溯源的——LLM（大语言模型）历史、工具调用和轮次结构都在日志上——所以 todo 列表也在那里。这免费获得了持久性、回放和恢复重建：重新打开的会话从最新的 `todo/write` 重新推导当前列表，无需独立的持久化后端、无需重新注水的内存服务、无需额外接线。一个内存中的 `ctx.todos` 服务需要重新发明以上所有。（全量 log 消费者直接获得这份重建；web 客户端的分页窗口则从尾页 history 携带的 host 计算投影获得——见 [web todo 展示 Note](2026-07-23-web-todo-display.md)。）
+列表作为 `todo/write` 事件追加到日志，携带完整的 `{ todos }` 快照。harness 是事件溯源的——LLM（大语言模型）历史、工具调用和轮次结构都在日志上——所以 todo 列表也在那里。这免费获得了持久性、回放和恢复重建：重新打开的会话从「其后没有更晚 `turn/start`」的最近一次 `todo/write` 重新推导站立计划（[计划条生命周期](2026-07-28-todo-plan-clears-on-next-turn.md)），无需独立的持久化后端、无需重新注水的内存服务、无需额外接线。一个内存中的 `ctx.todos` 服务需要重新发明以上所有。（全量 log 消费者直接获得这份重建；web 客户端的分页窗口则从尾页 history 携带的 host 计算投影获得——见 [web todo 展示 Note](2026-07-23-web-todo-display.md)。）

 ### 不是 surface 事件

@@ -2,5 +2,5 @@
 # side as of the last confirmed-consistent state. Both languages carry equal authority;
 # after editing either side, bring the other along and re-record with:
 #   pnpm run verify-translation-pairing --write .agents/notes/implemented/feature/2026-07-05-skill-system.md
-2026-07-05-skill-system.md: 4fc621a9fdfa8042ebf3eb1975f0930cb1bf116c
-2026-07-05-skill-system.zh.md: 0dbebd211a1fa9e434d3f0a189c936f2b1c76574
+2026-07-05-skill-system.md: 242650ec8ba64fd0801a958711d5790fae07b259
+2026-07-05-skill-system.zh.md: da5f4af4b2f8bc0144be0a7ed608de7edd9a9947
@@ -18,13 +18,13 @@ Provider plugins register synchronously during `apply()`. Provider membership is

 The local provider scans cwd-sensitive project roots, custom roots, and user roots in first-wins rank order: project `.dsh`, project `.agents`, `customSkillDirs`, user `.dsh`, then user `.agents`. The user `.dsh/skills` scan skips `.system` so a system-owned directory is not treated as normal user content. DeepSeek Harness does not ship built-in system skills; embedded or remote providers supply additional skills when configured.

-Each skill is either `<name>/SKILL.md` or `<name>.md` with YAML frontmatter. `name` and `description` are required; `whenToUse`, `disableModelInvocation`, and `metadata` are optional. Names are kebab-case. YAML frontmatter is parsed with the `yaml` package instead of `js-yaml` or a hand-written parser: `yaml` is the already-declared modern parser for this package's limited frontmatter needs, and a narrow parser would either reject valid YAML users expect to work or grow into an unreviewed YAML subset.
+Each skill is either `<name>/SKILL.md` or `<name>.md` with YAML frontmatter. `name` and `description` are required; `whenToUse`, `metadata`, `disable-model-invocation`, and `user-invocable` are optional. Names are kebab-case. The invocation fields project into a typed nested policy as defined by the [independent model and user invocation decision](2026-07-28-skill-invocation-policy.md); the parser rejects the old camel-case spellings. YAML frontmatter is parsed with the `yaml` package instead of `js-yaml` or a hand-written parser: `yaml` is the already-declared modern parser for this package's limited frontmatter needs, and a narrow parser would either reject valid YAML users expect to work or grow into an unreviewed YAML subset.

 Local skill filesystem I/O goes through `ctx.fs` when a filesystem service is loaded: project-root lookup probes `.git` with `resolve` and `stat`, root discovery uses `listDir`, and skill reads use `readText`. The Node filesystem remains a fallback for minimal contexts that mount `dsh-skill-local` without the fs seam. Missing roots, unreadable or malformed skill files, and transient provider `list()` failures degrade to warn-and-skip so one bad source does not make every agent request fail; malformed candidates still fail fast because they are provider contract violations.

 `dsh-tool-skill` injects one durable user-role `<system-reminder>` catalog as a sourced `user/message` at the session's first `agent/step`, and only when that agent's tool view resolves this plugin's exact `skill` registration. The catalog contains sorted skill name and description only; it excludes bodies, paths, sources, providers, and routing hints. Descriptions are whitespace-normalized, XML-escaped, and capped by `catalogDescriptionMaxLength`, whose default is `500` and minimum is `3`. Full skill bodies are never included in the catalog. (The catalog originally rode the request-only [session-prefix seam](../../archived/feature/2026-07-07-session-prefix.md), archived; the [unified sourced-message decision](../architecture/2026-07-22-unified-send-and-coalesced-user-messages.md) moved it into durable history.)

-The `skill({ name })` tool loads one full skill for the current agent cwd and returns a tool result containing `<skill_content name="...">`, `<skill_resources>`, and `<skill_instructions>`. `resourceBase` supplies a directory, URL, or opaque provider-managed base for explicitly referenced scripts, references, and assets; resources load only as needed, without directory enumeration. An unresolved name reports that the skill is unknown or no longer available; invalid names and skills marked `disableModelInvocation` retain distinct tool errors. The tool result is the model-visible disclosure path.
+The registry's `list()` returns every winning summary, while model and user consumers apply the invocation predicates owned by the [independent invocation-policy decision](2026-07-28-skill-invocation-policy.md). The `skill({ name })` tool loads one model-invocable skill for the current agent cwd and returns a tool result containing `<skill_content name="...">`, `<skill_resources>`, and `<skill_instructions>`. `resourceBase` supplies a directory, URL, or opaque provider-managed base for explicitly referenced scripts, references, and assets; resources load only as needed, without directory enumeration. An unresolved name reports that the skill is unknown or no longer available; invalid names and skills with `invocation.modelInvocable: false` retain distinct tool errors. The tool result is the model-visible disclosure path.

 The data structures and catalog/tool contract are documented in [skills.md](../../../../docs/core-data-structures/skills.md), with service signatures in the generated [services catalog](../../../../docs/cordis-catalog/services.md).

@@ -52,4 +52,4 @@ The catalog is deterministic for a fixed root set and runtime registration revis

 ## Deferred

-Forked skill contexts (`context: fork`), parameter declarations and hints (`arguments` and `argument-hint`), and per-skill tool constraints (`allowed-tools` and `disallowed-tools`) are outside the shipped contract. The registry, local provider, and model-facing tool do not parse, advertise, or enforce these fields, and the `user-invocable` frontmatter field is likewise unparsed. Direct user invocation itself ships as a consumer-side affordance instead: the TUI front door offers a manual `/skill:<name>` command over the registry's existing `list()` and `get()` methods, without a registry, provider, or tool contract change — see [the TUI skill slash command](2026-07-21-tui-skill-slash-command.md).
+Forked skill contexts (`context: fork`), parameter declarations and hints (`arguments` and `argument-hint`), and per-skill tool constraints (`allowed-tools` and `disallowed-tools`) are outside the shipped contract. The registry, local provider, and model-facing tool do not parse, advertise, or enforce these fields. Direct user invocation ships as a TUI affordance over the shared invocation policy and trusted `get()` primitive; see [the TUI skill slash command](2026-07-21-tui-skill-slash-command.md).
@@ -18,13 +18,13 @@ DeepSeek Harness 使用同一原语，使项目特定的评审、插件编写和

 本地提供方按先到先得的排名顺序扫描 cwd 敏感的项目根目录、自定义根目录和用户根目录：项目 `.dsh`、项目 `.agents`、`customSkillDirs`、用户 `.dsh`，然后是用户 `.agents`。用户 `.dsh/skills` 扫描跳过 `.system`，以免系统拥有的目录被当作普通用户内容处理。DeepSeek Harness 不随附内置系统 skill；嵌入式或远程提供方在配置后提供额外 skill。

-每个 skill 是 `<name>/SKILL.md` 或带 YAML frontmatter 的 `<name>.md`。`name` 和 `description` 为必填；`whenToUse`、`disableModelInvocation` 和 `metadata` 为可选。名称采用 kebab-case。YAML frontmatter 使用 `yaml` 包（package）解析，而非 `js-yaml` 或手写解析器：`yaml` 是本包有限 frontmatter 需求已声明的现代解析器，窄解析器要么拒绝用户预期可用的合法 YAML，要么膨胀为一个未经评审的 YAML 子集。
+每个 skill 是 `<name>/SKILL.md` 或带 YAML frontmatter 的 `<name>.md`。`name` 和 `description` 为必填；`whenToUse`、`metadata`、`disable-model-invocation` 和 `user-invocable` 为可选。名称采用 kebab-case。调用字段会投影到类型化的嵌套策略中，具体由[模型与用户独立调用决策](2026-07-28-skill-invocation-policy.md)定义；解析器会拒绝旧的驼峰拼写。YAML frontmatter 使用 `yaml` 包（package）解析，而非 `js-yaml` 或手写解析器：`yaml` 是本包有限 frontmatter 需求已声明的现代解析器，窄解析器要么拒绝用户预期可用的合法 YAML，要么膨胀为一个未经评审的 YAML 子集。

 本地 skill 的文件系统 I/O 在加载了文件系统服务时通过 `ctx.fs` 进行：项目根目录查找使用 `resolve` 和 `stat` 探测 `.git`，根目录发现使用 `listDir`，skill 读取使用 `readText`。Node 文件系统作为后备，供在不挂载 fs seam 的最小上下文中加载 `dsh-skill-local` 时使用。缺失的根目录、不可读或格式错误的 skill 文件、以及提供方 `list()` 的瞬态失败均降级为警告并跳过，使一个坏源不会导致所有 agent 请求失败；格式错误的候选项仍然快速失败，因为它们违反了提供方契约。

 `dsh-tool-skill` 在会话的第一个 `agent/step` 注入一个持久化的 user-role `<system-reminder>` 目录，作为带来源的 `user/message`，且仅当该 agent 的工具视图解析到本插件精确的 `skill` 注册时才注入。该目录仅包含排序后的 skill 名称与描述；不包含正文、路径、来源、提供方和路由提示。描述经过空白规范化、XML 转义，并受 `catalogDescriptionMaxLength` 上限约束，其默认值为 `500`，最小值为 `3`。完整的 skill 正文从不包含在目录中。（目录最初通过仅请求的[会话前缀 seam](../../archived/feature/2026-07-07-session-prefix.md)（已归档）传递；[统一带来源消息的决策](../architecture/2026-07-22-unified-send-and-coalesced-user-messages.md)将其移入持久化历史。）

-`skill({ name })` 工具为当前 agent cwd 加载一个完整 skill，返回包含 `<skill_content name="...">`、`<skill_resources>` 和 `<skill_instructions>` 的工具结果。`resourceBase` 提供一个目录、URL 或不透明的提供方管理的基路径，用于显式引用的脚本、参考资料和资产；资源仅按需加载，不进行目录枚举。无法解析的名称报告该 skill 未知或不再可用；无效名称和标记了 `disableModelInvocation` 的 skill 保留不同的工具错误。工具结果是面向模型的可见披露路径。
+注册表的 `list()` 返回全部胜出摘要，而模型与用户消费方应用[独立调用策略决策](2026-07-28-skill-invocation-policy.md)定义的调用判定。`skill({ name })` 工具为当前 agent cwd 加载一个模型可调用的 skill，返回包含 `<skill_content name="...">`、`<skill_resources>` 和 `<skill_instructions>` 的工具结果。`resourceBase` 提供一个目录、URL 或不透明的提供方管理的基路径，用于显式引用的脚本、参考资料和资产；资源仅按需加载，不进行目录枚举。无法解析的名称报告该 skill 未知或不再可用；无效名称和 `invocation.modelInvocable` 为 `false` 的 skill 保留不同的工具错误。工具结果是面向模型的可见披露路径。

 数据结构与目录/工具契约记录在 [skills.md](../../../../docs/core-data-structures/skills.md) 中，服务签名见生成的[服务目录](../../../../docs/cordis-catalog/services.md)。

@@ -52,4 +52,4 @@ agent-core 主干包含一个目录贡献者、一个本地提供方和一个面

 ## 延后

-Fork 的 skill 上下文（`context: fork`）、参数声明与提示（`arguments` 和 `argument-hint`）、以及逐 skill 的工具约束（`allowed-tools` 和 `disallowed-tools`）不在已交付的契约范围内。注册表、本地提供方和面向模型的工具不解析、不广播、也不执行这些字段，`user-invocable` frontmatter 字段同样不会被解析。直接用户调用本身则作为消费方层面的能力交付：TUI 前门基于注册表现有的 `list()` 与 `get()` 方法提供手动 `/skill:<name>` 命令，无需变更注册表、提供方或工具契约——见 [TUI skill 斜杠命令](2026-07-21-tui-skill-slash-command.md)。
+Fork 的 skill 上下文（`context: fork`）、参数声明与提示（`arguments` 和 `argument-hint`）、以及逐 skill 的工具约束（`allowed-tools` 和 `disallowed-tools`）不在已交付的契约范围内。注册表、本地提供方和面向模型的工具不解析、不广播、也不执行这些字段。直接用户调用作为 TUI 功能交付，基于共享调用策略和受信的 `get()` 原语；见 [TUI skill 斜杠命令](2026-07-21-tui-skill-slash-command.md)。
@@ -1,6 +1,6 @@
 # Bilingual-pair consistency record (docs/i18n/README.md): the git blob hash of each
 # side as of the last confirmed-consistent state. Both languages carry equal authority;
 # after editing either side, bring the other along and re-record with:
-#   pnpm run verify-translation-pairing --write
-2026-07-06-approval-seam.md: 70ccd4d486ad6e0126fa2eb638a064e9fc89bba6
-2026-07-06-approval-seam.zh.md: d218f79888957735305db14cd97cc74480297d29
+#   pnpm run verify-translation-pairing --write .agents/notes/implemented/feature/2026-07-06-approval-seam.md
+2026-07-06-approval-seam.md: efb4159d736779af28edc1ae6091de4669c92f31
+2026-07-06-approval-seam.zh.md: 9a4656f30a43473fe90cde9c56d45f48943e5d10
@@ -123,7 +123,7 @@ Costs and accepted limits:
 - **Who decides whether a call asks in the first place?** Policy producers: a hook returning `permissionDecision: ask`, any `tools/pre-execute` listener, or the sandbox escalation gate. The seam and the bridge only route and answer; neither injects its own judgment about what deserves a prompt.
 - **What happens when the user dismisses the prompt, or the turn aborts mid-ask?** Dismissal maps to `cancelled` with its own deny text. An already-aborted signal settles `cancelled` without dispatching; an abort during the ask discards the late answer. When both audit appends commit, either path records one pair, never two.
 - **What if the client answers with an option the harness never offered?** Any selection other than the offered `allow_once` maps to `rejected` — an unknown optionId from a non-conforming client can never grant.
- **How do subagents' approvals route?** An agent no answerer owns delegates through the whole waterfall and fails closed — in-process subagents are deliberately unanswerable. `subagent-acp`'s child-side auto-answer is separate; routing a child's asks to the parent controller is deferred (§ Deferred).
+- **How do subagents' approvals route?** An agent no answerer owns delegates through the whole waterfall and fails closed — in-process subagents are deliberately unanswerable. A `'never'` parent seeds that override into each in-process child's log ([decision](2026-07-25-subagent-policy-inheritance.md)), so the child is told up front instead of asking into the empty waterfall. `subagent-acp`'s child-side auto-answer is separate; routing a child's asks to the parent controller is deferred (§ Deferred).
 - **What does `policy: 'never'` actually change at runtime?** The service resolves every ask for that session to `rejected` before dispatching any answerer (in-service, so no registration order can bypass it); the system prompt states the policy; switches are narrated at boundaries; each successful auto-rejection records the audit pair.
 - **What happens across a hot reload, or when an answerer unloads mid-session?** Answerers dispose with their owning fiber, so the next ask degrades to `unavailable` instead of hanging on a dead channel; remounting re-registers the answerer with no catch-up state.
 - **Where does a client get approval context?** The request carries the exact `callId` and the asker's human-readable `reason`; channel adapters may correlate richer tool-call state without duplicating arguments in the approval seam.
@@ -123,7 +123,7 @@ ACP 桥只应答其会话映射所拥有的精确 agent 对象。它携带既有
 - **谁决定一次调用是否需要 ask？** 策略生产者：返回 `permissionDecision: ask` 的钩子、任何 `tools/pre-execute` 监听器、或沙箱升级门禁。seam 和桥只负责路由和应答；二者都不注入自己对「什么值得弹出提示」的判断。
 - **用户关闭提示或轮次在 ask 进行中中止时会发生什么？** 关闭映射为 `cancelled` 并携带自己的拒绝文本。已中止的 signal 直接结算为 `cancelled` 而不派发；ask 进行中的中止丢弃迟到的应答。当两个审计追加都提交时，任一路径都记录恰好一对事件，绝不会两对。
 - **如果客户端以 harness 从未提供的选项应答呢？** 除已提供的 `allow_once` 之外的任何选项都映射为 `rejected`——来自不合规客户端的未知 optionId 永远不能授权。
- **subagent 的审批如何路由？** 没有应答者拥有的 agent 穿过整个 waterfall 委派并失败关闭——进程内 subagent 被刻意设计为不可应答。`subagent-acp` 的子侧自动应答是独立的；将子 agent 的 ask 路由到父控制器已延后（§ 延后）。
+- **subagent 的审批如何路由？** 没有应答者拥有的 agent 穿过整个 waterfall 委派并失败关闭——进程内 subagent 被刻意设计为不可应答。`'never'` 父级会把该覆盖项预置到每个进程内子 agent 的日志中（[决策](2026-07-25-subagent-policy-inheritance.md)），因此子 agent 一开始就会得知，而不是向空的 waterfall 发出 ask。`subagent-acp` 的子侧自动应答是独立的；将子 agent 的 ask 路由到父控制器已延后（§ 延后）。
 - **`policy: 'never'` 在运行时实际改变了什么？** 服务在派发任何应答者之前，将该会话的每次 ask 解析为 `rejected`（在服务内部，因此没有注册顺序能绕过它）；系统提示词声明该策略；切换在边界处被叙述；每次成功的自动拒绝都会记录审计对。
 - **热重载或应答者在会话中途卸载时会发生什么？** 应答者随其拥有的 fiber 一起 dispose，因此下一次 ask 降级为 `unavailable` 而非挂在死通道上；重新挂载会重新注册应答者，无需追赶状态。
 - **客户端从哪里获得审批上下文？** 请求携带精确的 `callId` 和发起方的人类可读 `reason`；通道适配器可自行关联更丰富的工具调用状态，而无需在审批 seam 中重复携带参数。
@@ -2,5 +2,5 @@
 # side as of the last confirmed-consistent state. Both languages carry equal authority;
 # after editing either side, bring the other along and re-record with:
 #   pnpm run verify-translation-pairing --write .agents/notes/implemented/feature/2026-07-06-sandbox.md
-2026-07-06-sandbox.md: c3c61ed4539bcbca359f84f3dcc020e0bd41ae79
-2026-07-06-sandbox.zh.md: 39bf92aa20c697d2ad5f4c0926caeff0a0b60a1d
+2026-07-06-sandbox.md: 42b78ad8341dd52c4dd146a2207a5ae909d28f1e
+2026-07-06-sandbox.zh.md: dfa3349e4d74d6f2c4944414c25fe3726d4a9b5a
@@ -89,10 +89,10 @@ Left open: what a durable grant's scope identity is beyond the sandbox mode —
 #### Per-session modes: the session log as the store

 ```
-effective(session) = findLast(the session's own knob events)?.value ?? the composition-config default
+effective(session) = findLast(the session's knob events)?.value ?? the composition-config default
 ```

-The default is composition config (`cordis.yml`) — operator-owned, process-wide. A runtime switch is a session-scoped override recorded as one log-only event in that session's own log. Restart immunity (resuming a session replays its log, so overrides come back with zero catch-up machinery) and multi-session isolation both fall out by construction, and no external config store exists anywhere.
+The default is composition config (`cordis.yml`) — operator-owned, process-wide. A runtime switch is a session-scoped override recorded as one log-only event in that session's log. Restart immunity and multi-session isolation follow from replay, with no external config store. The in-process subagent driver snapshots a parent's explicit override at delegation and seeds a source-tagged event after the child's optional fork prefix, so delegation cannot fall back to a wider default ([decision](2026-07-25-subagent-policy-inheritance.md)).

 **One event per knob, owned by its domain** — the merge-extensible `SessionEventMap` idiom every existing event family already follows (`approval/*` in `dsh-user-approval`, `hook/*` in the hooks packages):

@@ -89,10 +89,10 @@ Landlock launcher 源码和包工作区位于 `native/landlock-run`，与 harnes
 #### 按会话模式：会话日志即存储

 ```
-effective(session) = findLast(the session's own knob events)?.value ?? the composition-config default
+effective(session) = findLast(the session's knob events)?.value ?? the composition-config default
 ```

-默认值是组合配置（`cordis.yml`）——运维人员拥有，进程范围。运行时切换是会话范围的覆盖，记录为该会话自身日志中的一条仅日志事件。重启免疫（恢复会话时回放其日志，覆盖自然恢复，无需追赶机制）和多会话隔离都是构造性的自然结果，且不存在任何外部配置存储。
+默认值是组合配置（`cordis.yml`）——由运维人员拥有、作用于整个进程。运行时切换是会话范围的覆盖，以一条仅日志事件记录在该会话的日志中。重启免疫与多会话隔离由回放自然保证，且不存在任何外部配置存储。进程内 subagent 驱动器在委派时对父级的显式覆盖项获取快照，并在子 agent 可选的 fork 前缀之后预置一条带来源标记的事件，因此委派无法回退到更宽的默认值（[决策](2026-07-25-subagent-policy-inheritance.md)）。

 **每个旋钮一种事件，由其领域拥有**——这是每个既有事件族已遵循的可合并扩展 `SessionEventMap` 惯用法（`dsh-user-approval` 中的 `approval/*`、hooks 包中的 `hook/*`）：

@@ -1,6 +1,6 @@
 # Bilingual-pair consistency record (docs/i18n/README.md): the git blob hash of each
 # side as of the last confirmed-consistent state. Both languages carry equal authority;
 # after editing either side, bring the other along and re-record with:
-#   pnpm run verify-translation-pairing --write
-2026-07-07-mcp-client-plugin.md: 2a8ad62e3bad1f2ae47100294f5dba25ceb47d12
-2026-07-07-mcp-client-plugin.zh.md: 9860d5ec805b394eb0c5f59f54349264593a5599
+#   pnpm run verify-translation-pairing --write .agents/notes/implemented/feature/2026-07-07-mcp-client-plugin.md
+2026-07-07-mcp-client-plugin.md: 756a5c4dc9f1152ecb1955d93b8dc47fcf07c661
+2026-07-07-mcp-client-plugin.zh.md: d2ff04d68033402fbb2718f8624e44d35d860db3
@@ -149,7 +149,7 @@ A unified `execute` handler for all tools from one MCP server:

 ### Subprocess environment (stdio transport)

-Replicate the `buildChildEnv` + `SENSITIVE_ENV_PATTERN` scrub from `dsh-subagent-acp`: filter ambient env (strip credential-shaped vars matching `/KEY|SECRET|TOKEN/i`), then merge `config.env` on top. Explicit env overrides survive the scrub.
+Build the child environment from the subprocess seam's shared `scrubbedParentEnv()` base, which removes ambient names matching `/KEY|PASSWORD|SECRET|TOKEN/i` and ambient `DSH_*` names, then merge `config.env` on top. Explicit env overrides survive the scrub.

 ### Disconnection / crash

@@ -149,7 +149,7 @@ MCP 仅保证工具名在[单个服务器内](https://modelcontextprotocol.io/sp

 ### 子进程环境（stdio 传输）

-复用 `dsh-subagent-acp` 的 `buildChildEnv` + `SENSITIVE_ENV_PATTERN` 清洗逻辑：过滤环境变量（剥离匹配 `/KEY|SECRET|TOKEN/i` 的凭证形变量），然后将 `config.env` 覆盖合并到顶层。显式配置的 env 不受清洗影响。
+以子进程服务边界共享的 `scrubbedParentEnv()` 为基础构建子进程环境；该基础环境会移除环境中匹配 `/KEY|PASSWORD|SECRET|TOKEN/i` 的名称以及 `DSH_*` 名称，然后在其上合并 `config.env`。显式配置的 env 覆盖在清洗后仍会保留。

 ### 断连 / 崩溃

@@ -1,6 +1,6 @@
 # Bilingual-pair consistency record (docs/i18n/README.md): the git blob hash of each
 # side as of the last confirmed-consistent state. Both languages carry equal authority;
 # after editing either side, bring the other along and re-record with:
-#   pnpm run verify-translation-pairing --write
-2026-07-10-parallel-tool-call-execution.md: c67ae61939a3e7974f9bf729058a57f5576308a1
-2026-07-10-parallel-tool-call-execution.zh.md: a80317aa951cbf3a9cae0651348c99712a4193d5
+#   pnpm run verify-translation-pairing --write .agents/notes/implemented/feature/2026-07-10-parallel-tool-call-execution.md
+2026-07-10-parallel-tool-call-execution.md: 19f5dc189821433052edfa72613980a2e94e2cae
+2026-07-10-parallel-tool-call-execution.zh.md: e90e357180ae3684500adf7cba41c5fc1dac5743
@@ -10,7 +10,7 @@ An assistant message may contain several sibling `tool-call` blocks. Running the

 Concurrency is a host scheduling concern, not model-facing tool metadata. The loop needs to decide which calls may overlap without hardcoding tool names or exposing scheduler policy in the JSON schema.

-The session log remains authoritative: every started call has an audit event, every started call receives a result, and model history observes results in the original call order regardless of completion order.
+The session log remains authoritative: every started call has an audit event, ordinary completion and cancellation pair calls with results, and model history observes committed results in the original call order regardless of completion order.

 ## Decision

@@ -46,7 +46,7 @@ Only dispatch and the tool body overlap. `tools/pre-execute` and `tools/post-exe

 Each started call appends `tool/call` immediately before its pre-execute gate. Completed dispatches occupy model-order slots, and a commit cursor appends `tool/result` and collects `additionalContexts` only when the next slot is ready. Live surfaces may show several pending calls, but results and post-tool context remain model-ordered.

-An abort before a group starts records no calls from that group. An abort during a group stops replenishment, waits for already-started calls, commits their results in order, drains accepted batch context after those results, and then ends the step through the existing abort path. Calls that never start have no audit event.
+An abort before a group starts records no calls from that group. An abort during a group stops replenishment, waits for already-started calls, commits their results in order, drains accepted batch context after those results, and then ends the step through the existing abort path. Calls that never start have no audit event. An unexpected scheduler failure stops new dispatches, waits for every already-started dispatch to settle, and rethrows the first failure. Because that failure is terminal internal state rather than a tool outcome, the loop does not invent tool results for rejected or uncommitted calls.

 Code Mode remains outside this scheduler because the model emits one native `run_code` call. `run_code` and its internal dispatch queue remain serial; native sibling calls in `mode: 'both'` use the normal scheduler.

@@ -66,7 +66,7 @@ Filesystem read relies on a narrow recorder exception: its synchronous observati

 ## Verification

-Unit coverage pins fail-closed classification, typed argument validation, grouping, barriers, live reclassification after registry replacement, the rolling cap, distinct execution objects, middleware order, ordered results and context, and abort draining. First-party tests pin each parallel declaration.
+Unit coverage pins fail-closed classification, typed argument validation, grouping, barriers, live reclassification after registry replacement, the rolling cap, distinct execution objects, middleware order, ordered results and context, abort draining, and scheduler-failure quiescence. First-party tests pin each parallel declaration.

 Snapshot coverage pins the visible multi-call transcript: pending calls may overlap while completed results remain model-ordered. Code Mode coverage pins its serial boundary. No provider-backed e2e is required because scheduling is deterministic loop behavior.

@@ -84,6 +84,8 @@ Snapshot coverage pins the visible multi-call transcript: pending calls may over

 **Expose staged methods or a scheduling waterfall.** Public `prepare` / `dispatch` / `finalize` methods or a `tools/execution-mode` event add extension surface before another consumer needs it. The loop uses an internal scheduler view, while `executionMode(exec)` leaves an insertion point for a policy seam.

+**Convert scheduler failures into tool results.** AgentLoop cannot determine whether a rejected dispatch invoked the tool body; ToolRegistry owns body-invocation state and typed tool outcomes. Internal scheduler failures therefore remain terminal instead of being reclassified as `ABORTED` results.
+
 **Start calls while the model streams.** This may reduce latency further but changes assistant-message authority, replay, and call/result pairing. The scheduler starts only after the assistant message is complete.

 **Use fixed-size windows.** Waiting for every call in one window before starting the next leaves capacity idle behind a slow call. The rolling pool preserves the cap without that delay.
@@ -101,3 +103,5 @@ Ordered commits may hold a fast result behind a slow earlier sibling. This prese
 Concurrent external calls can compete for quota or process capacity. Providers own their capacity controls; the loop cap only limits calls from one agent step.

 Tool registration is a scheduling boundary. Registry mutations affect not-yet-started calls because the scheduler reclassifies after each barrier and before every pool replenishment. Already-started calls retain the scheduling decision under which they entered the pool.
+
+A terminal scheduler failure may leave recorded calls without results before the failed step closes. Waiting for live dispatches preserves quiescence without misreporting those internal failures as tool outcomes.
@@ -10,7 +10,7 @@ Status: implemented

 并发属于宿主调度范畴，不是面向模型的工具元数据。循环需要在不硬编码工具名称、不向 JSON Schema 暴露调度策略的前提下，判断哪些调用可以重叠执行。

-会话日志仍是权威记录：每个已启动的调用都有审计事件，都会获得结果；无论完成顺序如何，模型历史都按原始调用顺序观察结果。
+会话日志仍是权威记录：每个已启动的调用都有审计事件，正常完成和取消都会使调用与结果配对；无论完成顺序如何，模型历史都按原始调用顺序观察已提交的结果。

 ## 决策

@@ -46,7 +46,7 @@ Status: implemented

 每个已启动的调用都会在进入 pre-execute 门禁之前立即追加 `tool/call`。已完成的派发占据模型顺序的槽位；提交游标只有在下一个槽位就绪时，才会追加 `tool/result` 并收集 `additionalContexts`。实时界面可以显示多个待处理调用，但结果和工具执行后的上下文仍按模型顺序排列。

-如果在一组启动前中止，系统不会记录该组的任何调用。如果在一组执行期间中止，系统会停止补充池，等待已启动的调用，按顺序提交其结果，在这些结果之后排空已接受的批次上下文，然后通过现有中止路径结束该步骤。从未启动的调用没有审计事件。
+如果在一组启动前中止，系统不会记录该组的任何调用。如果在一组执行期间中止，系统会停止补充池，等待已启动的调用，按顺序提交其结果，在这些结果之后排空已接受的批次上下文，然后通过现有中止路径结束该步骤。从未启动的调用没有审计事件。调度器发生意外故障时，会停止新的派发，等待每项已启动的派发结算，并重新抛出第一个故障。由于该故障是内部终态，而非工具结果，循环不会为被拒绝或未提交的调用虚构工具结果。

 Code Mode 仍不使用此调度器，因为模型只会发出一个原生 `run_code` 调用。`run_code` 及其内部派发队列仍按串行方式执行；`mode: 'both'` 中的原生并列调用使用常规调度器。

@@ -66,7 +66,7 @@ Code Mode 仍不使用此调度器，因为模型只会发出一个原生 `run_c

 ## 验证

-单元测试覆盖固定了安全退化的分类、类型化参数验证、分组、屏障、替换注册表后的实时重新分类、滚动上限、独立执行对象、中间件顺序、有序结果与上下文，以及中止排空。第一方测试固定每项并行声明。
+单元测试覆盖固定了安全退化的分类、类型化参数验证、分组、屏障、替换注册表后的实时重新分类、滚动上限、独立执行对象、中间件顺序、有序结果与上下文、中止排空，以及调度器故障后的完全停稳。第一方测试固定每项并行声明。

 快照覆盖固定了可见的多调用 transcript（文本记录）：待处理调用可以重叠执行，已完成结果仍按模型顺序排列。Code Mode 覆盖固定其串行边界。此调度属于确定性循环行为，因此无需依赖提供方的 e2e 测试。

@@ -84,6 +84,8 @@ Code Mode 仍不使用此调度器，因为模型只会发出一个原生 `run_c

 **公开分阶段方法或调度 waterfall。** 公开的 `prepare` / `dispatch` / `finalize` 方法或 `tools/execution-mode` 事件，会在出现另一个消费方之前扩大扩展接口。循环使用内部调度器视图，而 `executionMode(exec)` 为策略 seam 保留了插入点。

+**将调度器故障转换为工具结果。** AgentLoop 无法判断被拒绝的派发是否已调用工具主体；ToolRegistry 负责工具主体调用状态和类型化工具结果。因此，内部调度器故障保持为终态，而不会被重新分类为 `ABORTED` 结果。
+
 **在模型流式输出时启动调用。** 这可能进一步降低延迟，但会改变 assistant 消息的权威性、回放以及调用/结果配对。调度器只在 assistant 消息完成后才启动。

 **使用固定大小的窗口。** 如果在启动下一个窗口前等待当前窗口的每个调用，一个缓慢调用就会使容量闲置。滚动池在保持上限的同时避免了这项延迟。
@@ -101,3 +103,5 @@ Code Mode 仍不使用此调度器，因为模型只会发出一个原生 `run_c
 并发外部调用可能会争用配额或进程容量。提供方负责自身容量控制；循环上限只限制一个 agent 步骤中的调用数量。

 工具注册是调度边界。调度器会在每个屏障之后以及每次补充池之前重新分类，因此注册表变更会影响尚未启动的调用。已启动的调用保留它们进入池时所依据的调度决策。
+
+终态调度器故障可能会在故障步骤关闭前留下已记录但没有结果的调用。等待仍在运行的派发可确保完全停稳，而不会将这些内部故障误报为工具结果。
@@ -2,5 +2,5 @@
 # side as of the last confirmed-consistent state. Both languages carry equal authority;
 # after editing either side, bring the other along and re-record with:
 #   pnpm run verify-translation-pairing --write .agents/notes/implemented/feature/2026-07-16-persistent-pty-sessions.md
-2026-07-16-persistent-pty-sessions.md: 691ccd4341837d63bb48d27c2a3fb007657fe7ba
-2026-07-16-persistent-pty-sessions.zh.md: 86d57f79ee74429542730ebcad60bc9f3c9f15ca
+2026-07-16-persistent-pty-sessions.md: d7d06dc8517780a37889e4f94dd0b99475dc5432
+2026-07-16-persistent-pty-sessions.zh.md: 7783fa4b4f3056ab3b3088a4e9618f45d030ccbf
@@ -40,7 +40,7 @@ Agent-scope disposal closes registrations first, then awaits quiescent teardown

 A registered `shell` backend constrains how a terminal starts; it does not constrain commands typed after startup. `dsh-pty-local` therefore applies two protections before spawning:

- It builds a scrubbed child environment using the same credential-shaped-name policy as `bash-local`, removing ambient `*KEY*`, `*SECRET*`, `*TOKEN*`, and harness-managed variables unless an explicit trusted mapping supplies them.
+- It builds a scrubbed child environment using the same credential-shaped-name policy as `bash-local`, removing ambient `*KEY*`, `*PASSWORD*`, `*SECRET*`, `*TOKEN*`, and harness-managed variables unless an explicit trusted mapping supplies them.
 - It requires `ctx.sandbox` and the shared `ctx.sandboxPolicy`. At spawn, the backend resolves the owner's effective session mode over the deployment default and wraps the shell argv once; that mode and workspace root remain the process boundary for the PTY lifetime. A write that would change the effective `sandbox/mode` is rejected before commit while the owner has any open PTY or unpublished spawn, with an instruction to wait for creation to settle and close those sessions first; same-effective-mode writes remain valid. The pending reservation spans backend setup through publication, so there is no race in which a wider terminal appears after a downgrade. `danger-full-access` is the existing explicit unconfined choice rather than a PTY-specific bypass.

 Sandboxing confines local process effects but does not make arbitrary shell input safe: network calls and other external side effects remain governed by deployment policy. Tool descriptions state that PTY sessions are less auditable than one-shot tools and should be used only when persistence or interactive stdin is necessary.
@@ -40,7 +40,7 @@ agent scope dispose 时先关闭注册，再等待全部所属 PTY 静默退出

 注册的 `shell` 后端只约束终端如何启动，不约束启动后输入的命令。因此 `dsh-pty-local` 在 spawn 前应用两层保护：

- 它使用与 `bash-local` 相同的凭证形态名称策略构建清洗后的子进程环境，移除环境中的 `*KEY*`、`*SECRET*`、`*TOKEN*` 和 harness 管理的变量，除非显式的可信映射提供这些值。
+- 它使用与 `bash-local` 相同的凭证形态名称策略构建清洗后的子进程环境，移除环境中的 `*KEY*`、`*PASSWORD*`、`*SECRET*`、`*TOKEN*` 和 harness 管理的变量，除非显式的可信映射提供这些值。
 - 它要求 `ctx.sandbox` 和共享的 `ctx.sandboxPolicy`。后端在 spawn 时，以部署默认值为底折叠 owner 的有效 session mode，并只包装一次 shell argv；该 mode 与 workspace root 在 PTY 的整个生命周期中充当进程边界。只要 owner 有任何已打开的 PTY 或尚未发布的 spawn，任何会改变生效 `sandbox/mode` 的写入都会在提交前被拒绝，并提示先等待创建操作结算，再关闭这些会话；不会改变生效模式的写入仍然有效。这项进行中的预留从后端 setup 持续到发布完成，因此不存在降级后又出现权限更宽的终端这一竞态。`danger-full-access` 是现有的显式无约束选择，不另设 PTY 私有 bypass。

 沙箱限制本地进程副作用，但不会让任意 shell 输入自动安全：网络调用和其他外部副作用仍由部署策略治理。工具描述会说明 PTY 会话比一次性工具更难审计，只应在确实需要持久状态或交互式 stdin 时使用。
@@ -1,6 +1,6 @@
 # Bilingual-pair consistency record (docs/i18n/README.md): the git blob hash of each
 # side as of the last confirmed-consistent state. Both languages carry equal authority;
 # after editing either side, bring the other along and re-record with:
-#   pnpm run verify-translation-pairing --write
-2026-07-20-dsh-cli-personal-config.md: 9525aa811d792a918f03a52c21bc273e92fb8be7
-2026-07-20-dsh-cli-personal-config.zh.md: f21d4b1f22b3a3807b6b4155969282343f6048f5
+#   pnpm run verify-translation-pairing --write .agents/notes/implemented/feature/2026-07-20-dsh-cli-personal-config.md
+2026-07-20-dsh-cli-personal-config.md: cc965438214b68078647596af5a28fd666e7bd95
+2026-07-20-dsh-cli-personal-config.zh.md: 02f9c578061e1c25044c377c3ec8f80594275ae7
@@ -12,7 +12,7 @@ A developer's own preferences — which provider and model the TUI uses, persona

 Two coupled pieces, aligned with the `apps/` assembly tier proposed by the `dsh web` PR (#443):

-**The `dsh` CLI (`apps/cli`, npm name `@deepseek-ai/dsh`).** `apps/*` joins the workspaces as the product-assembly tier over `packages/*` libraries. The bin's dispatch reserves `web` and `-p`/`--prompt` for PR #443 (they exit with a pointer) so the two branches merge as a near-union; everything else runs the default surface: the interactive TUI, booting the shipped `examples/tui-agent/cordis.yml` (or an explicit config argument) with the invoking directory as the workspace. The committed `bin/dsh` launcher resolves the checkout through its own real path and runs the bin **from source** via the repo's tsx, so `ln -sf "$(pwd)/bin/dsh" ~/.local/bin/dsh` installs a command that always executes the current working tree. `pnpm run demo:tui` runs the same entry.
+**The `dsh` CLI (`apps/cli`, npm name `@deepseek-ai/dsh`).** `apps/*` joins the workspaces as the product-assembly tier over `packages/*` libraries. The bin's dispatch reserves `web` and `-p`/`--prompt` for PR #443 (they exit with a pointer) so the two branches merge as a near-union; everything else runs the default surface: the interactive TUI, booting the shipped `examples/tui-agent/cordis.yml` (or an explicit config argument) with the invoking directory as the workspace. The committed `bin/dsh` launcher resolves the checkout through its own real path and runs the bin **from source** through Node's native TypeScript transform plus the app-owned tsconfig-paths loader, so `ln -sf "$(pwd)/bin/dsh" ~/.local/bin/dsh` installs a command that always executes the current working tree. `pnpm run demo:tui` runs the same entry.

 **Personal config (`dsh-app-boot`).** The personal overlay lives in the Harness home — `$DSH_HOME`, else `~/.dsh` — resolved by the shared [`resolveDshHome`](../architecture/2026-07-24-single-harness-home-resolver.md) (`@deepseek-ai/dsh-paths`), the same single root skills and AGENTS.md resolve against. The dsh TUI surface consumes its two optional files; the demo bins boot their committed trees verbatim:

@@ -12,7 +12,7 @@ Status: implemented

 两个耦合的部分，与 `dsh web` PR（#443）提出的 `apps/` 装配层对齐：

-**`dsh` CLI（`apps/cli`，npm 名 `@deepseek-ai/dsh`）。** `apps/*` 作为 `packages/*` 库之上的产品装配层加入 workspaces。bin 的分发把 `web` 和 `-p`/`--prompt` 保留给 PR #443（它们以指引退出），使两个分支能以接近并集的方式合并；其余一切都运行默认表面：交互式 TUI，加载随仓库提供的 `examples/tui-agent/cordis.yml`（或显式的配置参数），并以调用目录为工作区。已提交的 `bin/dsh` 启动器通过自身真实路径解析 checkout，用仓库的 tsx **从源码**运行该 bin，因此 `ln -sf "$(pwd)/bin/dsh" ~/.local/bin/dsh` 安装的命令永远执行当前工作树。`pnpm run demo:tui` 运行同一入口。
+**`dsh` CLI（`apps/cli`，npm 名 `@deepseek-ai/dsh`）。** `apps/*` 作为 `packages/*` 库之上的产品装配层加入 workspaces。bin 的分发把 `web` 和 `-p`/`--prompt` 保留给 PR #443（它们以指引退出），使两个分支能以接近并集的方式合并；其余一切都运行默认表面：交互式 TUI，加载随仓库提供的 `examples/tui-agent/cordis.yml`（或显式的配置参数），并以调用目录为工作区。已提交的 `bin/dsh` 启动器通过自身真实路径解析 checkout，通过 Node 的原生 TypeScript 转换和应用自身持有的 tsconfig-paths loader **从源码**运行该 bin，因此 `ln -sf "$(pwd)/bin/dsh" ~/.local/bin/dsh` 安装的命令永远执行当前工作树。`pnpm run demo:tui` 运行同一入口。

 **个人配置（`dsh-app-boot`）。** 个人 overlay 存放在 Harness home——`$DSH_HOME`，否则 `~/.dsh`——由共享的 [`resolveDshHome`](../architecture/2026-07-24-single-harness-home-resolver.md)（`@deepseek-ai/dsh-paths`）解析，与 skills、AGENTS.md 解析所依据的单一根目录相同。dsh 的 TUI 表面消费其中两个可选文件；各示例 bin 仍然逐字节按已提交的配置树启动：

@@ -2,5 +2,5 @@
 # side as of the last confirmed-consistent state. Both languages carry equal authority;
 # after editing either side, bring the other along and re-record with:
 #   pnpm run verify-translation-pairing --write .agents/notes/implemented/feature/2026-07-21-cross-session-references.md
-2026-07-21-cross-session-references.md: a6c7c3b3e89e032d6d807a25a7661a875f9f7bec
-2026-07-21-cross-session-references.zh.md: ef78cd7ecee3e2625b446139d7f22e16d0bf1dfa
+2026-07-21-cross-session-references.md: 61ea30cabb2abf4d4a9b4391891b5987affb91d0
+2026-07-21-cross-session-references.zh.md: 6fba44103942d29428fd820591815743dfb5d96d
@@ -10,7 +10,7 @@ TUI users need to bring relevant work from another conversation into one new mes

 ## Decision

-`@deepseek-ai/dsh-session-reference` is one context consumer service at `ctx.sessionReferences`. Hosts normalize their protocol into `SessionReferenceInput[]` and call `prepare()` before delivery. The service returns detached readable content plus an optional sourced `UserMessageData` snapshot; core agent packages do not parse session URIs or read another log.
+`@deepseek-ai/dsh-session-reference` is one context consumer service at `ctx.sessionReferences`. Hosts normalize their protocol into `SessionReferenceInput[]` and call `prepare()` before delivery. The service returns detached readable content plus an optional identified, frozen `UserMessage` snapshot; core agent packages do not parse session URIs or read another log.

 `dsh-session:<base64url(JSON.stringify(sessionId))>` is the canonical host-independent identifier. JSON string encoding precedes base64url so quotes, slashes, backslashes, Unicode, newlines, and every other JavaScript string value round-trip without delimiter ambiguity. TUI renders that URI inside `@[label](uri)`; text-only clients may use the same inline mention. Explicit Markdown mentions reject malformed URIs. Bare text becomes a reference only for a non-empty base64url-shaped payload, whose decode must still be canonical; empty or punctuation-only uses remain ordinary discussion text.

@@ -10,7 +10,7 @@ TUI 用户需要把另一场对话中的相关工作带入一条新消息，但

 ## 决策

-`@deepseek-ai/dsh-session-reference` 是注册在 `ctx.sessionReferences` 上的单一上下文消费服务。宿主先把各自的协议规范化为 `SessionReferenceInput[]`，并在交付前调用 `prepare()`。该服务返回分离的可读内容和一份可选的、带来源信息的 `UserMessageData` 快照；核心 agent 包既不解析会话 URI，也不读取其他日志。
+`@deepseek-ai/dsh-session-reference` 是注册在 `ctx.sessionReferences` 上的单一上下文消费服务。宿主先把各自的协议规范化为 `SessionReferenceInput[]`，并在交付前调用 `prepare()`。该服务返回分离的可读内容和一份可选的、带标识且冻结的 `UserMessage` 快照；核心 agent 包既不解析会话 URI，也不读取其他日志。

 `dsh-session:<base64url(JSON.stringify(sessionId))>` 是与宿主无关的规范标识符。系统先执行 JSON 字符串编码，再执行 base64url 编码，因此引号、正斜杠、反斜杠、Unicode、换行符以及其他任意 JavaScript 字符串值都能无损往返，不会因分隔符产生歧义。TUI 把该 URI 渲染到 `@[label](uri)` 中；纯文本客户端可以使用同一种行内提及标记。显式 Markdown 提及标记会拒绝格式错误的 URI。裸文本只有在负载非空且形状符合 base64url 时才会成为引用，而且解码过程仍须通过规范性校验；空负载或只含标点符号的用法仍按普通讨论文本处理。

@@ -2,5 +2,5 @@
 # side as of the last confirmed-consistent state. Both languages carry equal authority;
 # after editing either side, bring the other along and re-record with:
 #   pnpm run verify-translation-pairing --write .agents/notes/implemented/feature/2026-07-21-log-backed-session-titles.md
-2026-07-21-log-backed-session-titles.md: 1bd58e35ec625fb0b04c0c119ce425ff30a64881
-2026-07-21-log-backed-session-titles.zh.md: 37ec95efbca334f71d19d2bc3e18c22d50d9b5fb
+2026-07-21-log-backed-session-titles.md: 81ac687c6f55dd0ca1eaeb9d84c811edcfe17b5c
+2026-07-21-log-backed-session-titles.zh.md: b0c7e9d76a1b9365fa16dcb223b390b5aec3e174
@@ -36,9 +36,13 @@ Model providers require explicit word, CJK-character, input-byte, output-token,

 Automatic provider failures are nonfatal warnings and retain the latest title. Explicit refresh failures reject to the caller. Output must be non-empty text with unique ordered seqs drawn from the fixed request; the service normalizes and byte-limits it before log acceptance.

+### Explicit rename
+
+`rename(session, title)` accepts a user title synchronously: it normalizes the text under the accepted-title byte limit, supersedes in-flight automatic work, and appends a `session/title` event with the third source kind, `user`. A user-sourced latest title pins the session: `onUserMessage` schedules no automatic revision while it stands, under either cadence. An explicit `refresh()` remains the deliberate unpin — it appends a provider or fallback event over the pinned one whenever a replacement title is derivable (an underivable fallback, e.g. under a tiny byte cap, leaves the pin standing). The Web host exposes this as the `session.rename` unary method (resuming cold sessions first) and returns the normalized title plus its event seq so the client settles its `title` projection cell before the push frame arrives.
+
 ### Forks and consumers

-A fork inherits seed title events unchanged, like the rest of its source log. The first-message provider does not automatically retitle a fork. The all-messages provider may append a child-owned revision after a later child prompt, using inherited and new eligible messages.
+A fork inherits seed title events unchanged, like the rest of its source log — a pinned (user-sourced) title stays pinned in the child until an explicit refresh. The first-message provider does not automatically retitle a fork. The all-messages provider may append a child-owned revision after a later child prompt, using inherited and new eligible messages.

 `ctx.sessionQuery.readTitle()` folds one live-preferred or persisted log without loading titles during `listSessions()`. The TUI uses the latest title as its header subtitle and sets the terminal window title to `<session title> — <configured product title>` after terminal-safe rendering. The Web host folds the same log state into a validated mux control frame after each attached-session subscription baseline and immediately after forwarding a live raw title event. The browser retains only newer title event seqs even when the control frame precedes list or session-instance creation; sidebar labels, search, breadcrumbs, and the browser title then react to the projected revision. `session.list` remains metadata-only, so a cold persisted session uses the cwd basename or id until opening or resuming it attaches the log. The browser title uses `<session title> — <existing HTML title>` only for a selected titled session and otherwise preserves the product title. Consumers reporting agent completion use the core `findLastMessageTurnEnd()` fold, so a later between-turn title record cannot replace the preceding message-triggered outcome.

@@ -59,4 +63,4 @@ A fork inherits seed title events unchanged, like the rest of its source log. Th
 - A fallback appears immediately. Each fresh Web session adds one first-message auxiliary call; other compositions choose whether better titles justify model cost and whether later prompts should retitle a session.
 - Auxiliary request records and late accepted titles consume event seqs without consuming turn numbers, so persistence exposes both attempted dispatches and accepted updates even though model history and KV-cache identity do not change.
 - One provider and monotonic per-session revisions make disposal, supersession, and stale-result rejection explicit, at the cost of leaving multi-strategy precedence to a composite provider.
- Manual rename, deletion, generated-versus-user precedence, search, and list indexing remain outside the capability.
+- Deletion (unpinning without an explicit refresh), search, and list indexing remain outside the capability.
@@ -36,9 +36,13 @@ Status: implemented

 自动提供方故障只会发出非致命警告，并保留最新标题。显式刷新失败则会向调用方返回拒绝。输出必须是非空文本，并包含来自固定请求、唯一且有序的 seq；服务会在日志接受前对其进行规范化并施加字节限制。

+### 显式重命名
+
+`rename(session, title)` 同步接受用户标题：按已接受标题的字节上限规范化文本、取代在途自动工作，并追加一条第三种来源 `user` 的 `session/title` 事件。最新标题来源为 user 即钉住该会话：只要它还在，`onUserMessage` 在任一节奏下都不再安排自动修订。显式 `refresh()` 仍是有意的解钉手段——只要能推导出替代标题，它就在被钉住的标题之上追加提供方或回退事件（推导不出回退标题时，例如字节上限过小，钉住状态保持不变）。Web host 将其暴露为 `session.rename` unary 方法（冷会话先恢复），并返回规范化后的标题及其事件 seq，使 client 在推送帧到达前就结算自己的 `title` 投影格。
+
 ### Fork 与消费方

-与源日志的其他部分相同，fork 会原样继承作为种子的标题事件。首消息提供方不会自动为 fork 重新生成标题。全部消息提供方可以在子会话出现后续提示词后追加一项归子会话所有的修订，并使用继承的合格消息和新增的合格消息。
+与源日志的其他部分相同，fork 会原样继承作为种子的标题事件——被钉住（user 来源）的标题在子会话中保持钉住，直到显式 refresh。首消息提供方不会自动为 fork 重新生成标题。全部消息提供方可以在子会话出现后续提示词后追加一项归子会话所有的修订，并使用继承的合格消息和新增的合格消息。

 `ctx.sessionQuery.readTitle()` 会折叠一份实时优先或已持久化的日志，而不会在 `listSessions()` 期间加载标题。TUI 使用最新标题作为其标题栏副标题，并在完成终端安全渲染后，将终端窗口标题设置为 `<session title> — <configured product title>`。Web host 会在每个已附加会话的订阅基线之后，以及转发实时原始标题事件后立即，将同一份日志状态折叠为经过校验的 mux 控制帧。即使控制帧先于列表或会话实例创建抵达，浏览器也只保留标题事件 seq 较新的版本；侧边栏标签、搜索、面包屑和浏览器标题会随投影后的修订更新。`session.list` 仍只包含元数据，因此尚未打开的持久化会话会继续以 cwd 基名或 id 作为回退，直至打开或恢复会话时附加其日志。浏览器仅在选中已有标题的会话时将标题设置为 `<session title> — <existing HTML title>`，否则保留产品标题。报告 agent 完成情况的消费方使用核心的 `findLastMessageTurnEnd()` 折叠逻辑，因此后续的轮次间标题记录无法取代此前由消息触发的结果。

@@ -59,4 +63,4 @@ Status: implemented
 - 回退标题会立即出现。每个新建的 Web 会话都会增加一次针对首消息的辅助调用；其他组合可以自行决定更优标题是否值得模型成本，以及后续提示词是否需要重新生成会话标题。
 - 辅助请求记录和延迟接受的标题会占用事件 seq，但不会占用轮次编号，因此持久化会同时呈现尝试发起的调用与已接受的更新，尽管模型历史和 KV 缓存标识保持不变。
 - 单个提供方和每会话单调递增的修订号让释放、取代和陈旧结果拒绝行为明确可见，但多策略优先级必须由复合提供方负责。
- 手动重命名、删除、生成标题与用户标题的优先级、搜索和列表索引不在此功能范围内。
+- 删除（不经显式 refresh 的解钉）、搜索和列表索引不在此功能范围内。
@@ -1,6 +1,6 @@
 # Bilingual-pair consistency record (docs/i18n/README.md): the git blob hash of each
 # side as of the last confirmed-consistent state. Both languages carry equal authority;
 # after editing either side, bring the other along and re-record with:
-#   pnpm run verify-translation-pairing --write
-2026-07-21-tui-skill-slash-command.md: 8370ab61f552a6a60177b6da0b598dd142d21960
-2026-07-21-tui-skill-slash-command.zh.md: 66edec6ecd5c2a974b56e08bd9e924729302cc4a
+#   pnpm run verify-translation-pairing --write .agents/notes/implemented/feature/2026-07-21-tui-skill-slash-command.md
+2026-07-21-tui-skill-slash-command.md: 872e1f109728731e0d55e81a538c81e794724856
+2026-07-21-tui-skill-slash-command.zh.md: 772e25745ea7ab25f715208a9c6b1d10cf0c6e65
@@ -10,17 +10,17 @@ The [skill system](2026-07-05-skill-system.md) shipped with model-initiated load

 ## Decision

-The [`@deepseek-ai/dsh-tui`](../../../../packages/ui/tui/README.md) front door owns a `/skill:<name> [instructions]` command. On submit it loads the named skill and delivers one text block as a user turn — sent with `agent.send()` while idle and `agent.steer()` while running, the same rule as ordinary editor input. The block is `renderSkillInvocation(skill, instructions)`: a `<skill name="…">` element wrapping the skill body, preceded by one resource-base line when the provider exposes one, with the user's trailing text appended after a blank line. The command is a TUI-only affordance; it adds no model-facing tool and changes no skill-system package contract.
+The [`@deepseek-ai/dsh-tui`](../../../../packages/ui/tui/README.md) front door owns a `/skill:<name> [instructions]` command. On submit it loads the named skill and delivers one text block as a user turn — sent with `agent.send()` while idle and `agent.steer()` while running, the same rule as ordinary editor input. The block is `renderSkillInvocation(skill, instructions)`: a `<skill name="…">` element wrapping the skill body, preceded by one resource-base line when the provider exposes one, with the user's trailing text appended after a blank line. The command is a TUI-only affordance; it adds no model-facing tool. Its visibility and loading policy comes from the shared [independent model and user skill invocation policy](2026-07-28-skill-invocation-policy.md).

 The TUI reads the skill service through `ctx.get('skills')`, not a declared injection, because skills mount conditionally: a deployment without the registry keeps a working front door, and `/skill:` there reports that skills are unavailable rather than failing to mount. `createTuiChat` is synchronous while `ctx.skills.list()` is async, so autocomplete seeds the static slash commands immediately and rebuilds the provider with `skill:<name>` entries once the catalog resolves; a resolution that arrives after disposal is dropped, and a rejected lookup keeps the base commands.

-Autocomplete lists only model-invocable skills — it is built from `list()`, which omits `disableModelInvocation` skills — while manual submission resolves through `get()`, which the skill registry documents as the trusted-caller path that returns disabled skills too. So a person can load any skill by typing its exact name, but the completion menu never advertises a skill the model is meant not to see. Each completion entry is labeled with its winning source's scope — `(project)` for the `project-` sources, `(user)` for every other source — in the slash-command argument-hint slot, which the menu shows but selection never inserts, so trailing instructions still follow the completed name. An unknown name, an empty name after the prefix, and a lookup failure each surface as a transcript notice without sending anything.
+Autocomplete filters the invocation-neutral `list()` result with `isUserInvocable`, and manual submission applies the same predicate after trusted `get()` resolves the definition. A user-only skill can therefore appear and load even when model invocation is disabled, while a user-disabled skill is neither advertised nor loadable by exact name. Each completion entry is labeled with its winning source's scope — `(project)` for the `project-` sources, `(user)` for every other source — in the slash-command argument-hint slot, which the menu shows but selection never inserts, so trailing instructions still follow the completed name. An unknown name, an empty name after the prefix, a user-disabled name, and a lookup failure each surface as a transcript notice without sending anything.

 `renderSkillInvocation` and the resource-base line are the TUI's own, deliberately not reused from `dsh-tool-skill`'s `skill` tool result. The tool wraps a body in `<skill_content>`/`<skill_resources>`/`<skill_instructions>` for a *tool result*; a manual invocation is a *user turn*, and coupling the two renderers would force one model-facing shape to serve both surfaces. The cost is two renderers that both format a skill body; the benefit is that each surface's model-facing text evolves independently, and each is pinned where it is produced.

 ## Alternatives considered

-**Add a `user-invocable` frontmatter field and enforce it in the registry.** Rejected for this change. The skill-system note defers that field, and manual invocation does not need it: the TUI is a trusted local caller, so `get()` already authorizes loading any skill, and autocomplete visibility keys off the existing `disableModelInvocation`. A new per-skill field would add a contract to the registry, local provider, and tool with no current consumer beyond visibility, which `disableModelInvocation` already covers.
+**Add a `user-invocable` frontmatter field only inside the original TUI change.** Rejected there because a TUI-only field would have changed the registry, provider, and tool contract without a shared invocation model. The later [independent invocation-policy decision](2026-07-28-skill-invocation-policy.md) adds it across every relevant consumer and preserves `get()` as a trusted primitive.

 **Declare `skills` as a TUI injection.** Rejected because skills mount conditionally; a declared injection would make the front door require the registry and refuse to mount without it, contradicting the package's optional-service stance. `ctx.get('skills')` reads the global store and tolerates absence.

@@ -10,17 +10,17 @@ Status: implemented

 ## Decision

-[`@deepseek-ai/dsh-tui`](../../../../packages/ui/tui/README.md) 前门拥有一条 `/skill:<name> [instructions]` 命令。提交时它加载指定的 skill，并投递一个文本块作为用户轮次——空闲时用 `agent.send()` 发送、运行中用 `agent.steer()` 中途引导，与普通编辑器输入遵循同一规则。该文本块由 `renderSkillInvocation(skill, instructions)` 生成：一个包裹 skill 正文的 `<skill name="…">` 元素，当提供方暴露资源基址时在其前加一行资源基址行，用户尾随的文本在空行之后追加。该命令是 TUI 独有的能力；它不新增任何面向模型的工具，也不改动任何 skill 系统包的契约。
+[`@deepseek-ai/dsh-tui`](../../../../packages/ui/tui/README.md) 前门拥有一条 `/skill:<name> [instructions]` 命令。提交时它加载指定的 skill，并投递一个文本块作为用户轮次——空闲时用 `agent.send()` 发送、运行中用 `agent.steer()` 中途引导，与普通编辑器输入遵循同一规则。该文本块由 `renderSkillInvocation(skill, instructions)` 生成：一个包裹 skill 正文的 `<skill name="…">` 元素，当提供方暴露资源基址时在其前加一行资源基址行，用户尾随的文本在空行之后追加。该命令是 TUI 独有的功能；它不新增任何面向模型的工具。其可见性和加载策略来自共享的[模型与用户独立 skill 调用策略](2026-07-28-skill-invocation-policy.md)。

 TUI 通过 `ctx.get('skills')` 读取 skill 服务，而非声明式注入，因为 skill 是条件挂载的：没有注册表的部署仍保有可用的前门，此时 `/skill:` 会报告 skill 不可用，而不是挂载失败。`createTuiChat` 是同步的，而 `ctx.skills.list()` 是异步的，所以自动补全先立即种入静态斜杠命令，待目录解析完成后再用 `skill:<name>` 条目重建 provider（提供方）；在 dispose（资源释放）之后才到达的解析结果会被丢弃，而被拒绝的查找会保留基础命令。

-自动补全只列出模型可调用的 skill——它基于 `list()` 构建，而 `list()` 会略去 `disableModelInvocation` 的 skill——手动提交则通过 `get()` 解析，skill 注册表将其记录为返回被禁用 skill 的可信调用方路径。因此用户可以通过键入 skill 的确切名称加载任意 skill，但补全菜单绝不会宣传一个本不该让模型看见的 skill。每个补全条目都以其胜出来源的作用域为标签——`project-` 来源标为 `(project)`，其他一切来源标为 `(user)`——标签置于斜杠命令的参数提示位，菜单会显示它，但选中时绝不会插入，因此尾随指令仍然跟在补全后的名称之后。未知名称、前缀之后为空的名称、以及查找失败，都会各自呈现为 transcript（文本记录）中的一条通知，且不发送任何内容。
+自动补全使用 `isUserInvocable` 过滤与调用策略无关的 `list()` 结果；手动提交则在受信的 `get()` 解析定义后应用相同判定。因此，即使模型调用已禁用，仅供用户调用的 skill 仍会显示并可加载；用户禁用的 skill 既不会展示，也无法按精确名称加载。每个补全条目都以其胜出来源的作用域为标签——`project-` 来源标为 `(project)`，其他一切来源标为 `(user)`——标签置于斜杠命令的参数提示位，菜单会显示它，但选中时绝不会插入，因此尾随指令仍然跟在补全后的名称之后。未知名称、前缀之后为空的名称、用户禁用的名称以及查找失败，都会各自呈现为 transcript（文本记录）中的一条通知，且不发送任何内容。

 `renderSkillInvocation` 及资源基址行是 TUI 自有的，刻意不复用 `dsh-tool-skill` 的 `skill` 工具结果。该工具把正文包进 `<skill_content>`/`<skill_resources>`/`<skill_instructions>` 是为了一个*工具结果*；而手动调用是一个*用户轮次*，把两个渲染器耦合起来会迫使一种面向模型的形态同时服务两个界面。代价是两个都在格式化 skill 正文的渲染器；收益是各界面面向模型的文本可以独立演进，且各自在其产出处被固定。

 ## Alternatives considered

-**新增 `user-invocable` frontmatter 字段并在注册表中强制执行。** 本次改动否决。skill 系统 note 把该字段列为待办，而手动调用并不需要它：TUI 是可信的本地调用方，`get()` 已经授权加载任意 skill，自动补全的可见性以既有的 `disableModelInvocation` 为准。新增一个逐 skill 字段会给注册表、本地提供方和工具都加上一条契约，而除了可见性之外没有任何现有消费方，可见性又已由 `disableModelInvocation` 覆盖。
+**仅在最初的 TUI 变更内新增 `user-invocable` frontmatter 字段。** 当时未采纳，因为 TUI 独有的字段会在没有共享调用模型的情况下改变注册表、提供方和工具契约。后续的[独立调用策略决策](2026-07-28-skill-invocation-policy.md)将其扩展到每个相关消费方，并保留 `get()` 作为受信原语。

 **把 `skills` 声明为 TUI 注入。** 否决，因为 skill 是条件挂载的；声明式注入会使前门必须依赖注册表，缺少它就拒绝挂载，与本包可选服务的立场相悖。`ctx.get('skills')` 读取全局存储并容忍其缺失。

@@ -0,0 +1,6 @@
+# Bilingual-pair consistency record (docs/i18n/README.md): the git blob hash of each
+# side as of the last confirmed-consistent state. Both languages carry equal authority;
+# after editing either side, bring the other along and re-record with:
+#   pnpm run verify-translation-pairing --write .agents/notes/implemented/feature/2026-07-22-docked-web-goal-bar.md
+2026-07-22-docked-web-goal-bar.md: 110aea299a260896b0098f10b337734e0c0aebcf
+2026-07-22-docked-web-goal-bar.zh.md: cc0a5eda6815e6c02e97fd02197659764d4f2d69
@@ -0,0 +1,39 @@
+# Agent Note: Docked web goal bar
+
+Status: implemented
+
+English | [中文](2026-07-22-docked-web-goal-bar.zh.md)
+
+## Problem
+
+The web UI had no goal surface at all: the goal stack shipped with model tools, the TUI/ACP adapters, and the `/goal` command, but the browser client exposed none of it — no runtime verbs, no indicator. This change introduces the client goal verbs (runtime session methods over RPC) and the first goal UI together. Placement follows the redesign's premise that goal presence belongs to the composer's context: the goal is a property of the work the user is about to prompt, so its indicator docks directly above the message composer as a rounded-top strip tucked under the composer card's top edge. The mock keeps only a sparkle, a phase word ("Ongoing/Paused/Blocked Goal"), the truncated objective, and edit/clear icon actions, with resume appearing only on a paused goal.
+
+## Decision
+
+`GoalBar` (`packages/client/ui-goal/src/client/GoalBar.tsx`) is a new props-driven, self-contained component; `ConversationRoot` mounts it immediately before the composer `InputBar`. The strip's CSS mirrors the composer's horizontal geometry (32px side padding, 776px centered cap) plus the mock's 12px inset, and a -10px bottom margin eats InputBar's 8px top padding and tucks its square bottom edge 2px under the composer card's top edge. All strip states share one fixed 38px height so switching between them never resizes it. Loading (`goal === undefined`), absent (`goal === null`), and `phase === 'complete'` render nothing — a completed goal is history, not chrome.
+
+Visibility drives the label and actions: active shows "Ongoing Goal" with pause/edit/clear; paused shows "Paused Goal" and swaps pause for a resume icon button; blocked shows "Blocked Goal" and carries `blockedReason.message` as the strip's `title` tooltip. Goal creation lives on the `/goal` command, not in the bar. The pencil swaps the strip for an inline edit form prefilled with the current objective: Enter or the check button saves through `GoalBarActions.onEdit(objective)`, Esc cancels, and an all-whitespace objective keeps save disabled. The form closes only when the edit succeeds; a failure preserves the draft and displays the error in the bar. Resume and clear failures are displayed there as well. Clear otherwise calls `onClear` directly with no confirmation — a clear keeps a durable tombstone, so nothing is unrecoverable. An effect keyed on the goal's id drops the edit form when the goal's identity changes, so a surviving draft can never be written over the goal that replaced it.
+
+`GoalBarActions` lives in ui-goal's slot contract (`packages/client/ui-goal/src/client/slots.ts`) and carries exactly the rendered verbs: `onEdit`/`onPause`/`onResume`/`onClear`. Each callback asynchronously returns an explicit success/failure result so `GoalBar` owns its transitions and error display. `apply.ts` wires them to the runtime session methods; the runtime session resolves the current goal's compare-and-set ref internally, so the UI passes no ref.
+
+The runtime session gains the goal surface the strip (and future UI) needs: `fetchGoal` populates the snapshot on open, and a live `context/message` carrying `goal/change` meta triggers a coalesced refetch — concurrent triggers share the in-flight `goal.get`, while a trigger received during that read schedules one coalesced trailing read so independently ordered notifications and GET responses cannot leave stale state. Window replays never refetch, and matching the meta kind (rather than a goal key) also catches clear tombstones written by other clients. The six mutation verbs fold transport failures into `{ ok: false }` results like every sibling session method, and a get result older than a mutation response that landed mid-flight is dropped.
+
+The strip's background is `--dsw-alias-interactive-bg-hover` rather than the mock's literal `#F5F6F7`: the translucent hover gray resolves to that value over the white light-theme base and lifts the strip off the composer card in dark mode, where a static light token would sink. All colors are `--dsw-*` tokens.
+
+## Testing
+
+`packages/client/ui-goal/tests/goalbar.spec.tsx` pins the behavior through props alone: loading/absent/complete render nothing, the active strip renders label/objective and fires clear, the edit form prefills, rejects empty, saves on Enter, cancels on Esc, and resets when the goal's identity changes, the active strip fires pause, the paused strip fires resume, and the blocked strip exposes the reason tooltip. Component failure-path cases prove that a failed edit preserves its draft and that edit/resume/clear errors remain visible in the bar. The skeleton specs mount `ConversationRoot` with and without `goalActions`; the undefined case is seeded with an active goal, so the missing gate — not the missing goal — is what hides the strip. Runtime session specs pin the folded-error results, the live-only in-flight-plus-trailing refetch, and the stale-read guard. A keyless real-browser smoke boots the assembled application through `boot → RPC → runtime → GoalBar` and records an inline snapshot of the rendered label, objective, and actions.
+
+## Alternatives considered
+
+- **Put the strip in the session header** — rejected because the redesign's premise is that goal presence belongs to the composer's context; a header strip cannot dock into the composer card.
+- **Render a "Loading goal…" placeholder for `undefined`** — rejected: the strip would flash and collapse on every session open, chrome noise for a sub-second state.
+- **Include an inline create affordance when no goal is set** — rejected after implementation review: goal creation lives on the `/goal` command, matching the pattern where the model creates goals on request; the bar is a status indicator, not a creation surface.
+- **Carry the full verb set (`onComplete` included) in `GoalBarActions`** — rejected as speculative generality: the interface carries only the rendered verbs (`onPause` joined it when the active strip gained its pause action).
+
+## Consequences
+
+- Goal presence in the web UI is a composer-docked strip: sparkle, phase label, truncated objective, and pause/edit/clear (resume replacing pause when paused) — the browser client's first goal surface.
+- The runtime session exposes the goal verbs over RPC with folded transport errors, and refreshes the snapshot's goal on open and on live goal-change meta (coalesced, guarded against stale reads).
+- Objective editing is reachable from the UI for the first time, through `goal.edit` with the runtime-owned ref; complete remains available to other surfaces (`/goal`, model tools).
+- `goal === null` renders nothing; the composer carries no persistent create affordance — creation is the `/goal` command's job.
@@ -0,0 +1,39 @@
+# Agent Note: 停靠式 Web 目标条
+
+Status: implemented
+
+[English](2026-07-22-docked-web-goal-bar.md) | 中文
+
+## 问题
+
+Web UI 此前没有任何目标相关的界面：目标栈已随模型工具、TUI/ACP 适配器和 `/goal` 命令交付，但浏览器客户端完全不接触它——既没有运行时动词，也没有指示器。本变更同时引入客户端目标动词（基于 RPC 的运行时会话方法）和第一个目标 UI。摆放位置遵循重新设计的前提：目标的存在感属于输入框的上下文——目标是用户即将提交的工作的属性，因此它的指示器停靠在消息输入框正上方，呈现为一条圆角顶部的横条，收进输入框卡片顶边之下。设计稿只保留一个闪光图标、一个阶段词（"Ongoing/Paused/Blocked Goal"）、截断后的目标内容，以及编辑／清除图标操作，恢复按钮仅在目标暂停时出现。
+
+## 决策
+
+`GoalBar`（`packages/client/ui-goal/src/client/GoalBar.tsx`）是一个新的、由 props 驱动的自包含组件；`ConversationRoot` 将它挂载在输入框 `InputBar` 紧上方。横条的 CSS 对齐输入框的水平几何（两侧 32px 内边距、776px 居中上限），再加上设计稿的 12px 内缩，并用 -10px 的下外边距吃掉 InputBar 的 8px 上内边距，使它方形的底边收进输入框卡片顶边之下 2px。横条的所有状态共享固定的 38px 高度，状态切换不会引起尺寸变化。加载中（`goal === undefined`）、无目标（`goal === null`）和 `phase === 'complete'` 时不渲染任何内容：已完成的目标是历史记录，不是常驻界面元素。
+
+可见性决定标签和操作：active 状态显示 "Ongoing Goal" 并提供暂停／编辑／清除；paused 状态显示 "Paused Goal"，把暂停换成一个恢复图标按钮；blocked 状态显示 "Blocked Goal"，并把 `blockedReason.message` 作为横条的 `title` 悬浮提示。创建目标的入口在 `/goal` 命令上，不在横条里。铅笔图标把横条切换为内联编辑表单，预填当前目标内容：Enter 或勾选按钮通过 `GoalBarActions.onEdit(objective)` 保存，Esc 取消，目标内容全为空白字符时保存按钮保持禁用。编辑成功后表单才会关闭；编辑失败时保留草稿，并在横条中显示错误。恢复和清除失败也显示在横条中。除此之外，清除直接调用 `onClear`，不做确认——清除会保留 durable 墓碑，没有不可恢复的损失。一个以目标 id 为键的 effect 会在目标身份变化时丢弃编辑表单，因此存留的草稿绝不可能覆盖掉替换它的新目标。
+
+`GoalBarActions` 位于 ui-goal 的槽位契约（`packages/client/ui-goal/src/client/slots.ts`），只携带实际渲染的动词：`onEdit`/`onPause`/`onResume`/`onClear`。每个回调都会异步返回显式成功／失败结果，因此 `GoalBar` 自行负责界面转换和错误显示。`apply.ts` 把它们接到运行时会话方法上；运行时会话在内部解析当前目标的 compare-and-set ref，因此 UI 不传 ref。
+
+运行时会话获得了横条（以及未来 UI）所需的目标表面：`fetchGoal` 在打开时填充快照；携带 `goal/change` 元数据的 live `context/message` 触发合并重新拉取——并发触发器共享正在执行的 `goal.get`，读取期间收到的触发器会安排一次合并后的尾随读取，避免彼此独立排序的通知和 GET 响应留下陈旧状态。窗口重放绝不触发重新拉取，且匹配元数据 kind（而不是 goal 键）还能捕获其他客户端写入的清除墓碑。六个变更动词与所有同类会话方法一样，把传输层失败折叠为 `{ ok: false }` 结果；比在拉取途中落地的变更响应更旧的 get 结果会被丢弃。
+
+横条的背景色用 `--dsw-alias-interactive-bg-hover`，而不是设计稿里的字面值 `#F5F6F7`：这个半透明的悬浮灰在浅色主题的白色底上正好解析为该值，而在深色模式下能把横条从输入框卡片上衬托出来，静态的浅色 token 在深色模式下会沉进去。所有颜色都是 `--dsw-*` token。
+
+## 测试
+
+`packages/client/ui-goal/tests/goalbar.spec.tsx` 仅通过 props 固定这些行为：加载中／无目标／已完成时不渲染；active 横条渲染标签和目标内容并触发清除；编辑表单预填内容、拒绝空值、按 Enter 保存、按 Esc 取消，并在目标身份变化时重置；active 横条触发暂停；paused 横条触发恢复；blocked 横条暴露原因悬浮提示。组件失败路径用例证明编辑失败时保留草稿，并且编辑／恢复／清除错误持续显示在横条中。skeleton 规格测试分别挂载带与不带 `goalActions` 的 `ConversationRoot`；未定义的情形预置了一个 active 目标，因此隐藏横条的是缺失的挂载门，而不是缺失的目标。运行时会话规格测试固定了折叠错误结果、仅 live 的执行中读取加尾随读取，以及陈旧读取守卫。一个无密钥真实浏览器冒烟测试通过 `boot → RPC → runtime → GoalBar` 启动组装后的应用，并以内联快照记录渲染出的标签、目标内容和操作。
+
+## 考虑过的替代方案
+
+- **把横条放在会话头部**：不予采纳，因为重新设计的前提是目标的存在感属于输入框的上下文；放在头部的横条无法停靠进输入框卡片。
+- **为 `undefined` 渲染 "Loading goal…" 占位**：不予采纳，每次打开会话横条都会闪现再坍缩，对一个不到一秒的状态来说只是界面噪音。
+- **未设置目标时在横条内提供内联创建入口**：实现评审后不予采纳，创建目标的职责在 `/goal` 命令上，与模型按请求创建目标的模式一致；横条是状态指示器，不是创建入口。
+- **在 `GoalBarActions` 中携带完整动词集合（含 `onComplete`）**：作为投机性泛化不予采纳，接口只携带实际渲染的动词（active 横条获得暂停操作后，`onPause` 随之加入）。
+
+## 后果
+
+- Web UI 中目标的存在形式是停靠在输入框上方的横条：闪光图标、阶段标签、截断的目标内容，以及暂停／编辑／清除（暂停时恢复取代暂停）——这是浏览器客户端的第一个目标界面。
+- 运行时会话通过 RPC 暴露目标动词并折叠传输层错误，且在打开时和 live 目标变更元数据到达时刷新快照中的目标（合并拉取，带陈旧读取守卫）。
+- 目标内容首次可以从 UI 编辑，经由 `goal.edit`，ref 由运行时持有；完成对其他界面（`/goal`、模型工具）照常可用。
+- `goal === null` 时不渲染任何内容；输入框不提供常驻的创建入口，创建是 `/goal` 命令的职责。
@@ -0,0 +1,6 @@
+# Bilingual-pair consistency record (docs/i18n/README.md): the git blob hash of each
+# side as of the last confirmed-consistent state. Both languages carry equal authority;
+# after editing either side, bring the other along and re-record with:
+#   pnpm run verify-translation-pairing --write
+2026-07-23-web-permission-and-approval.md: cd402a039e55e7a24a038055dab5793aa0d08438
+2026-07-23-web-permission-and-approval.zh.md: ce4964789bc94a0962796bb2f5fbf1a94e8f5145
@@ -0,0 +1,33 @@
+# Agent Note: Web UI permission presets and approval answering
+
+Status: implemented
+
+English | [中文](2026-07-23-web-permission-and-approval.zh.md)
+
+## Problem
+
+The web host booted an unconfined agent: `bootHost` composed `dsh-bash-local` and `dsh-fs-local`, so every web session ran with full file access, no approval channel, and no permission control — while the ACP composition had shipped the complete sandboxed product path (sandbox provider + policy home + confined bash/fs + approval + presets) for months. The web wire contract had already reserved the seats — `approval/requested`/`approval/resolved` mux frames, `POST /api/respond` with `ApprovalResponsePayload`, client-side `pendingBuffers` — but the host `respond` was a stub, no answerer bridged `ctx.approval` to the stream, no RPC exposed the permission select, and the PendingCard rendered approvals as visible-but-unanswerable.
+
+## Decision
+
+The web host composes the same sandboxed product path as the acp-agent composition: `dsh-sandbox-local`, `dsh-sandbox-policy`, `dsh-bash-sandbox`, `dsh-fs-sandbox`, `dsh-user-approval`, and `dsh-permission`, with `BootHostOptions.sandbox` supplying the deployment defaults (`mode`, default `workspace-write`; `approvalPolicy`, default `ask`).
+
+`createApiProxy` owns the approval pending registry. Its `approval/request` waterfall answerer reads the approval id from the session's just-appended `approval/asked` audit event (an ask with no audit event is a foreign channel and delegates), mints one stable rpcId per question, broadcasts the answerable `approval/requested` frame to every open mux stream, and replays still-pending frames verbatim on each mux open — the refresh-recovery baseline the contract already promised. `respond` routes by the echoed rpcId, validates `ApprovalResponsePayload` with the existing zod schema, cross-checks the payload's audit correlation against the routed entry, resolves the answerer, and broadcasts `approval/resolved`; the ask's abort signal withdraws the question as `cancelled`.
+
+The permission select rides two new unary RPCs, `session.permissions` and `session.setPermission`, projecting `ctx.permission` into a protocol-owned `PermissionOption` DTO (the ACP bridge precedent: each protocol owns its presentation shape). A permission-less composition serves an empty select and clients hide the control. Idle switches are held last-write-wins in a proxy-side pending map and flushed on `agent/prompt-submit`, because knob events must stay turn-enclosed for durable replay; the shared `hasOpenTurn` fold moved to `dsh-session` and replaced the private copies in `dsh-user-approval`, the ACP bridge, and the proxy.
+
+Client-side, `Session` gained `permissions` and `setPermission`, and approval answering rides the runtime's `PendingWait` carrier. Per the designer draft, a pending approval takes over the composer: `ApprovalPanel` registers as a selector-routed entry of the conversation-declared `conversation.composer` chain (the ui-question pattern), replacing the InputBar with the justification headline, the paired command, and one-shot refuse/allow buttons; the `PendingApproval` domain face in ui-conversation's contract owns the `ApprovalResponsePayload` wire encoding over the carrier, and the broadcast resolved frame settles the wait and restores the composer. Question placeholders stay in the message flow. The sidebar mirrors the blocked state with an amber warning dot that outranks the running ring: the manager tracks per-session outstanding approvalIds (idempotent under mux-open replays, cleared per connection generation so the reopen replay is authoritative) rather than reading Session instances, so the dot lights for sessions never instantiated. The composer's bottom-row chip hosts the `PermissionSelect` control fed through the conversation inject face. The connection fixture mirrors the host: its resident approval is answerable once, and its permission select persists per session.
+
+## Alternatives considered
+
+**Reuse the ACP `session/set_config_option` shape on the web wire.** Rejected: the web contract's unary method registry (`RpcMethodMap` + per-method zod schemas) is its own dialect; a generic config-option surface would bypass the compiler-locked schema table for one select. A dedicated method pair keeps both sides derivable from the signature.
+
+**A session event for pending approvals instead of a proxy-side registry.** Rejected: approval requests are transient interaction state, not durable session data — the `approval/asked`/`decided` audit pair already logs the durable half. Persisting requested frames would re-ask dead questions on replay.
+
+**Registering the answerer only when a mux subscriber exists.** Rejected: the pending entry must survive client disconnects (refresh recovery is the point), so the registry outlives any one stream; a subscriber-gated answerer would fail asks closed during a reload window.
+
+**Optimistic card removal on click.** Rejected: the broadcast resolved frame is the truth; removing on click would hide a question that a rejected receipt or transport failure left standing. The panel disables its buttons locally and re-arms them on failure instead.
+
+## Consequences
+
+Web sessions now start confined (`workspace-write` + `ask` by default) and a sandbox-denial escalation reaches the browser as an answerable card; the deployment can widen or narrow the default through `BootHostOptions.sandbox` without touching the assembly. Question answering shipped separately through the same registry pattern (ui-question over the question pending table). The permission select reads once per mount; live refresh from another client's switch is deferred. Coverage: proxy registry and permission RPC unit suites, session-object and fixture unit suites, and the keyless web smoke exercises the fixture-mode approval answer and preset switch in a real browser.
@@ -0,0 +1,33 @@
+# Agent Note: Web UI 权限预设与审批应答
+
+Status: implemented
+
+[English](2026-07-23-web-permission-and-approval.md) | 中文
+
+## 问题
+
+Web 承载层启动的是一个不受限的 agent（智能体）：`bootHost` 组合了 `dsh-bash-local` 与 `dsh-fs-local`，因此每个 Web 会话都以完整文件访问权限运行，既无审批通道，也无权限管控——而 ACP 组合早在数月前就已交付完整的沙箱化产品路径（沙箱提供方 + 策略归属 + 受限的 bash/fs + 审批 + 预设）。Web 协议契约其实早已预留了对应位置——`approval/requested`/`approval/resolved` 的 mux 帧、携带 `ApprovalResponsePayload` 的 `POST /api/respond`、client 侧的 `pendingBuffers`——但 host 的 `respond` 只是一个 stub，没有应答者把 `ctx.approval` 桥接到流上，没有 RPC 暴露权限选择，PendingCard 把审批渲染成可见却无法应答的样子。
+
+## 决策
+
+Web 承载层组合与 acp-agent 相同的沙箱化产品路径：`dsh-sandbox-local`、`dsh-sandbox-policy`、`dsh-bash-sandbox`、`dsh-fs-sandbox`、`dsh-user-approval` 与 `dsh-permission`，由 `BootHostOptions.sandbox` 提供部署默认值（`mode`，默认 `workspace-write`；`approvalPolicy`，默认 `ask`）。
+
+`createApiProxy` 拥有审批 pending 注册表。它的 `approval/request` waterfall（瀑布式事件）应答者从会话刚追加的 `approval/asked` 审计事件中读取审批 id（没有审计事件的 ask 属于外部通道，予以委托），为每个问题 mint 一个稳定的 rpcId，向每个打开的 mux 流广播可应答的 `approval/requested` 帧，并在每次 mux 打开时原样重放仍处于 pending 的帧——这正是契约早已承诺的刷新恢复基线。`respond` 按回显的 rpcId 路由，用既有的 zod schema 校验 `ApprovalResponsePayload`，将载荷的审计关联与所路由的条目交叉核对，解析应答者，并广播 `approval/resolved`；ask 的中断信号会以 `cancelled` 撤回该问题。
+
+权限选择依托两个新的一元 RPC，`session.permissions` 与 `session.setPermission`，把 `ctx.permission` 投影为一个由协议拥有的 `PermissionOption` DTO（沿用 ACP bridge 的先例：每个协议拥有自己的呈现形状）。无权限的组合提供空的选择项，client 隐藏该控件。空闲期的切换以后写胜出（last-write-wins）的方式保存在 proxy 侧的 pending map 中，并在 `agent/prompt-submit` 时冲刷，因为旋钮事件必须保持轮次内闭合以支持持久回放；共享的 `hasOpenTurn` 折叠迁入 `dsh-session`，取代了 `dsh-user-approval`、ACP bridge 与 proxy 中各自的私有副本。
+
+在 client 侧，`Session` 新增了 `permissions` 与 `setPermission`，审批应答则依托运行时的 `PendingWait` 载体。按照设计师草稿，处于 pending 的审批会接管 composer：`ApprovalPanel` 注册为由会话声明的 `conversation.composer` 链中一个按选择器路由的条目（即 ui-question 模式），以理由标题、配对的命令与一次性的拒绝／允许按钮取代 InputBar；ui-conversation 契约中的 `PendingApproval` 领域面拥有 `ApprovalResponsePayload` 在该载体上的协议编码（wire encoding），广播的 resolved 帧使该等待落定并恢复 composer。问题占位符仍留在消息流中。侧边栏用一枚琥珀色警示圆点同步呈现这一阻塞状态，且其优先级高于表示运行中的圆环：manager 跟踪每个会话尚未解决的 approvalId（对 mux 打开时的回放幂等，并按连接代次清除，以保证重开后的回放才是权威依据），而非读取 Session 实例，因此从未实例化过的会话也能点亮该圆点。composer 底行的 chip 经会话注入面挂载 `PermissionSelect` 控件。连接 fixture（测试前置数据）与 host 保持一致：它的常驻审批可应答一次，其权限选择项按会话持久保存。
+
+## 曾考虑的替代方案
+
+**在 Web 协议上复用 ACP 的 `session/set_config_option` 形状。** 不予采纳：Web 契约的一元方法注册表（`RpcMethodMap` + 逐方法的 zod schema）是它自成一体的方言；一个通用的 config-option 接口会为一个选择项绕开编译期锁定的 schema 表。一对专用方法让两侧都能从签名推导得出。
+
+**用一个会话事件承载 pending 审批，而非 proxy 侧注册表。** 不予采纳：审批请求是瞬态的交互状态，而非持久的会话数据——`approval/asked`/`decided` 审计对已经记录了持久的那一半。持久化 requested 帧会在回放时重新问出已经作废的问题。
+
+**仅在存在 mux 订阅者时才注册应答者。** 不予采纳：pending 条目必须在 client 断连后依然存活（刷新恢复正是要点所在），因此注册表的生命周期长于任何单个流；一个受订阅者门控的应答者，会让在重载窗口期间关闭的 ask 落空。
+
+**点击即乐观移除卡片。** 不予采纳：广播的 resolved 帧才是真相；点击即移除会隐藏一个因拒绝回执或传输失败而仍然悬置的问题。面板改为在本地禁用其按钮，并在失败时重新启用。
+
+## 后果
+
+Web 会话现在从受限状态启动（默认 `workspace-write` + `ask`），一次沙箱拒绝的升级会以可应答的卡片形式抵达浏览器；部署方可以通过 `BootHostOptions.sandbox` 放宽或收紧默认值，无需触动装配。问题应答已通过同一注册表模式单独交付（ui-question 基于问题 pending 表）。权限选择在每次挂载时读取一次；来自另一个 client 切换的实时刷新暂缓实现。覆盖情况：proxy 注册表与权限 RPC 的单元测试套件、会话对象与 fixture 的单元测试套件，以及无密钥 Web 冒烟测试在真实浏览器中演练 fixture 模式的审批应答与预设切换。
@@ -2,5 +2,5 @@
 # side as of the last confirmed-consistent state. Both languages carry equal authority;
 # after editing either side, bring the other along and re-record with:
 #   pnpm run verify-translation-pairing --write .agents/notes/implemented/feature/2026-07-23-web-todo-display.md
-2026-07-23-web-todo-display.md: 5fe08cc40c1d23ff3a9b8c6d766fea6d3694c30d
-2026-07-23-web-todo-display.zh.md: c121ffc27e3d0a93707c2c22b2f180023ebae5be
+2026-07-23-web-todo-display.md: 7223ff9adbf1fa6dca39c9eb4949b6d3861bdd6d
+2026-07-23-web-todo-display.zh.md: ac773a3c587183aa9a152caff2812686bc79f3c7
@@ -14,7 +14,7 @@ Consume `todo/write` as a Session side effect, not a surface node, and render it

 ### Side-effect channel, converging with window replay

-`applyEventSideEffects` gains a `todo/write` case (whole list, last write wins). Unlike partial/openCalls, `rebuildDerivedFromWindow` deliberately does NOT reset it: the value is session-level — taken from the tail history page's full-log projection — and an arbitrary window may not contain the latest write, so an older-page prepend keeps it and only an in-window or live write overwrites it. Every `installWindow` caller is a tail request (`doOpen`, its gap re-pull, `repairGap`; `loadOlder` prepends without it), which the host answers with the projection or omits it only when the full log holds no `todo/write` — so an absent field is the authoritative empty list and is assigned as such. That distinction matters on rollback: a live write whose host crashed before persisting leaves the log empty, and preserving the prior value instead would strand the rolled-back plan on screen indefinitely. `ConversationSnapshot.todos` is the read surface. This follows the event's own contract ("log-only UI state; never derived history"): surfacing each write as a conversation node would render superseded lists as if they were still standing.
+`applyEventSideEffects` gains a `todo/write` case (whole list, last write wins) and clears on `turn/start` ([turn-scoped plan lifetime](2026-07-28-todo-plan-clears-on-next-turn.md)). `rebuildDerivedFromWindow` sweeps the window from an empty plan and restores the tail-page seed only when the window never determined the plan (no `todo/write` and no `turn/start`); otherwise the in-window write/`turn/start` fold wins. Every `installWindow` caller is a tail request (`doOpen`, its gap re-pull, `repairGap`; `loadOlder` prepends without reseeding), which the host answers with the projection or omits it when no plan stands — so an absent field is the authoritative empty list and is assigned as such. That distinction matters on rollback: a live write whose host crashed before persisting leaves the log empty, and preserving the prior value instead would strand the rolled-back plan on screen indefinitely. `ConversationSnapshot.todos` is the read surface. This follows the event's own contract ("log-only UI state; never derived history"): surfacing each write as a conversation node would render superseded lists as if they were still standing.

 ### TodoPanel: the durable list as a persistent strip

@@ -33,4 +33,4 @@ The dedicated `todo_write` chat row is a plain registrant plugin (`todoToolview`

 ## Consequences

-Replay correctness is owned by one code path: any future change to window rebuild keeps todos consistent for free, and the fixture (fx-alpha turn 65) plus the assembled keyless snapshot (`apps/web/tests/todo-display.snapshot.ts`) pin the full chain (row summary and state, dock panel content, collapse round-trip) over the built client graph. `todos` is a required `ConversationSnapshot` field, so scripted fakes in specs must carry it. The TUI panel is untouched (the automation-only ACP bridge deliberately omits todo presentation); the web surfaces render the same event, adding one wire field and no new event type. That field is how cold-load reconstruction stays host-backed: the tail history page carries `todos` — the full-log latest `todo/write`, computed independently of the page window (the same backscan posture the view pairing uses) — so a reopened session restores the plan even when the last write precedes the window; that value survives an older-page prepend, is overwritten by any later write, and resets to empty when a tail response carries no projection.
+Replay correctness is owned by one code path: any future change to window rebuild keeps todos consistent for free, and the fixture (fx-alpha turn 65) plus the assembled keyless snapshot (`apps/web/tests/todo-display.snapshot.ts`) pin the full chain (row summary and state, dock panel content, collapse round-trip) over the built client graph. `todos` is a required `ConversationSnapshot` field, so scripted fakes in specs must carry it. The TUI panel shares the same turn-scoped lifetime (the automation-only ACP bridge deliberately omits todo presentation); the web surfaces render the same event, adding one wire field and no new event type. That field is how cold-load reconstruction stays host-backed: the tail history page carries `todos` — the full-log standing plan (latest `todo/write` with no later `turn/start`), computed independently of the page window (the same backscan posture the view pairing uses) — so a reopened session restores the plan when it still stands and the last write precedes the window; that value survives an older-page prepend, is overwritten by any later write, clears on a later `turn/start`, and resets to empty when a tail response carries no projection.
@@ -14,7 +14,7 @@ Status: implemented

 ### 副作用通道，与窗口回放收敛

-`applyEventSideEffects` 新增一个 `todo/write` 分支（整份列表，后写覆盖先写）。与 partial/openCalls 不同，`rebuildDerivedFromWindow` 刻意不重置它：该值是会话级的——取自尾页 history 携带的全量 log 投影——而任意窗口未必包含最近一次写入，因此往前翻页保留它，只有窗口内或实时的写入才会覆盖。`installWindow` 的每个调用方都是尾页请求（`doOpen`、其补洞重拉、`repairGap`；`loadOlder` 只往前拼接、不走它），而 host 对尾页请求要么带上投影、要么仅在全量 log 没有任何 `todo/write` 时省略——因此字段缺失就是权威的空列表，直接照此赋值。这个区分在回滚场景上要紧：实时写入若在 host 持久化前崩溃，log 里就是空的，此时保留旧值会让已回滚的计划永远留在屏幕上。`ConversationSnapshot.todos` 是读取面。这遵循事件自身的契约（「仅日志 UI 状态，绝非派生历史」）：把每次写入作为对话节点呈现，会让已被取代的列表看起来仍然有效。
+`applyEventSideEffects` 新增一个 `todo/write` 分支（整份列表，后写覆盖先写），并在 `turn/start` 清空（[按 turn 界定的计划生命周期](2026-07-28-todo-plan-clears-on-next-turn.md)）。`rebuildDerivedFromWindow` 从空计划扫过窗口，仅当窗口从未判定计划（无 `todo/write` 且无 `turn/start`）时恢复尾页种子；否则以窗口内写入／`turn/start` 折叠为准。`installWindow` 的每个调用方都是尾页请求（`doOpen`、其补洞重拉、`repairGap`；`loadOlder` 只往前拼接、不再播种），而 host 对尾页请求要么带上投影、要么在无站立计划时省略——因此字段缺失就是权威的空列表，直接照此赋值。这个区分在回滚场景上要紧：实时写入若在 host 持久化前崩溃，log 里就是空的，此时保留旧值会让已回滚的计划永远留在屏幕上。`ConversationSnapshot.todos` 是读取面。这遵循事件自身的契约（「仅日志 UI 状态，绝非派生历史」）：把每次写入作为对话节点呈现，会让已被取代的列表看起来仍然有效。

 ### TodoPanel：长驻列表作为一条常驻横条

@@ -33,4 +33,4 @@ Status: implemented

 ## Consequences

-回放正确性由一条代码路径掌管：未来对窗口重建的任何改动都免费保持 todos 一致；fx-alpha 第 65 轮的 fixture（测试前置数据）加 assembled keyless snapshot（`apps/web/tests/todo-display.snapshot.ts`）在构建产物客户端全图上钉住整条链（行摘要与状态、dock 面板内容、折叠往返）。`todos` 是 `ConversationSnapshot` 的必填字段，所以 spec 里脚本化的 fake 必须带上它。TUI 面板未受改动（自动化专用的 ACP 桥接刻意不做 todo 呈现）；Web 各面渲染同一个事件，只新增一个协议字段，不新增事件类型。冷加载重建正是靠这个字段由 host 兜底：history 尾页附带 `todos`——全量 log 上最新一次 `todo/write` 的投影，独立于分页窗口计算（与 view 配对同一种 backscan 姿势）——因此重开会话时即使最后一次写入落在窗口之前，计划也照常恢复；该值跨往前翻页保留，之后的任何写入照常覆盖，而尾页响应不带投影时复位为空。
+回放正确性由一条代码路径掌管：未来对窗口重建的任何改动都免费保持 todos 一致；fx-alpha 第 65 轮的 fixture（测试前置数据）加 assembled keyless snapshot（`apps/web/tests/todo-display.snapshot.ts`）在构建产物客户端全图上钉住整条链（行摘要与状态、dock 面板内容、折叠往返）。`todos` 是 `ConversationSnapshot` 的必填字段，所以 spec 里脚本化的 fake 必须带上它。TUI 面板共用同一按 turn 界定的生命周期（自动化专用的 ACP 桥接刻意不做 todo 呈现）；Web 各面渲染同一个事件，只新增一个协议字段，不新增事件类型。冷加载重建正是靠这个字段由 host 兜底：history 尾页附带 `todos`——全量 log 上的站立计划（其后没有更晚 `turn/start` 的最近一次 `todo/write`），独立于分页窗口计算（与 view 配对同一种 backscan 姿势）——因此重开会话时若计划仍站立且最后一次写入落在窗口之前，计划也照常恢复；该值跨往前翻页保留，之后的任何写入照常覆盖，更晚的 `turn/start` 会清空，而尾页响应不带投影时复位为空。
@@ -0,0 +1,6 @@
+# Bilingual-pair consistency record (docs/i18n/README.md): the git blob hash of each
+# side as of the last confirmed-consistent state. Both languages carry equal authority;
+# after editing either side, bring the other along and re-record with:
+#   pnpm run verify-translation-pairing --write .agents/notes/implemented/feature/2026-07-25-subagent-policy-inheritance.md
+2026-07-25-subagent-policy-inheritance.md: aeff83795eedead9c75de6bbb74c1da1945092ca
+2026-07-25-subagent-policy-inheritance.zh.md: 7a09fd972f53b46655df93ddf9625455f2077460
@@ -0,0 +1,36 @@
+# Agent Note: In-process subagent policy inheritance — the child starts under the parent's sandbox and approval overrides
+
+Status: implemented
+
+English | [中文](2026-07-25-subagent-policy-inheritance.zh.md)
+
+## Problem
+
+Sandbox and approval overrides are per-session log folds. An in-process subagent gets a new session, so a spawn child once fell back to deployment defaults and a fork child saw only switches inside its completed-turn prefix. Delegation could therefore widen a parent that had switched to `read-only`, or turn a parent's unattended `'never'` approval stance back into prompting behavior.
+
+## Decision
+
+The shared in-process driver snapshots `sandboxPolicy.overrideOf(parent.session)` and `approval.overrideOf(parent.session)` before its first await. A later parent switch belongs to the parent's future; cancel-and-redelegate takes a new snapshot. Both services are optional, and only explicit session overrides are copied, never deployment defaults or one-shot grants.
+
+Each captured value becomes a source-tagged `sandbox/mode` or `approval/policy` event appended during the child factory's unpublished setup. The session constructor has already fixed `Session.firstLiveSeq` at the fork-prefix length, so the inherited facts follow fork history, reach telemetry when the child is announced, and leave `SessionHeader.seedLength` at the prefix length. Existing last-event-wins folds therefore make the delegation snapshot beat stale fork history and let a later child switch beat the snapshot. A grandchild folds its parent's logged state, so the rule composes without another inheritance mechanism.
+
+Ordinary session appends validate the inherited events before publication, and persistence captures the complete unpublished log when the session is announced. Any materialized child log therefore stores the inherited events with its first batch; there is no second policy store, schema field, or query index. The `source: 'delegation'` marker lets approval narration distinguish inheritance from a child-side user switch.
+
+### What a blocked child experiences
+
+A confined child gets the ordinary denial marker. No answerer currently owns an in-process child, so an escalation request fails closed and the child reports upward; a controller-owned parent may widen its own session and delegate again. An inherited `'never'` policy tells the child not to request escalation in its first system prompt.
+
+## Alternatives considered
+
+- **Generic `SessionHeader` policy fields** — rejected: they duplicate an event-sourced fact in metadata and require propagation through core session types, persistence backends, query indexes, collision identity, and every policy consumer. Unpublished setup events have the required ordering and reuse the existing durable store.
+- **Combining new policy facts with constructor history** — rejected: `Session.firstLiveSeq` classifies the complete constructor seed as replayed history, so telemetry would skip child-only facts. Unpublished setup keeps history and new facts on their existing sides of that boundary without another session option.
+- **A first-prompt listener** — rejected: it introduces listener ordering and a later timing boundary even though the creation transaction already permits log appends before publication.
+- **Copying deployment defaults** — rejected: defaults remain operator-owned and may change; an unswitched parent stamps nothing, so its child follows the current deployment.
+- **Live resolution walking `parentSession` at each call** — rejected: it breaks the "two sessions never see each other's state" isolation invariant, requires the parent session to stay loaded for the child's lifetime, and makes a mid-run parent switch retroactively change a running child. Snapshot-at-delegation is the semantic: the child keeps the policy it was handed; cancel-and-respawn picks up a tightening.
+- **Forcing `'never'` or routing asks to the root controller** — rejected as inheritance behavior. A forced value forecloses a future child answerer; parent routing needs parent-chain ownership and the spawning `callId`, and remains deferred in [the approval-seam Agent Note](2026-07-06-approval-seam.md).
+
+## Consequences
+
+- Spawn, fork, and nested in-process children retain a parent's explicit sandbox and approval overrides. The focused suite proves real filesystem denial, stale-fork precedence, delegation-time capture, the live-event boundary, default omission, and context disposal.
+- The keyless headless snapshot is the assembled regression: only the parent is `read-only`, the deployment default is `workspace-write`, and the child's persisted event plus denied disk write both fail if capture is removed.
+- Each delegation adds at most two log-only events. `dsh-subagent-inprocess` has optional peer types for the two policy services; compositions without either service behave unchanged. Out-of-process children retain their own deployment policy, and a running child does not follow later parent switches.