Merge remote-tracking branch 'origin/master' into codex/pr-555-ci-fix

packages/host/apiproxy/README.i18n.yaml

@@ -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 packages/host/apiproxy/README.md
-README.md: c29f30b85c5579f278ac9b40a0422347502eeb8f
-README.zh.md: 92b866bafd71902c55bf0bad14c6b9e761421cf8
+README.md: 0e98520149297732da8a4d06608f9b04204a444f
+README.zh.md: 27dc0ad668bd2da3d340fd081d698a7d090a94d4

packages/host/apiproxy/README.md

@@ -50,7 +50,7 @@ The `agentPreset.list` domain exposes the deployment's preset roster so a browse
 
 `agentPreset.read`, `copy`, `openDocument`, and `remove` manage the compositions themselves. `read` reports the text with its `trust`, for the read-only viewer. Authoring is copy-only: `copy` takes `{ from, agentPreset, name? }` — two ids the Host resolves against its own roots plus an optional display name — and copies the source's whole directory, so no composition text crosses the wire and a copy is exactly as loadable as its source; an uncontainable or already-taken id answers `agent-preset-invalid`, and `remove` refuses a shipped preset as `agent-preset-read-only`. `openDocument` hands one locally authored preset's DIRECTORY to the platform opener — the request carries an id, never a path, so no browser payload can select an arbitrary filesystem target; where the deployment has no native opener the reply is `{ opened: false, path }` for the surface to show as text, a shipped preset is refused like `remove`, and the gateway's `nativeOpen` config pins the capability where platform detection (`canOpenNativePath`) would mislead. These four are loopback-pinned in [`dsh-client-connection`](../../client/connection/README.md): a composition names the plugins a session runs, so reading one is reconnaissance, and copy/remove/openDocument manage the roster and drive the host desktop. `list` and `select` stay ordinary — the roster carries ids and trust and every preset picker needs it, and choosing a preset grants nothing `session.create`'s own `agentPreset` did not, over a default that already carries bash. `list` reports two path-free capability flags: `authorable`, whether the deployment configures a root a new preset could be copied to, and `hasDocument`, whether `openDocument` would open natively rather than answer a path.
 
-The `command.*` and `skill.*` domains expose the host command registry and skill catalog to clients. Every method addresses one session's agent by `sessionId` (a served session always has an Agent; `command.*` resumes cold sessions through the same path as `session.*`, while `skill.list` resolves the project root from the session header without touching the Agent registry). `skill.list` serves the composer's menu: it returns every user-invocable skill with its `modelInvocable` flag, so menus can mark user-only (`disable-model-invocation`) entries whose only invocation path is the slash gesture. Listing is the skill domain's only RPC — invocation itself is an ordinary `session.prompt` whose whitespace-bounded `/name` tokens `dsh-tool-skill` recognizes at the pre-step boundary and answers with injected `<skill_content>` context, so every entry point (Web, TUI, and ACP) shares one deterministic path—including for hand-typed text—with no dedicated invocation wire. `command.execute` runs a slash-command line host-side with pure admission semantics: the response reports whether the line resolved to a handler plus the minted lifecycle `commandId` when it did (correlating the acknowledgment with the flow node), while the outcome rides the durably logged `command/run`/`command/done` lifecycle pair broadcast on the mux stream. Command handlers may legitimately outlast the 30-second transport health deadline, so `command.execute` carries only caller/connection cancellation; that signal cancels the running handler. `host/commands-changed` is the catalog invalidation frame: clients refetch `command.list` instead of diffing.
+The `command.*` and `skill.*` domains expose the host command registry and skill catalog to clients. Every method addresses one session's agent by `sessionId` (a served session always has an Agent; `command.*` resumes cold sessions through the same path as `session.*`, while `skill.list` resolves the project root from the session header without touching the Agent registry). `skill.list` serves the composer's menu: it returns every user-invocable skill with its `modelInvocable` flag, so menus can mark user-only (`disable-model-invocation`) entries whose only invocation path is the slash gesture. Listing is the skill domain's only RPC — invocation itself is an ordinary `session.prompt` whose whitespace-bounded `/name` tokens `dsh-tool-skill` recognizes at the pre-step boundary and answers with injected `<skill_content>` context, so every entry point (Web, TUI, and ACP) shares one deterministic path—including for hand-typed text—with no dedicated invocation wire. `command.execute` runs a slash-command line host-side with pure admission semantics: the response reports whether the line resolved to a handler plus the minted lifecycle `commandId` when it did (correlating the acknowledgment with the flow node), while the outcome rides the durably logged `command/run`/`command/done` lifecycle pair broadcast on the mux stream. Command handlers may legitimately outlast the 30-second transport health deadline, so `command.execute` carries only caller/connection cancellation; that signal cancels the running handler. `host/commands-changed` is the registry-wide catalog invalidation frame: clients refetch `command.list` instead of diffing. `host/session-preset-changed` is its per-session counterpart, framed off the logged `agent-preset/selected` commit: recomposing a blank session's agent re-parents its scope without registering anything, so both catalogs that session's composition decides (`command.list`, `skill.list`) go stale with no registry change to announce it.
 
 The `settings.*`, `credentials.*`, and `llm.*` domains are the configuration-page wire. The settings domain serves the namespaces addressed by registered configurable providers (`ctx.llm.listConfigurableProviders()`) plus a small explicit allowlist — the Web preference `permission` and the product-owned `ui-onboarding`; adding a Settings registration alone never makes it remotely readable or writable. Any other namespace answers `settings-not-exposed` — the same answer an unregistered namespace gets, so no caller can enumerate the registry by probing. `settings.describe` returns each exposed namespace's serialized schemastery schema, redacted layered values (resolved/`base`/`user` — a field's presence in `user` marks it user-overridden), the `secrets` slot list, the section's `revision`, and the boolean `hasDocument` capability flag. The browser receives no Host path: pathless `settings.openDocument` asks the provider to materialize its document and then hands the Host-resolved result to the native opener, so no browser payload can select an arbitrary filesystem target. `settings.update`/`settings.replace` write the user layer; `settings.mutate` applies path ops (`set`/`unset`) against the section as stored, which is the removal path for a client holding the redacted view — rebuilding a section from it and replacing wholesale would delete the secrets the wire never returned. Any write may carry `expectedRevision`; a stale one answers `settings-conflict` with both revisions rather than overwriting the writer that landed first, and every other seam refusal folds into `settings-rejected`. Secret-role values never ride any response in any layer; a secret crosses the wire in exactly one direction — inside an `update`/`mutate` payload or `credentials.set`. `credentials.describe` returns value-free views (`configured`/`source`/`writable`), and `credentials.set`/`credentials.unset` map a shadowed-reference refusal onto `credential-rejected`. `llm.providers` merges the configurable-provider directory with live routes (dormant entries carry `active: false`; undeclared live routes append with no settings address) and `llm.models` is the session-independent catalog. `llm.discoverModels` interrogates a provider endpoint the page is still drafting: `settingsNs` selects the adapter family that knows how to read the listing, and the endpoint, protocol, and key come from the form rather than from storage. It writes nothing — the reply is candidates, and only a later `settings.mutate` decides what a route serves — so its `apiKey` is the third payload on which a secret may ride, alongside `settings.update`/`mutate` and `credentials.set`. The host never stores or returns it; like the other two it does ride the client's outgoing envelope, which `subscribeEnvelopes()` observers can see, and redacting that tap is a configuration-plane-wide change rather than this method's to make alone. Every refusal (an unserved namespace, a protocol with no readable listing, an unreachable endpoint, a rejected credential) folds into `model-discovery-failed`, whose message is the adapter's own text and whose details name the endpoint asked but never the credential offered. Three invalidation frames keep every surface converged without polling: `host/settings-changed {ns}` (`settings/document-updated` passthrough, so a raw change whose resolved value is unchanged still reaches clients), `host/credentials-changed {ref}` (reference names only, never values), and `host/models-changed` — fired by `llm/adapters-updated` and by a change to a configurable-provider namespace, whose settings carry that provider's catalog and endpoint; a `permission` or `ui-onboarding` change emits only its settings invalidation. The browser carrier restricts the whole configuration plane, reads and native actions included (`settings.describe`/`openDocument`/`update`/`replace`/`mutate`, `credentials.describe`/`set`/`unset`), to loopback same-origin requests — the `host.pickDirectory` privileged set. A composition without a settings or credential provider answers those domains with an actionable `internal` error naming the missing plugin.
 

packages/host/apiproxy/README.zh.md

@@ -50,7 +50,7 @@ Workspace 列表与 Session 列表是相互独立的重连基线。`workspace.cr
 
 `agentPreset.read`、`copy`、`openDocument` 与 `remove` 负责管理组装本身。`read` 返回文本连同它的 `trust`，供只读查看器使用。创作只有复制一种写入：`copy` 接收 `{ from, agentPreset, name? }`——两个由 Host 对照自身根目录解析的 id 加一个可选显示名——并整目录复制来源，因此组装文本不经过传输层，副本与其来源同等可加载；不可约束或已被占用的 id 回答 `agent-preset-invalid`，`remove` 对随附 preset 回答 `agent-preset-read-only`。`openDocument` 把一个本地创作 preset 的**目录**交给平台打开器——请求只携带 id、绝不携带路径，因此没有任何浏览器载荷能选中任意文件系统目标；部署没有原生打开器时回答 `{ opened: false, path }` 供界面以文本展示，随附 preset 与 `remove` 一样被拒绝，而网关的 `nativeOpen` 配置可在平台探测（`canOpenNativePath`）失真处钉死该能力。这四个方法在 [`dsh-client-connection`](../../client/connection/README.md) 中被固定在环回地址：组装指明了一个会话所运行的插件，因此读取它是侦察，而 copy/remove/openDocument 管理名单并驱动宿主桌面。`list` 与 `select` 保持为普通方法——名单只携带 id 与信任级别，每个 preset 选择器都需要它；而选择一个 preset 并不比 `session.create` 自带的 `agentPreset` 多给任何能力，何况默认 preset 本就带着 bash。`list` 报告两个不含路径的能力标志：`authorable`，即部署是否配置了可供复制新 preset 的根目录；`hasDocument`，即 `openDocument` 会原生打开、还是回答一个路径。
 
-`command.*` 与 `skill.*` 领域向客户端暴露宿主命令注册表和 skill（技能）目录。每个方法都通过 `sessionId` 寻址一个会话的 Agent（被服务的会话必有 Agent；`command.*` 经由与 `session.*` 相同的路径恢复冷会话，而 `skill.list` 从会话头解析项目根目录，不触碰 Agent 注册表）。`skill.list` 服务于 composer 的菜单：它返回每一个用户可调用的 skill 及其 `modelInvocable` 标志，让菜单能够标出仅限用户（`disable-model-invocation`）的条目——斜杠手势是这类条目唯一的调用路径。列表是 skill 领域唯一的 RPC——调用本身就是一次普通的 `session.prompt`，`dsh-tool-skill` 会在 pre-step 边界识别其中以空白为界的 `/name` token，并以注入的 `<skill_content>` 上下文作答，因此所有入口（Web、TUI 与 ACP（Agent Client Protocol））共享同一条确定性路径，手动键入的文本也走该路径，且没有专设的调用协议。`command.execute` 在宿主侧运行一条斜杠命令行，语义为纯准入：响应报告该行是否解析到处理器，并在解析到时回带铸造的生命周期 `commandId`（将本次确认与流节点关联）；结局经由持久落账并在 mux 流广播的 `command/run`/`command/done` 生命周期事件对承载。命令处理器运行超过 30 秒的传输健康时限仍属正常，因此 `command.execute` 仅携带调用方／连接取消信号；该信号可取消正在运行的处理器。`host/commands-changed` 是目录失效帧：客户端重新拉取 `command.list` 而不是做差分。
+`command.*` 与 `skill.*` 领域向客户端暴露宿主命令注册表和 skill（技能）目录。每个方法都通过 `sessionId` 寻址一个会话的 Agent（被服务的会话必有 Agent；`command.*` 经由与 `session.*` 相同的路径恢复冷会话，而 `skill.list` 从会话头解析项目根目录，不触碰 Agent 注册表）。`skill.list` 服务于 composer 的菜单：它返回每一个用户可调用的 skill 及其 `modelInvocable` 标志，让菜单能够标出仅限用户（`disable-model-invocation`）的条目——斜杠手势是这类条目唯一的调用路径。列表是 skill 领域唯一的 RPC——调用本身就是一次普通的 `session.prompt`，`dsh-tool-skill` 会在 pre-step 边界识别其中以空白为界的 `/name` token，并以注入的 `<skill_content>` 上下文作答，因此所有入口（Web、TUI 与 ACP（Agent Client Protocol））共享同一条确定性路径，手动键入的文本也走该路径，且没有专设的调用协议。`command.execute` 在宿主侧运行一条斜杠命令行，语义为纯准入：响应报告该行是否解析到处理器，并在解析到时回带铸造的生命周期 `commandId`（将本次确认与流节点关联）；结局经由持久落账并在 mux 流广播的 `command/run`/`command/done` 生命周期事件对承载。命令处理器运行超过 30 秒的传输健康时限仍属正常，因此 `command.execute` 仅携带调用方／连接取消信号；该信号可取消正在运行的处理器。`host/commands-changed` 是注册表级目录失效帧：客户端重新拉取 `command.list` 而不是做差分。`host/session-preset-changed` 是它按会话粒度的对应物，由落账的 `agent-preset/selected` 提交点成帧：重组空会话的 agent 只是重新挂接其 scope，不产生任何注册，因此该会话组成所决定的两份目录（`command.list`、`skill.list`）都会失效，却没有任何注册表变化来宣告它。
 
 `settings.*`、`credentials.*` 与 `llm.*` 领域是配置页协议。settings 领域服务于已注册可配置提供方所指向的 namespace（`ctx.llm.listConfigurableProviders()`），并额外服务于一份小型、显式的 allowlist——Web 偏好 `permission` 与产品持有的 `ui-onboarding`；仅新增一项 Settings 注册，绝不会使其可被远程读取或写入。其他任何 namespace 都只会得到 `settings-not-exposed`——未注册的 namespace 得到的是同一个答复，因此没有调用方能靠逐个探测把注册表枚举出来。`settings.describe` 为每个已暴露 namespace 提供其序列化 schemastery schema、脱敏后的分层值（resolved/`base`/`user`——字段出现在 `user` 中即标记其被用户覆盖）、`secrets` 槽位列表、该分节的 `revision`，以及布尔型 `hasDocument` 能力标志。浏览器不会收到 Host 路径：无路径参数的 `settings.openDocument` 会请求提供方准备文档，再把由 Host 解析出的结果交给原生打开器，因此任何浏览器载荷都无法选择任意文件系统目标。`settings.update`/`settings.replace` 写入用户层；`settings.mutate` 则在已存分节上施加路径 op（`set`/`unset`），这是持有脱敏视图的客户端的删除路径——据此重建分节再整体替换，会删掉协议从未回传过的那些机密。任何写入都可携带 `expectedRevision`；陈旧的期望值会以 `settings-conflict` 连同两个 revision 作答，而不是覆盖先落地的那个写方，其余每种 seam 拒绝则折叠为 `settings-rejected`。secret 角色的值绝不在任何一层搭乘任何响应；secret 只沿一个方向跨越协议——在 `update`/`mutate` 载荷或 `credentials.set` 之内。`credentials.describe` 返回不含值的视图（`configured`/`source`/`writable`），`credentials.set`/`credentials.unset` 则把被遮蔽引用的拒绝映射为 `credential-rejected`。`llm.providers` 把可配置提供方目录与存活路由合并（休眠条目携带 `active: false`；未声明的存活路由追加在后，不带 settings 地址），`llm.models` 则是与会话无关的目录。`llm.discoverModels` 询问页面尚在起草的提供方端点：`settingsNs` 选出懂得读取该列表的适配器家族，端点、协议与密钥则来自表单而非存储。它什么都不写——回复是候选，只有随后的 `settings.mutate` 才决定路由服务什么——因此其 `apiKey` 是 secret 可以搭乘的第三个载荷（另两个是 `settings.update`/`mutate` 与 `credentials.set`），且绝不被存储或回显。host 从不存储或回传它；与另两者一样，它确实会搭乘客户端的出站信封，`subscribeEnvelopes()` 的观察者能看到——为该 tap 做脱敏是整个配置面的改动，而非本方法一家的事。每一种拒绝（无人服务的 namespace、没有可读列表的协议、不可达端点、被拒凭据）都折叠为 `model-discovery-failed`，其消息是适配器自己的文本，details 点名被询问的端点，绝不点名所提供的凭据。三个失效帧让每个面无需轮询即保持收敛：`host/settings-changed {ns}`（`settings/document-updated` 透传，因此解析值未变的原始变更同样能到达客户端）、`host/credentials-changed {ref}`（只带引用名，绝不带值），以及 `host/models-changed`——它由 `llm/adapters-updated` 和可配置提供方 namespace 的变更触发，因为该提供方的设置正承载着它的目录与端点；`permission` 或 `ui-onboarding` 变更只会发出自身的 settings 失效通知。浏览器载体把整个配置面（含读取与原生操作：`settings.describe`/`openDocument`/`update`/`replace`/`mutate` 与 `credentials.describe`/`set`/`unset`）限制为仅接受来自回环地址的同源请求——即 `host.pickDirectory` 所在的特权集合。未装 settings 或凭据 provider 的组合会以指名缺失插件、包含解决建议的 `internal` 错误应答这些领域。
 

packages/host/apiproxy/src/api-proxy.ts

@@ -3373,6 +3373,17 @@ export function createApiProxy(ctx: Context, defaults: ApiProxyDefaults): ApiPro
           ctx.on('commands/change', () => {
             queue.push(frame({ type: 'host/commands-changed' }))
           }),
+          // The recompose itself registers nothing (it re-parents the agent's
+          // scope onto a standing mount that may already exist), so the
+          // logged selection is the only commit point a client can follow.
+          ctx.on('session/event', (session: Session, event: SessionEvent) => {
+            if (event.type !== 'agent-preset/selected') return
+            queue.push(frame({
+              type: 'host/session-preset-changed',
+              sessionId: session.id,
+              agentPreset: event.data.agentPreset,
+            }))
+          }),
           ctx.on('settings/document-updated', (ns) => {
             // The RAW-section event, not the resolved one: a field going from
             // inherited to overridden leaves the resolved value equal, and a

packages/host/apiproxy/src/api/events.schema.ts

@@ -82,6 +82,7 @@ export const hostFrameSchema = z.discriminatedUnion('type', [
   z.object({ type: z.literal('host/workspace-removed'), workspaceId: workspaceIdSchema }),
   z.object({ type: z.literal('host/archived-sessions-changed'), archivedSessionIds: z.array(sessionIdSchema) }),
   z.object({ type: z.literal('host/commands-changed') }),
+  z.object({ type: z.literal('host/session-preset-changed'), sessionId: sessionIdSchema, agentPreset: z.string() }),
   z.object({ type: z.literal('host/settings-changed'), ns: z.string() }),
   z.object({ type: z.literal('host/credentials-changed'), ref: z.string() }),
   z.object({ type: z.literal('host/models-changed') }),

packages/host/apiproxy/src/api/events.ts

@@ -130,6 +130,18 @@ export type HostFrame =
    * background rather than diffing.
    */
   | { type: 'host/commands-changed' }
+  /**
+   * One blank session was recomposed onto another agent preset (the logged
+   * `agent-preset/selected` commit point, read off the session stream). The
+   * registry-wide `host/commands-changed` cannot stand in for it: recomposing
+   * re-parents that agent's scope without registering anything, so a
+   * preset already mounted for another session produces no registry change
+   * at all. Clients refetch the catalogs this session's composition decides
+   * (`command.list`, `skill.list`) for this sessionId alone, and fold the
+   * preset id into their session row — the RPC echo reaches only the client
+   * that issued the switch, so the row is where every other one learns it.
+   */
+  | { type: 'host/session-preset-changed'; sessionId: SessionId; agentPreset: string }
   /**
    * One settings namespace's resolved value changed (`settings/updated`
    * passthrough) — an RPC write, an external `settings.yaml` edit, or a

packages/host/apiproxy/tests/api-proxy-agent-preset.spec.ts

@@ -14,6 +14,7 @@ import type { Agent } from '@deepseek-ai/dsh-agent'
 import SessionStore, { SessionId, type Session } from '@deepseek-ai/dsh-session'
 import UserInteractionService from '@deepseek-ai/dsh-user-interaction'
 import { RpcId, type RpcRequest } from '../src/api/rpc.ts'
+import type { HostFrame } from '../src/api/events.ts'
 import {
   InvalidPresetIdError, PresetExistsError, resolveSessionPreset, UnknownPresetError,
 } from '@deepseek-ai/dsh-agent-presets'
@@ -350,6 +351,37 @@ describe('agentPreset.select', () => {
       .toBe('core-web')
   })
 
+  it('frames the committed switch so clients can drop that session\'s catalogs', async () => {
+    const { api, ctx } = await harness(['standard', 'minimal'])
+    await api.sessions.create(request({ sessionId: SessionId('sel-frame'), agentPreset: 'standard' }))
+    // The host-stream opener reads the committed-workspace baseline; this
+    // spec owns preset identity, so the stub suffices (api-proxy-commands
+    // precedent).
+    ctx.provide('workspace', { list: () => [] } as never)
+    const abort = new AbortController()
+    const frames: HostFrame[] = []
+    const stream = api.events.host(request({}), abort.signal)
+    const consume = (async () => {
+      for await (const frame of stream) {
+        if (frame.payload.type === 'host/session-preset-changed') frames.push(frame.payload)
+      }
+    })()
+
+    await api.agentPresets.select(
+      request({ sessionId: SessionId('sel-frame'), agentPreset: 'minimal' }))
+    // The queue push rides the synchronous append, so one turn of the loop is
+    // enough to deliver it; closing the stream bounds the read either way.
+    await new Promise(resolve => setTimeout(resolve, 0))
+    abort.abort()
+    await consume
+
+    // Recomposing registers nothing, so this frame — not the registry-wide
+    // commands one — is what tells a client its cached catalogs are stale.
+    expect(frames).toEqual([
+      { type: 'host/session-preset-changed', sessionId: 'sel-frame', agentPreset: 'minimal' },
+    ])
+  })
+
   it('serializes two concurrent selects on one session', async () => {
     const { api, ctx } = await harness(['standard', 'core-web'])
     await api.sessions.create(request({ sessionId: SessionId('sel-race'), agentPreset: 'standard' }))

packages/host/apiproxy/tests/rpc-schemas.spec.ts

@@ -493,6 +493,7 @@ describe('events frame schemas', () => {
       } },
       { type: 'host/workspace-removed', workspaceId: 'w' },
       { type: 'host/commands-changed' },
+      { type: 'host/session-preset-changed', sessionId: 's', agentPreset: 'minimal' },
       { type: 'stream/error', error: { code: 'internal', message: 'm', details: {} } },
     ]
     for (const frame of frames) expect(hostFrameSchema.parse(frame)).toMatchObject({ type: frame.type })