docs(agent): document next-step acceptance window

packages/context/session-reference/README.i18n.yaml

@@ -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
-README.md: c995256511742c193e064cf808fc89194b444974
-README.zh.md: e2e67cfee745c84d6c85e8792e53c50bf2648293
+#   pnpm run verify-translation-pairing --write packages/context/session-reference/README.md
+README.md: 2ca461f88b266b4dec1ffa4132c8cb17455f4b8e
+README.zh.md: 9f8fd0bace9b37b2f7885eded7686ecac8c625ba

packages/context/session-reference/README.md

@@ -2,19 +2,19 @@
 
 English | [中文](README.zh.md)
 
-`ctx.sessionReferences` prepares bounded, read-only snapshots of other sessions as prompt-prefix context. It consumes `ctx.sessionQuery` and the backend-independent compact checkpoint marker; SQLite FTS is not required. The standard TUI bundle mounts it, while other hosts may call the service directly.
+`ctx.sessionReferences` prepares bounded, read-only snapshots of other sessions as sourced model-facing context. It consumes `ctx.sessionQuery` and the backend-independent compact checkpoint marker; SQLite FTS is not required. The standard TUI bundle mounts it, while other hosts may call the service directly.
 
 ## Public API
 
 - `listCandidates(agent, query?, limit?)` lists sessions other than `agent.id`, filters case-insensitively by id or cwd, and ranks same-cwd, cwd-less, then other-cwd records while preserving `listSessions()` creation order within each group. Each selected candidate uses its latest log-backed title as the mention label and falls back to the session id; titles and message bodies are not searched.
-- `prepare(agent, content, references, signal?)` preserves first-mention order, deduplicates ids, rejects self-reference and more than the configured distinct-source limit, reads every source in parallel, and returns detached content plus zero or one aggregated `HookContext`. Any invalid reference, failed read, cancellation, or budget failure rejects before the host calls `followup()` or `steer()`.
+- `prepare(agent, content, references, signal?)` preserves first-mention order, deduplicates ids, rejects self-reference and more than the configured distinct-source limit, reads every source in parallel, and returns detached content plus zero or one aggregated `UserMessageData` context. Any invalid reference, failed read, cancellation, or budget failure rejects before the host calls `followup()` or `steer()`.
 - `encodeSessionReferenceUri()` and `decodeSessionReferenceUri()` implement `dsh-session:<base64url(JSON.stringify(sessionId))>` so every JavaScript string id round-trips exactly. `formatSessionReferenceMention()` emits `@[label](uri)`, and `parseSessionReferenceText()` replaces Markdown mentions or bare canonical URIs with readable `@label` text while returning structured references. Explicit Markdown mentions reject every malformed URI; bare text is considered a reference only when a non-empty base64url-shaped payload follows the scheme, and a matching noncanonical candidate still fails. Empty or punctuation-only scheme mentions remain ordinary discussion text.
 
 ## Snapshot semantics
 
 Preparation calls `ctx.sessionQuery.readSurface()` once per distinct source and never rereads it after enqueue. It projects only direct-user `user/message`, direct-user `steering/message`, assistant text, and `user/message` checkpoints carrying the canonical `dsh-compact` source marker from the folded current surface. For a source prompt that already contains baked prefix context, projection reads only its model-hidden display content, preventing recursive snapshot propagation. Shadowed pre-compaction events, tools, reasoning, context, plugin-generated user messages other than marked compact checkpoints, and unfinished assistant chunks are excluded. A compacted source therefore contributes its latest checkpoint plus retained later conversation, not restored shadowed text.
 
-The context source is `{ kind: 'plugin', plugin: 'session-reference' }` with `placement: 'prompt-prefix'`. Its metadata records version `1`, source ids and labels, capture seqs, compact presence, retained/omitted message counts, omitted UTF-8 bytes, and truncation state. AgentLoop writes the snapshot, `## My request:` delimiter, and effective prompt into one `user/message` or `steering/message`; the same event's model-hidden envelope retains the direct prompt and metadata for UI replay. Later source mutation, compaction, or deletion cannot change target replay.
+The context source is `{ kind: 'session-reference', version: 1, references }`; each reference records its source id and label, capture seq, compact presence, retained/omitted message counts, omitted UTF-8 bytes, and truncation state. The standard TUI preserves admission ownership without attaching context to the generic inbox record: outside the next-step acceptance window, a one-shot `agent/prompt-submit` wrapper adds the snapshot only to an allowed decision; during prompt admission or an open turn, `inject()` and `steer()` stage beside each other for the same safe boundary. The target log therefore records a sourced context `user/message` followed by the readable direct `user/message` or `steering/message`. Later source mutation, compaction, or deletion cannot change target replay.
 
 ## Configuration
 
@@ -32,7 +32,7 @@ Retention applies `maxReferenceBytes` independently to each source, keeps compac
 
 #### What the model sees
 
-The model sees one user-role message in this order: the `## Referenced sessions` untrusted snapshot, the `## My request:` delimiter, then the current message with its readable `@label`. The warning forbids following instructions, permission claims, or tool requests from the snapshot unless the current user repeats them. Labels, cwd values, ids, and conversation text are serialized as JSON inside `<referenced-sessions>` tags; every data `<` is emitted as the lossless JSON escape `\u003c`, so source text cannot spell a framing tag.
+The model sees two consecutive user-role messages: the `## Referenced sessions` untrusted snapshot, then the current message with its readable `@label`. The warning forbids following instructions, permission claims, or tool requests from the snapshot unless the current user repeats them. Labels, cwd values, ids, and conversation text are serialized as JSON inside `<referenced-sessions>` tags; every data `<` is emitted as the lossless JSON escape `\u003c`, so source text cannot spell a framing tag.
 
 #### Token effect
 
@@ -40,7 +40,7 @@ Each referenced message adds the fixed warning plus up to three serialized snaps
 
 #### KV Cache effect
 
-The combined snapshot and request are append-only at the target message boundary and preserve earlier cacheable history. Different references or source capture contents change the new suffix only; later target compaction may invalidate reuse from its replacement boundary.
+The snapshot and request are consecutive append-only target messages and preserve earlier cacheable history. Different references or source capture contents change the new suffix only; later target compaction may invalidate reuse from its replacement boundary.
 
 ## Known Limitations and Deferred Work
 

packages/context/session-reference/README.zh.md

@@ -2,19 +2,19 @@
 
 [English](README.md) | 中文
 
-`ctx.sessionReferences` 会把其他会话准备为有界、只读快照，作为提示词前缀上下文。它消费 `ctx.sessionQuery` 与后端无关的 compact 检查点标记；不需要 SQLite FTS。标准 TUI bundle 会装载它，其他宿主也可直接调用该服务。
+`ctx.sessionReferences` 会把其他会话准备为有界、只读快照，作为带来源信息、面向模型的上下文。它消费 `ctx.sessionQuery` 与后端无关的 compact 检查点标记；不需要 SQLite FTS。标准 TUI bundle 会装载它，其他宿主也可直接调用该服务。
 
 ## 公开 API
 
 - `listCandidates(agent, query?, limit?)` 会列出 `agent.id` 之外的会话，按 id 或 cwd 进行不区分大小写的筛选，再按同 cwd、无 cwd、其他 cwd 记录排序，同时保持每组内的 `listSessions()` 创建顺序。每个已选候选会话都使用最新的日志支持标题作为 mention label，并回退到会话 id；不搜索标题与消息主体。
-- `prepare(agent, content, references, signal?)` 会保留首次 mention 顺序、对 id 去重，并拒绝自引用或超过已配置不同源上限的情况。它会并行读取所有源，返回与输入脱离的内容，外加零个或一个聚合 `HookContext`。任何无效引用、读取失败、取消或预算失败都会在宿主调用 `followup()` 或 `steer()` 之前被拒绝。
+- `prepare(agent, content, references, signal?)` 会保留首次 mention 顺序、对 id 去重，并拒绝自引用或超过已配置不同源上限的情况。它会并行读取所有源，返回与输入脱离的内容，外加零个或一个聚合 `UserMessageData` 上下文。任何无效引用、读取失败、取消或预算失败都会在宿主调用 `followup()` 或 `steer()` 之前被拒绝。
 - `encodeSessionReferenceUri()` 与 `decodeSessionReferenceUri()` 实现 `dsh-session:<base64url(JSON.stringify(sessionId))>`，因此每个 JavaScript 字符串 id 都能精确往返。`formatSessionReferenceMention()` 发出 `@[label](uri)`，`parseSessionReferenceText()` 将 Markdown mention 或裸规范 URI 替换为可读的 `@label` 文本，并返回结构化引用。显式 Markdown mention 会拒绝每个格式错误的 URI；只当 scheme 后跟非空、符合 base64url 形状的 payload 时，裸文本才被视为引用，匹配但非规范的候选项仍会失败。空 scheme mention 或只含标点符号的 scheme mention 仍是普通讨论文本。
 
 ## 快照语义
 
 准备阶段会对每个不同源调用一次 `ctx.sessionQuery.readSurface()`，入队后绝不重读。它仅投影折叠后当前表层中的直接 user `user/message`、直接 user `steering/message`、assistant 文本，以及 `user/message` 检查点；这类检查点携带规范 `dsh-compact` 源标记。对于已经包含烘焙前缀上下文的源提示词，投影只读取其对模型隐藏的显示内容，以防止快照递归传播。已遮蔽的压缩前事件、工具、reasoning、上下文、除已标记 compact 检查点外的插件生成 user 消息，以及未完成的 assistant chunk 均会被排除。因此，已压缩源贡献的是最新检查点与之后保留的会话，而非已恢复的遮蔽文本。
 
-上下文源为 `{ kind: 'plugin', plugin: 'session-reference' }`，并携带 `placement: 'prompt-prefix'`。其元数据会记录版本 `1`、源 id 与 label、捕获 seq、是否存在 compact、已保留／已省略消息数、已省略 UTF-8 字节数与截断状态。AgentLoop 将快照、`## My request:` 分隔符和有效提示词写入同一个 `user/message` 或 `steering/message`；同一事件的模型隐藏 envelope 保留直接提示词与元数据，用于 UI 回放。后续源变更、压缩或删除都无法改变目标回放。
+上下文源为 `{ kind: 'session-reference', version: 1, references }`；每条引用会记录其源 id 与 label、捕获 seq、是否存在 compact、已保留／已省略消息数、已省略 UTF-8 字节数与截断状态。标准 TUI 在不把上下文附加到通用 inbox 记录的情况下保留接纳归属：next-step 接收窗口之外，一次性 `agent/prompt-submit` 包装层只为获准决策添加快照；提示词接纳期间或轮次打开时，`inject()` 与 `steer()` 会并排暂存到同一安全边界。目标日志因此会先记录一条带来源信息的上下文 `user/message`，再记录可读的直接 `user/message` 或 `steering/message`。后续源变更、压缩或删除都无法改变目标回放。
 
 ## 配置
 
@@ -32,7 +32,7 @@
 
 #### 模型看到的内容
 
-模型会按此顺序看到一条 user 角色消息：`## Referenced sessions` 不受信任快照、`## My request:` 分隔符，随后是带可读 `@label` 的当前消息。警告禁止遵循快照中的指令、权限声明或工具请求，除非当前 user 重复这些内容。Label、cwd 值、id 与会话文本作为 JSON 在 `<referenced-sessions>` 标签中序列化；每个数据 `<` 都发出为无损 JSON 转义 `\u003c`，因此源文本无法拼出框定标签。
+模型会看到两条连续的 user 角色消息：先是 `## Referenced sessions` 不受信任快照，再是带可读 `@label` 的当前消息。警告禁止遵循快照中的指令、权限声明或工具请求，除非当前 user 重复这些内容。Label、cwd 值、id 与会话文本作为 JSON 在 `<referenced-sessions>` 标签中序列化；每个数据 `<` 都发出为无损 JSON 转义 `\u003c`，因此源文本无法拼出框定标签。
 
 #### Token 影响
 
@@ -40,7 +40,7 @@
 
 #### KV Cache 影响
 
-组合快照与请求在目标消息边界处仅追加，并保留较早的可缓存历史。不同引用或源捕获内容只改变新后缀；后续目标压缩可能使从替换边界起的复用失效。
+快照与请求是两条连续、仅追加的目标消息，并保留较早的可缓存历史。不同引用或源捕获内容只改变新后缀；后续目标压缩可能使从替换边界起的复用失效。
 
 ## 已知限制与暂缓事项
 

packages/cordis/tool-cordis/src/api-catalog.ts

@@ -1016,8 +1016,8 @@ export const EVENT_API: readonly EventApiEntry[] = [
   {
     name: 'agent/inbox/enqueue',
     mode: 'emit',
-    signature: '\'agent/inbox/enqueue\'(this: Scoped<Agent>, agent: Agent, message: AgentMessage): void',
-    jsDoc: '/**\n * An item entered the queued or steering inbox.\n * @param agent - the owning agent.\n * @param message - accepted content, source, and correlation identity.\n * Scope-filtered dispatch (`@deepseek-ai/dsh-scope`): agent-scoped listeners receive only that agent.\n * @mode emit\n */',
+    signature: '\'agent/inbox/enqueue\'(this: Scoped<Agent>, agent: Agent, message: AgentMessage, placement: InboxPlacement): void',
+    jsDoc: '/**\n * An item entered the queued or steering inbox. `placement` is the\n * acceptance-time routing result; listeners must not reconstruct it from\n * later agent or session state.\n * @param agent - the owning agent.\n * @param message - accepted content, source, and correlation identity.\n * @param placement - resolved queued or steering placement.\n * Scope-filtered dispatch (`@deepseek-ai/dsh-scope`): agent-scoped listeners receive only that agent.\n * @mode emit\n */',
     summary: 'An item entered the queued or steering inbox.',
   },
   {
@@ -1320,7 +1320,7 @@ export const EVENT_API: readonly EventApiEntry[] = [
 export const TYPE_API: readonly TypeApiEntry[] = [
   {
     name: 'Agent',
-    declaration: 'export interface Agent {\n    readonly id: SessionId;\n    readonly options: AgentOptions;\n    readonly session: Session;\n    readonly status: AgentStatus;\n    readonly ctx: Context;\n    send(input: UserMessageData, options: SendOptions): AgentMessageId;\n    cancel(cause: AgentCancelCause, options?: CancelOptions): void;\n    whenIdle(): Promise<void>;\n    followup(input: UserMessageData): AgentMessageId;\n    steer(input: UserMessageData): AgentMessageId;\n    inject(input: UserMessageData): AgentMessageId;\n    retry(): void;\n}',
+    declaration: 'export interface Agent {\n    readonly id: SessionId;\n    readonly options: AgentOptions;\n    readonly session: Session;\n    readonly status: AgentStatus;\n    readonly acceptsNextStep: boolean;\n    readonly ctx: Context;\n    send(input: UserMessageData, options: SendOptions): AgentMessageId;\n    cancel(cause: AgentCancelCause, options?: CancelOptions): void;\n    whenIdle(): Promise<void>;\n    followup(input: UserMessageData): AgentMessageId;\n    steer(input: UserMessageData): AgentMessageId;\n    inject(input: UserMessageData): AgentMessageId;\n    retry(): void;\n}',
   },
   {
     name: 'AgentCancelCause',

packages/core/agent-loop/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/core/agent-loop/README.md
-README.md: bdf1367446bb9345d56f0a44e65e43467dd82144
-README.zh.md: 3540499bfc64fc465853bc26b9a9c7f3cd1df15e
+README.md: 95b4410aaaf847dac73e9cca327c304af66aed7f
+README.zh.md: acd8dd875856bd1df79030e301d3a2174e391bd7

packages/core/agent-loop/README.md

@@ -54,7 +54,7 @@ Configured agents start automatically. A model call requires both `provider` and
 
 The concrete `ReactLoopAgent`, its queued input, outbox, and run controls are package-internal. The package root exports only the plugin/service/config contract, and the package exports map exposes no `./src/*` escape hatch; lifecycle owners create agents through `ctx.agents` rather than naming, constructing, or starting driver internals. One prepared session can be claimed by only one concrete driver, and everything observable happens through session events and the `agent/*` event taxonomy.
 
-The unified `send()` primitive routes content and source by (`target` × `wakeup`); `followup`/`steer`/`inject` are its fixed-preset aliases. A `next-turn` item joins the queued FIFO, waking the driver unless `wakeup: false`; admission happens before any turn opens. An allowed prompt and the prompt waterfall's `additionalContexts` enter the outbox as separate messages, then `run()` opens the turn and drains them together. A running `next-step`/wakeup `steer()` enters the same outbox without prompt admission and normally causes another step. A `next-step`/no-wakeup `inject()` waits there only while a turn is open; while idle it appends a `user/message` immediately without opening a turn or running the model. Persistence owns the resulting eager drain. Every inbox enqueue publishes `agent/inbox/enqueue`; taking it publishes `agent/inbox/dequeue`, and `cancel()` without `keepInbox` publishes `agent/inbox/discard`.
+The unified `send()` primitive routes content and source by (`target` × `wakeup`); `followup`/`steer`/`inject` are its fixed-preset aliases. A `next-turn` item joins the queued FIFO, waking the driver unless `wakeup: false`; admission happens before any turn opens. The loop opens a private next-step acceptance window before `agent/prompt-submit` and closes it before `turn/end`. During that window, `steer()` and `inject()` stage in one outbox; an allowed admission opens the turn, records the prompt and returned `additionalContexts`, then drains the staged input before the first request. A blocked or failed admission writes no prompt or hook-produced context and leaves caller-staged input for retry or a later admitted prompt. Outside the window, steering becomes a waking queued prompt and injection immediately appends `user/message` without opening a turn or running the model. Every inbox enqueue publishes `agent/inbox/enqueue` with the resolved queued-or-steering placement; taking it publishes `agent/inbox/dequeue`, and `cancel()` without `keepInbox` publishes `agent/inbox/discard`.
 
 ### Loop lifecycle (`agent.ts`)
 

packages/core/agent-loop/README.zh.md

@@ -54,7 +54,7 @@ interface Config {
 
 实体 `ReactLoopAgent`、其排队输入、outbox 与运行控制均为包内部实现。包根只导出插件／服务／配置契约，包导出映射不提供 `./src/*` 逃逸路径；生命周期拥有方通过 `ctx.agents` 创建 agent，而不是点名、构造或启动驱动器内部组件。一个准备完成的会话只能由一个实体驱动器认领；所有可观测行为都通过会话事件和 `agent/*` 事件分类体系发生。
 
-统一的 `send()` 原语按（`target` × `wakeup`）路由内容与来源；`followup`/`steer`/`inject` 是它的固定预设别名。`next-turn` 项加入排队 FIFO，除非 `wakeup: false`，否则会唤醒驱动器；接纳发生在任何轮次开启之前。获准的提示词与提示词 waterfall（瀑布式事件）的 `additionalContexts` 会作为各自独立的消息进入 outbox，然后 `run()` 开启轮次并一起 drain 它们。运行期间的 `next-step`／wakeup `steer()` 进入同一个 outbox，不经过提示词接纳，通常导致再执行一步。`next-step`／不唤醒的 `inject()` 只在轮次打开时才在 outbox 中等待；空闲时它立即追加一条 `user/message`，不开启轮次也不运行模型。由此产生的立即 drain 由持久化负责。每次 inbox 入队都会发布 `agent/inbox/enqueue`；取走它会发布 `agent/inbox/dequeue`；`cancel()` 在不带 `keepInbox` 时会发布 `agent/inbox/discard`。
+统一的 `send()` 原语按（`target` × `wakeup`）路由内容与来源；`followup`/`steer`/`inject` 是它的固定预设别名。`next-turn` 项加入排队 FIFO，除非 `wakeup: false`，否则会唤醒驱动器；接纳发生在任何轮次开启之前。循环在 `agent/prompt-submit` 之前打开一个私有的 next-step 接收窗口，并在 `turn/end` 之前关闭它。在该窗口内，`steer()` 与 `inject()` 会暂存到同一个 outbox；接纳获准后会开启轮次，记录提示词及其返回的 `additionalContexts`，再于首次请求前排空暂存输入。接纳被阻止或失败时，不会写入提示词或钩子生成的上下文，并会保留调用方暂存的输入，以供重试或之后获准的提示词使用。窗口之外，steering（中途引导）会成为唤醒驱动器的排队提示词，而注入会立即追加 `user/message`，不开启轮次也不运行模型。每次 inbox 入队都会发布 `agent/inbox/enqueue`，并携带解析出的 queued 或 steering 路由归类；取走它会发布 `agent/inbox/dequeue`；`cancel()` 在不带 `keepInbox` 时会发布 `agent/inbox/discard`。
 
 ### 循环生命周期（`agent.ts`）
 
@@ -89,7 +89,7 @@ interface Config {
 
 #### Token 影响
 
-每个步骤都会再次计入系统文本与 schema。逐 agent 作用域决定贡献，而权威组装 waterfall 可以改变最终请求，并使其监听器负责保持协议连贯。
+每个步骤都会再次计入系统文本与 schema。逐 agent 作用域决定贡献，而权威组装 waterfall（瀑布式事件）可以改变最终请求，并使其监听器负责保持协议连贯。
 
 #### KV Cache 影响
 
@@ -99,7 +99,7 @@ interface Config {
 
 #### 模型所见
 
-已接纳的 user 消息、assistant 消息、工具调用与结果、注入上下文和 steering（中途引导）都会记录，并在后续步骤中发送。原始流分片、生命周期边界和其他仅写入日志的事件会被排除。
+已接纳的 user 消息、assistant 消息、工具调用与结果、注入上下文和 steering 都会记录，并在后续步骤中发送。原始流分片、生命周期边界和其他仅写入日志的事件会被排除。
 
 #### Token 影响
 

packages/core/agent/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/core/agent/README.md
-README.md: d3aced3c680f2d4ae8df9c3f2bf83ae01435d05b
-README.zh.md: e2c57ba50162deeac502d4f8eff23f7f2f7dab71
+README.md: 3ee1e7f1229d5e34c4e73896bbb30d2feb6ac10b
+README.zh.md: 60e7410d102cb61d0db81baf1f21a9cc51081b7c

packages/core/agent/README.md

@@ -58,10 +58,11 @@ Turn and step boundaries and the model token stream are durable `session/event`
 
 The handle every plugin programs against:
 
-- `agent.send(input, options)` — the one delivery primitive over the (`target` × `wakeup`) matrix. `input` is the existing `UserMessageData { content, source }`, while `SendOptions` requires only the routing policy `target` and `wakeup`. It returns the accepted message's opaque `AgentMessageId`, which the message's `agent/inbox/enqueue`/`dequeue`/`discard` events carry so a caller can correlate a queued item with its lifecycle. `target: 'next-turn'` queues one independent FIFO item that, if admitted, becomes the sole ordinary prompt in its turn. `target: 'next-step'` with `wakeup: true` submits steering, while `target: 'next-step'` with `wakeup: false` injects durable context without running the model. The [one-send-one-turn Agent Note](../../../.agents/notes/implemented/simplification/2026-07-17-one-send-one-turn.md) owns the turn rationale.
+- `agent.send(input, options)` — the one delivery primitive over the (`target` × `wakeup`) matrix. `input` is the existing `UserMessageData { content, source }`, while `SendOptions` requires only the routing policy `target` and `wakeup`. It returns the accepted message's opaque `AgentMessageId`, which the message's `agent/inbox/enqueue`/`dequeue`/`discard` events carry so a caller can correlate a queued item with its lifecycle; enqueue also carries the resolved `queued | steering` placement so listeners never reconstruct acceptance-time routing from later state. `target: 'next-turn'` queues one independent FIFO item that, if admitted, becomes the sole ordinary prompt in its turn. `target: 'next-step'` with `wakeup: true` submits steering, while `target: 'next-step'` with `wakeup: false` injects durable context without running the model. The [one-send-one-turn Agent Note](../../../.agents/notes/implemented/simplification/2026-07-17-one-send-one-turn.md) owns the turn rationale.
 - `agent.followup(input)` — the `next-turn`/wakeup preset of `send()`: queue an ordinary follow-up turn and wake the driver.
-- `agent.steer(input)` — the `next-step`/wakeup preset: while a turn is open, stage steering for its next safe boundary without dispatching `agent/prompt-submit`; when idle, delegate to a woken follow-up. Cancellation or disposal may discard pending steering.
-- `agent.inject(input)` — the `next-step`/no-wakeup preset: append model-facing context without running the model; the next request sees a verbatim user-role message whose provenance is carried by the required `input.source`. While a turn is open, injection waits in the outbox for the next safe boundary. While idle, it appends immediately without opening a turn; persistence reacts to `session/event` independently. Injection emits no `agent/inbox/*` event.
+- `agent.steer(input)` — the `next-step`/wakeup preset: during prompt admission or an open turn, stage steering for the next safe boundary without dispatching `agent/prompt-submit`; outside that acceptance window, delegate to a woken follow-up. Admission failure leaves staged steering for retry or a later admitted prompt, while cancellation or disposal may discard it.
+- `agent.inject(input)` — the `next-step`/no-wakeup preset: append model-facing context without running the model; the next request sees a verbatim user-role message whose provenance is carried by the required `input.source`. During prompt admission or an open turn, injection waits in the outbox for the next safe boundary. Outside that acceptance window, it appends immediately without opening a turn; persistence reacts to `session/event` independently. Injection emits no `agent/inbox/*` event.
+- `agent.acceptsNextStep` — whether a `next-step` send would currently join prompt admission or the open turn. Use this narrower routing predicate when a caller must choose between steering and a fresh admitted prompt; `status === 'running'` also covers admission exit and turn settlement.
 - `agent.cancel(cause, options?)` — cancel the active turn and, unless `options.keepInbox`, ALL pending work. Callers must choose the `user | parent` cause explicitly; an active holder copies its discriminant into a detached frozen signal reason before aborting. An effective call emits `agent/cancel-requested` with the cause before clearing queued and steering work; dropped items are reported on `agent/inbox/discard`, and observers may synchronize state but cannot veto cancellation. `keepInbox: true` aborts the turn but preserves queued and steering items (no discard, and un-started work is not dropped). The same-process typed seam adds no runtime validation or compatibility fallback for untyped callers. Repeated active-turn cancellation is first-wins for the signal, and idle cancellation is a safe no-op with no notification. ACP maps to `user`, while in-process parent propagation maps to `parent`. The cause is runtime-only; durable `turn/end` stays coarse `aborted`.
 - `agent.whenIdle()` — resolve once the agent reaches quiescence after settling out of `running` (idle → immediately; disposed → awaits the loop exit). A non-owner's quiescence-observation hook: it observes the work settling WITHOUT tearing the agent down. Teardown is separate — a lifecycle owner stops and unregisters via `AgentHandle.dispose()`, which awaits the loop exit directly.
 - `agent.session`, `agent.status`, `agent.options`, `agent.id`

packages/core/agent/README.zh.md

@@ -58,10 +58,11 @@ Agent *创建* 由实现 `AgentFactory` 的插件（`dsh-agent-loop`）提供，
 
 每个插件面向的 handle：
 
-- `agent.send(input, options)`：覆盖（`target` × `wakeup`）矩阵的唯一投递原语。`input` 是既有的 `UserMessageData { content, source }`，而 `SendOptions` 只要求路由策略 `target` 与 `wakeup`。它返回被接受消息的不透明 `AgentMessageId`，由该消息的 `agent/inbox/enqueue`/`dequeue`/`discard` 事件携带，调用方可据此把排队项与其生命周期关联。`target: 'next-turn'` 排队一条独立 FIFO 项，获准后成为其轮次中唯一的普通提示词。`target: 'next-step'` 且 `wakeup: true` 提交 steering（中途引导），而 `target: 'next-step'` 且 `wakeup: false` 注入持久上下文，不运行模型。轮次原理由 [one-send-one-turn Agent Note](../../../.agents/notes/implemented/simplification/2026-07-17-one-send-one-turn.md)拥有。
+- `agent.send(input, options)`：覆盖（`target` × `wakeup`）矩阵的唯一投递原语。`input` 是既有的 `UserMessageData { content, source }`，而 `SendOptions` 只要求路由策略 `target` 与 `wakeup`。它返回被接受消息的不透明 `AgentMessageId`，由该消息的 `agent/inbox/enqueue`/`dequeue`/`discard` 事件携带，调用方可据此把排队项与其生命周期关联；入队事件还会携带解析出的 `queued | steering` 路由归类，使监听器无需从后续状态重建接收时的路由。`target: 'next-turn'` 排队一条独立 FIFO 项，获准后成为其轮次中唯一的普通提示词。`target: 'next-step'` 且 `wakeup: true` 提交 steering（中途引导），而 `target: 'next-step'` 且 `wakeup: false` 注入持久上下文，不运行模型。轮次原理由 [one-send-one-turn Agent Note](../../../.agents/notes/implemented/simplification/2026-07-17-one-send-one-turn.md)拥有。
 - `agent.followup(input)`：`send()` 的 `next-turn`／wakeup 预设：排队一个普通后续轮次并唤醒驱动器。
-- `agent.steer(input)`：`next-step`／wakeup 预设：轮次打开时，为其下一个安全边界暂存 steering，且不分发 `agent/prompt-submit`；空闲时委托给会唤醒的后续轮次。取消或 dispose 可能丢弃待处理 steering。
-- `agent.inject(input)`：`next-step`／不唤醒预设：追加面向模型的上下文而不运行模型；下一次请求会看到一条逐字的 user role 消息，其来源由必填的 `input.source` 携带。轮次打开时，注入在 outbox 中等待下一个安全边界。空闲时，它立即追加而不开启轮次；持久化独立地响应 `session/event`。注入不发出 `agent/inbox/*` 事件。
+- `agent.steer(input)`：`next-step`／wakeup 预设：提示词接纳期间或轮次打开时，为下一个安全边界暂存 steering，且不分发 `agent/prompt-submit`；该接收窗口之外则委托给会唤醒的后续轮次。接纳失败会保留暂存的 steering，以供重试或之后获准的提示词使用，而取消或 dispose 可能丢弃它。
+- `agent.inject(input)`：`next-step`／不唤醒预设：追加面向模型的上下文而不运行模型；下一次请求会看到一条逐字的 user role 消息，其来源由必填的 `input.source` 携带。提示词接纳期间或轮次打开时，注入会在 outbox 中等待下一个安全边界。该接收窗口之外，它会立即追加而不开启轮次；持久化独立地响应 `session/event`。注入不发出 `agent/inbox/*` 事件。
+- `agent.acceptsNextStep`：当前发送 `next-step` 时，是否会加入提示词接纳或已打开的轮次。当调用方必须在 steering 与新接纳的提示词之间选择时，应使用这一更窄的路由判定；`status === 'running'` 还涵盖接纳收尾与轮次结算阶段。
 - `agent.cancel(cause, options?)`：取消活动轮次，并在未设置 `options.keepInbox` 时取消全部待处理工作。调用方必须显式选择 `user | parent` 原因；活动持有者会在中止前把其判别字段复制为已分离、冻结的信号原因。有效调用会在清除排队与 steering 工作前，随原因发出 `agent/cancel-requested`；丢弃项在 `agent/inbox/discard` 上报告，观察方可以同步状态，但不能 veto 取消。`keepInbox: true` 会中止轮次，但保留排队与 steering 项（不丢弃，且不删除尚未开始的工作）。同进程类型化 seam 不会为无类型调用方添加运行时校验或兼容回退。重复取消活动轮次时，首个信号生效；空闲取消是安全空操作，不发通知。ACP 映射到 `user`，进程内父传播映射到 `parent`。原因只存在于运行时；持久 `turn/end` 保持粗粒度的 `aborted`。
 - `agent.whenIdle()`：agent 从 `running` 结算后达到静默时解析（idle ⇒ 立即；disposed ⇒ 等待循环退出）。这是非拥有者的静默观测钩子：观察工作结算，但不 teardown agent。Teardown 独立存在；生命周期拥有者通过 `AgentHandle.dispose()` 停止并注销，并直接等待循环退出。
 - `agent.session`、`agent.status`、`agent.options`、`agent.id`

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

@@ -63,9 +63,9 @@ export type MuxFrame =
   | { type: 'question/requested'; sessionId: SessionId; questions: AskUserQuestionItem[] }
   | { type: 'question/resolved'; sessionId: SessionId; questionRpcId: RpcId; outcome: 'answered' | 'cancelled' }
   /**
-   * A message entered the addressed agent's inbox (`agent/queued` passthrough:
-   * a queued message is not model-visible, so there is no session event to
-   * ride — this transient frame is the only wire signal). On stream open the
+   * A message entered the addressed agent's inbox. A queued message is not
+   * model-visible, so there is no session event to carry it; this transient
+   * frame is the only wire signal. On stream open the
    * host replays the current queue snapshot for every attached session (same
    * refresh-recovery baseline as pending questions); queue clearing on cancel
    * has no dedicated frame — clients fold it from the status flip.