@deepseek-ai/dsh-session-snapshot 0.1.3-alpha.2 → 0.1.5-alpha.2

package/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/test-support/session-snapshot/README.md
-README.md: 12acb8d426a2966a852f98d85843d3c246ffeb5f
-README.zh.md: 0b30cbf205f6055ef7b11b791d47adb38f6f704d
+README.md: a73ea265516f965bec8d11c41d6eb5bb028c7ac6
+README.zh.md: 544694a9fe9678d89cffb1f3293ea4e648e89f5d

package/README.md

@@ -68,27 +68,30 @@ defineAcpSnapshotSuite({
 })
 ```
 
-Each recorded-session directory carries a closed `snapshot.yml` manifest plus canonical parent and contiguous child roles. Parent filenames are `session[.vN].jsonl`; children are `session.<ordinal>[.vN].jsonl`; v0 omits `.v0`, positive versions use lowercase `.vN`, and every filename agrees with its header. A role may retain older generations, but the harness selects the numerically highest one. An owning manifest may declare `sessionFormat.version` plus one or more closed `coverage` names to keep that historical generation as an explicit migration fixture; absence tracks the current writer. The manifest also names the scenario, shipped profile, composition/header class, recording source, and only the replay, platform, permission, environment, workspace, or input facts the completed Session cannot reconstruct. The adapter registers expected-output, Session-log, and optional `workspace.expected/` comparisons; guards reject orphan directories, missing roles, noncanonical names, absolute paths, malformed manifests, and platform-specific separators.
+Each recorded-session directory carries a closed `snapshot.yml` manifest plus canonical parent and contiguous child roles. Parent filenames are `session[.vN].jsonl`; children are `session.<ordinal>[.vN].jsonl`; v0 omits `.v0`, positive versions use lowercase `.vN`, and every filename agrees with its header. A role may retain older generations, but the harness selects the numerically highest one. An owning manifest may declare `sessionFormat.version` plus one or more closed `coverage` names to keep that historical generation as an explicit migration fixture; absence tracks the current writer. The manifest also names the scenario, shipped profile, composition/header class, recording source, and only the replay, platform, permission, environment, workspace, or input facts the completed Session cannot reconstruct. Storage guards check tool results and portable paths in every selected parent and child role. Prompt/schema scrubbing, message identity, and prompt-before-request ordering apply to current generations; retained predecessors keep their historical representation. The adapter registers expected-output, Session-log, and optional `workspace.expected/` comparisons; guards reject orphan directories, missing roles, noncanonical names, absolute paths, malformed manifests, and platform-specific separators.
 
-`normalizeSessionSnapshot` retains the complete Session header and event payloads but omits top-level `seq`/`time` envelopes from committed fixtures after normalizing paths and scrubbing request headers; it also normalizes embedded stream clocks and historical packed-row `seq0`/`time0` envelopes. Replay synthesizes the top-level envelopes in memory, while runtime persistence continues to write complete logs. Multi-session comparison restores every selected persisted or projected fixture through the current build-static Session format catalog before identity redaction and normalization, so retained v0/v1 replay input and fresh `session.v2.jsonl` writer output compare as one v2 logical Session without rewriting or renaming historical files. Expected and harvested logs use the same strict restoration path; source filenames cannot alter format validation. Versionless protocol-adapter unit fixtures remain outside the released Session format corpus. Current v2 fixtures use one row per event; retained v0/v1 fixtures may use canonical packed rows. The [temporary repository migrator](../../../scripts/migrate-packed-session-fixtures.ts) (`pnpm run migrate:packed-session-fixtures`) rewrites older historical layouts, and its [removal proposal](../../../.agents/notes/proposed/process/2026-07-26-remove-packed-session-fixture-migrator.md) owns its deletion.
+`normalizeSessionSnapshot` retains the complete Session header and event payloads but omits top-level `seq`/`time` envelopes from committed fixtures after normalizing paths and scrubbing system-prompt text and tool schemas; it also normalizes embedded stream clocks and historical packed-row `seq0`/`time0` envelopes, and catalog child-creation clocks. Event order and source-event references are preserved. Replay synthesizes the top-level envelopes in memory, while runtime persistence continues to write complete logs. Multi-session comparison validates expected and harvested logs through the strict build-static Session format catalog before identity redaction and normalization; source filenames cannot alter format validation. A retained historical replay input is not the native current-format writer-output oracle: structural migration preserves request meaning but can produce a different event layout. Normalization preserves unexpected request-header fields, including `system`, so regressions remain visible. Versionless protocol-adapter unit fixtures remain outside the released Session format corpus. Fixtures in the [current writer format](../../../docs/session-format-status.md) use one row per event; retained v0/v1 fixtures may use canonical packed rows. The [temporary repository migrator](../../../scripts/migrate-packed-session-fixtures.ts) (`pnpm run migrate:packed-session-fixtures`) rewrites older historical layouts, and its [removal proposal](../../../.agents/notes/proposed/process/2026-07-26-remove-packed-session-fixture-migrator.md) owns its deletion.
 
-Known snapshot spill paths normalize to stable locator tokens, including paths quoted inside JSON omission notices with JSON-escaped Windows separators. Refresh extraction preserves the matched serialized path spelling for literal replacement. Normalization changes only the locator: saved byte lengths and omission counts remain comparison evidence.
+Spill scenarios save through the real local provider under a private temporary root. Their fixture adapter exposes fixed-length logical locators and maps only this run’s saved locators back to real files for retrieval, preserving preview budgets without writing into a shared logical path. Known snapshot spill paths normalize to stable locator tokens, including paths quoted inside JSON omission notices with JSON-escaped Windows separators. Refresh extraction preserves the matched serialized path spelling for literal replacement. Normalization changes only the locator: saved byte lengths and omission counts remain comparison evidence.
+
+Retained historical scenarios keep their canonical Session files unchanged and selected for replay, with no newer canonical sibling in the pinned directory. Their exact normalized native current-format output is recorded separately in `writer.expected.jsonl` for the parent and `writer.<ordinal>.expected.jsonl` for children; these are output oracles, not replay generations. Retained SDK scenarios use `notifications.current.expected.jsonl` for current protocol output. Comparisons neither project current events back into a historical format nor strip structural differences. Independent migration tests verify the official transform rather than treating native writer layout as its expected event sequence.
 
 ### Record, replay, and refresh
 
-`pnpm run test:snapshot:record` calls the live LLM and writes the harvested current generation under its canonical versioned filename. Record and refresh never rename or delete a completed generation, including generations of a child role absent from a later run; reviewed source-tree curation removes a predecessor only after the same role has a verified current replacement. Scenarios with an explicit `sessionFormat` remain read-only in record mode. `pnpm run test:snapshot:refresh` stays keyless, runs the selected highest replay input, and writes stdout, owned prompt and tool-schema sidecars, and a fresh current-generation comparable Session output unless the manifest retains a historical generation. Each composition owner keeps its replay patch beside its live patch; top-level `snapshots/` owns Session-driven scenarios, while other expected outputs stay beside their owning package. [`dsh-llm-replay`](../llm-replay/README.md) serves the recorded streams selected through `DSH_SNAPSHOT_*` environment values.
+`pnpm run test:snapshot:record` calls the live LLM and writes the harvested current generation under its canonical versioned filename. Record and refresh never rename or delete a completed generation, including generations of a child role absent from a later run; reviewed source-tree curation removes a predecessor only after the same role has a verified current replacement. Scenarios with an explicit `sessionFormat` remain read-only in record mode. `pnpm run test:snapshot:refresh` stays keyless, runs the selected highest replay input, and writes stdout, owned prompt and tool-schema sidecars, and a fresh current-generation comparable Session output; retained historical scenarios write the separate writer-output oracles instead of a canonical current-format replay generation. Each composition owner keeps its replay patch beside its live patch; top-level `snapshots/` owns Session-driven scenarios, while other expected outputs stay beside their owning package. [`dsh-llm-replay`](../llm-replay/README.md) serves the recorded streams selected through `DSH_SNAPSHOT_*` environment values.
 
-### Pinning request headers
+### Pinning request headers and system prompts
 
-A pin owns its generated `system-prompt.expected.md` or `tool-schemas.expected.json` sidecar by default; `systemPromptSource` and `toolSchemasSource` name another pin when the complete corresponding sequence is identical, so each distinct version is committed once. The selected highest parent fixture stores `"system":"{{system}}","tools":"{{tools}}"` while retaining config, reason, and any model-visible prefix. A child Session whose own scope composes a different request declares it per fixture index with `pinsChildToolSchemas` and `pinsChildSystemPrompts`. A scenario that changes the request header mid-run declares `expectedHeaderChanges`.
+A pin owns its generated `system-prompt.expected.md` or `tool-schemas.expected.json` sidecar by default; `systemPromptSource` and `toolSchemasSource` name another pin when the complete corresponding sequence is identical, so each distinct version is committed once. The system prompt is surface node 0, logged as a `system/message` event before the step's first `request/header`; every fixture stores its text block as `"text":"{{system}}"` and the prompt sidecar keeps the full text. The pin's `request/header` events store `"tools":"{{tools}}"` while retaining config and reason, and the structured schema sidecar keeps the full catalogs. A child Session whose own scope composes a different request declares it per fixture index with `pinsChildToolSchemas` and `pinsChildSystemPrompts`. A scenario that changes the request header mid-run declares `expectedHeaderChanges`; a scenario whose prompt changes mid-run — replacing node 0, or appending after the cached history on an `in-history` route — declares `expectedPromptChanges`, and each change adds a `<!-- system/message change N -->` section to the prompt sidecar. The manifest spells these `header.changes` and `header.promptChanges`.
 
 ### Platform and composition variants
 
-A scenario requiring a non-Windows host declares `posixOnly`, which skips its run test on Windows while the fixture guards keep covering its committed files everywhere; a scenario whose composition needs a usable `pwsh` declares `pwshOnly`. `workspaceParent` moves the generated child cwd outside the platform temp directory when temporary-directory grants are themselves under test; a scenario's committed `workspace/` is copied into that child first, then `prepareWorkspace` runs against the generated cwd before the agent starts. Default generated workspaces are stored in session fixtures as `{{cwd}}`, so platform temp roots and random basenames do not affect recordings.
+A scenario requiring a non-Windows host declares `posixOnly`, which skips its run test on Windows while the fixture guards keep covering its committed files everywhere; a scenario whose composition needs a usable `pwsh` declares `pwshOnly`. `workspaceParent` moves the generated child cwd outside the platform temp directory when temporary-directory grants are themselves under test; a scenario's committed `workspace/` is copied into that child first, then `prepareWorkspace` runs against the generated cwd before the agent starts. Default generated workspaces are stored in session fixtures as `{{cwd}}`, so platform temp roots and random basenames do not affect recordings. Headless manifests use `workspace.parent: outside-temp` when the Session workspace grant itself is under test. The adapter allocates beside the platform temp root when its parent is writable and outside system temporary grants, otherwise under home, and rejects any generated cwd already covered by automatic temporary write grants.
 
 ### What can go wrong
 
-- **A fixture guard rejects the committed files** — orphan scenario dirs, missing files, multiple pins for one header class, duplicate sidecar content, unscrubbed JSONL headers, and malformed pinning headers all fail the suite before comparisons run.
+- **A child turn wait fails** — `waitForSubagentTurnEnd` identifies the child, requested turn, and deadline even when the first log harvest exceeds that deadline, and retains the underlying failure as the error cause.
+- **A fixture guard rejects the committed files** — orphan scenario dirs, missing files, multiple pins for one header class, duplicate sidecar content, unscrubbed prompt text or tool schemas, a `request/header` with no preceding `system/message`, and malformed pinning headers all fail the suite before comparisons run.
 - **The session harvest needs raw JSONL mode** — snapshot configs set the JSONL backend's `compression: 'none'`; compressed JSONL has no snapshot-harvest path.
 - **Built mode needs current artifacts** — run `pnpm run build` before selecting `DSH_EXAMPLE_MODE=lib`; source mode remains the zero-build path.
 
@@ -104,7 +107,7 @@ This section explains the design of the kit; the observable behavior is fully co
 
 ### Design
 
-The shared core owns manifests, generation-qualified role selection, workspace setup/comparison, typed identity mapping, normalizers, and fixture invariants. The ACP adapter adds four composable layers: launcher, scenario harness, normalizers, and suite factory. `launchAcpTestAgent` boots a source profile under tsx or a built `lib` profile under plain Node, connects the SDK client over a raw-byte stdout tee, collects Session updates and stderr, fails closed on unhandled permission requests, and owns shutdown. `runScenario` drives ACP JSON-RPC stdio and harvests the numerically highest persisted raw JSONL generation for every Session directory. The pure normalizers replace cwd paths and typed identities with stable tokens, zero times, expand physical provenance ranges, and scrub request-header bulk. `defineAcpSnapshotSuite` registers comparisons, generation-qualified fixture write-back, and the live uniformity guard.
+The shared core owns manifests, generation-qualified role selection, workspace setup/comparison, typed identity mapping, normalizers, and fixture invariants. The ACP adapter adds four composable layers: launcher, scenario harness, normalizers, and suite factory. `launchAcpTestAgent` boots a source profile under tsx or a built `lib` profile under plain Node, connects the SDK client over a raw-byte stdout tee, collects Session updates and stderr, fails closed on unhandled permission requests, and owns shutdown. `runScenario` drives ACP JSON-RPC stdio and harvests the numerically highest persisted raw JSONL generation for every Session directory. The pure normalizers replace cwd paths and typed identities with stable tokens, zero times, expand physical provenance ranges, and scrub system-prompt text and tool-schema bulk. `defineAcpSnapshotSuite` registers comparisons, generation-qualified fixture write-back, and the live uniformity guard.
 
 ### Source map
 
@@ -123,7 +126,7 @@ The shared core owns manifests, generation-qualified role selection, workspace s
 
 ### Data flow
 
-A scenario runs the agent under the launcher, feeds it the input script through the harness, and captures stdout plus the persisted logs. The normalizers canonicalize those captures — ids to first-seen sequence, generated cwd to `{{cwd}}`, header bulk to `{{system}}`/`{{tools}}` — so recorded and fresh runs compare structurally. The factory then compares normalized stdout and re-persisted logs against committed fixtures, or writes them back in record/refresh mode, and its guards reject malformed or drifting fixtures before any comparison result is trusted.
+A scenario runs the agent under the launcher, feeds it the input script through the harness, and captures stdout plus the persisted logs. The normalizers canonicalize those captures — ids to first-seen sequence, generated cwd to `{{cwd}}`, `system/message` text to `{{system}}` and header tool schemas to `{{tools}}` — so recorded and fresh runs compare structurally. The factory then compares normalized stdout and re-persisted logs against committed fixtures, or writes them back in record/refresh mode, and its guards reject malformed or drifting fixtures before any comparison result is trusted.
 
 </details>
 

package/README.zh.md

@@ -68,27 +68,30 @@ defineAcpSnapshotSuite({
 })
 ```
 
-每个已记录 Session 目录携带封闭的 `snapshot.yml` manifest，以及规范 parent 与连续 child 角色。parent 文件名是 `session[.vN].jsonl`；child 是 `session.<ordinal>[.vN].jsonl`；v0 省略 `.v0`，正版本使用小写 `.vN`，且每个文件名与其 header 一致。一个角色可以保留旧 generation，但 harness 会选择数值最高的一项。拥有 fixture 的 manifest 可以声明 `sessionFormat.version` 与一个或多个封闭 `coverage` 名称，把该历史 generation 保留为显式迁移 fixture；省略此字段时跟随当前 writer。manifest 还会指名场景、随附 profile、组合／header 类别、录制来源，以及已完成 Session 无法重建的 replay、平台、权限、环境、workspace 或输入事实。适配器注册预期输出、Session 日志与可选 `workspace.expected/` 比较；保护会拒绝遗留目录、缺失角色、非规范名称、绝对路径、malformed manifest 与平台专用分隔符。
+每个已记录 Session 目录携带封闭的 `snapshot.yml` manifest，以及规范 parent 与连续 child 角色。parent 文件名是 `session[.vN].jsonl`；child 是 `session.<ordinal>[.vN].jsonl`；v0 省略 `.v0`，正版本使用小写 `.vN`，且每个文件名与其 header 一致。一个角色可以保留旧 generation，但 harness 会选择数值最高的一项。拥有 fixture 的 manifest 可以声明 `sessionFormat.version` 与一个或多个封闭 `coverage` 名称，把该历史 generation 保留为显式迁移 fixture；省略此字段时跟随当前 writer。manifest 还会指名场景、随附 profile、组合／header 类别、录制来源，以及已完成 Session 无法重建的 replay、平台、权限、环境、workspace 或输入事实。存储保护检查每个选定 parent 与 child 角色的工具结果和可移植路径。提示词／schema 擦除、消息身份及提示词先于请求的顺序检查适用于当前 generation；保留的前代维持其历史表示。适配器注册预期输出、Session 日志与可选 `workspace.expected/` 比较；保护会拒绝遗留目录、缺失角色、非规范名称、绝对路径、malformed manifest 与平台专用分隔符。
 
-`normalizeSessionSnapshot` 在规范化路径并清理 request header 后，会保留完整 Session header 与事件 payload，但从已提交 fixture 中省略顶层 `seq`/`time` envelope；它还会规范化嵌入式 stream clock 与历史 packed-row 的 `seq0`/`time0` envelope。Replay 只在内存中合成顶层 envelope，而运行时持久化仍写入完整日志。多 Session 比较会先通过当前构建期静态 Session 格式目录恢复每个选定的持久化或投影 fixture，再进行身份脱敏与规范化，因此保留的 v0/v1 replay 输入与新生成的 `session.v2.jsonl` writer 输出会作为同一个 v2 logical Session 比较，且不会重写或重命名历史文件。预期日志与收集日志使用同一条严格恢复路径；来源文件名不能改变格式校验。无版本的协议适配器单元测试 fixture 不属于已发布 Session 格式语料。当前 v2 fixture 每个事件占一行；保留的 v0/v1 fixture 可以使用规范 packed row。[临时仓库迁移器](../../../scripts/migrate-packed-session-fixtures.ts)（`pnpm run migrate:packed-session-fixtures`）会改写更旧的历史布局，由其[移除提案](../../../.agents/notes/proposed/process/2026-07-26-remove-packed-session-fixture-migrator.zh.md)负责删除该迁移器。
+`normalizeSessionSnapshot` 在规范化路径并擦除系统提示文本与工具 schema 后，会保留完整 Session header 与事件 payload，但从已提交 fixture 中省略顶层 `seq`/`time` envelope；它还会规范化嵌入式 stream clock 与历史 packed-row 的 `seq0`/`time0` envelope 与 catalog child 创建时钟。事件顺序与来源事件引用保持不变。Replay 只在内存中合成顶层 envelope，而运行时持久化仍写入完整日志。多 Session 比较会先通过严格的构建期静态 Session 格式目录校验预期日志与收集日志，再进行身份脱敏与规范化；来源文件名不能改变格式校验。保留的历史 replay 输入不是原生当前格式 writer 输出的比较基准：结构迁移保留请求含义，但可以产生不同的事件布局。归一化保留意外的 request-header 字段（包括 `system`），使回归保持可见。无版本的协议适配器单元测试 fixture 不属于已发布 Session 格式语料。[当前写入器格式](../../../docs/session-format-status.zh.md)的 fixture 每个事件占一行；保留的 v0/v1 fixture 可以使用规范 packed row。[临时仓库迁移器](../../../scripts/migrate-packed-session-fixtures.ts)（`pnpm run migrate:packed-session-fixtures`）会改写更旧的历史布局，由其[移除提案](../../../.agents/notes/proposed/process/2026-07-26-remove-packed-session-fixture-migrator.zh.md)负责删除该迁移器。
 
-已知的快照 spill 路径会规范化为稳定的定位信息 token，包括 JSON 省略通知中带引号、使用 JSON 转义 Windows 分隔符的路径。刷新提取会保留匹配路径的序列化写法，以便进行字面替换。规范化只改变定位信息：保存字节数与省略计数仍作为比较证据。
+Spill 场景通过真实本地 provider 保存到私有临时根目录。夹具适配器提供固定长度的逻辑定位符，并仅将本次运行已保存的定位符映射回实际文件以供检索，在不写入共享逻辑路径的情况下保留预览预算。已知的快照 spill 路径会规范化为稳定的定位信息 token，包括 JSON 省略通知中带引号、使用 JSON 转义 Windows 分隔符的路径。刷新提取会保留匹配路径的序列化写法，以便进行字面替换。规范化只改变定位信息：保存字节数与省略计数仍作为比较证据。
+
+保留历史输入的场景保持规范 Session 文件不变，并继续选择它们进行回放；固定历史版本的目录中没有更新的规范同角色文件。其精确的规范化原生当前格式输出单独记录在父会话的 `writer.expected.jsonl` 和子会话的 `writer.<ordinal>.expected.jsonl` 中；这些是输出比较基准，而非 replay 代际。保留历史输入的 SDK 场景使用 `notifications.current.expected.jsonl` 记录当前协议输出。比较既不将当前事件反向投影为历史格式，也不剥除结构差异。独立迁移测试验证正式转换，而不把原生 writer 布局当作其预期事件序列。
 
 ### 录制、回放与刷新
 
-`pnpm run test:snapshot:record` 调用在线 LLM（大语言模型），并在规范具名版本文件下写入收集到的当前 generation。record 与 refresh 绝不重命名或删除已完成的 generation，即使后续运行不再产生某个 child 角色也一样；受审阅的源树整理只有在同角色存在已验证的当前替代文件后才移除前代。显式声明 `sessionFormat` 的场景在录制模式下保持只读。`pnpm run test:snapshot:refresh` 保持无密钥，运行选定的最高 replay 输入，并写入 stdout、各 pin 自有的 prompt 与工具 schema sidecar；只有 manifest 未保留历史 generation 时，才写入新鲜当前 generation 的可比较 Session 输出。每个组合 owner 把 replay patch 放在 live patch 旁；顶层 `snapshots/` 拥有 Session 驱动场景，其他预期输出留在其 package owner 旁。[`dsh-llm-replay`](../llm-replay/README.zh.md) 提供通过 `DSH_SNAPSHOT_*` 环境值选择的已记录流。
+`pnpm run test:snapshot:record` 调用在线 LLM（大语言模型），并在规范具名版本文件下写入收集到的当前 generation。record 与 refresh 绝不重命名或删除已完成的 generation，即使后续运行不再产生某个 child 角色也一样；受审阅的源树整理只有在同角色存在已验证的当前替代文件后才移除前代。显式声明 `sessionFormat` 的场景在录制模式下保持只读。`pnpm run test:snapshot:refresh` 保持无密钥，运行选定的最高 replay 输入，并写入 stdout、各 pin 自有的 prompt 与工具 schema sidecar，以及新鲜当前 generation 的可比较 Session 输出；保留历史输入的场景写入单独的 writer 输出比较基准，而非规范当前格式 replay 代际。每个组合 owner 把 replay patch 放在 live patch 旁；顶层 `snapshots/` 拥有 Session 驱动场景，其他预期输出留在其 package owner 旁。[`dsh-llm-replay`](../llm-replay/README.zh.md) 提供通过 `DSH_SNAPSHOT_*` 环境值选择的已记录流。
 
-### 固定请求 header
+### 固定请求 header 与系统提示
 
-每个 pin 默认拥有其生成的 `system-prompt.expected.md` 或 `tool-schemas.expected.json` sidecar；当完整的对应序列相同时，`systemPromptSource` 与 `toolSchemasSource` 指定另一个 pin 作为来源，因此每个不同版本只提交一次。选定的最高 parent fixture 存储 `"system":"{{system}}","tools":"{{tools}}"`，同时保留配置、原因与任何模型可见前缀。自身 scope 组合出不同请求的 child Session 按 fixture index 以 `pinsChildToolSchemas` 与 `pinsChildSystemPrompts` 单独声明。运行中改变请求 header 的场景声明 `expectedHeaderChanges`。
+每个 pin 默认拥有其生成的 `system-prompt.expected.md` 或 `tool-schemas.expected.json` sidecar；当完整的对应序列相同时，`systemPromptSource` 与 `toolSchemasSource` 指定另一个 pin 作为来源，因此每个不同版本只提交一次。系统提示是 surface 节点 0，作为 `system/message` 事件记录在该步骤第一个 `request/header` 之前；每个 fixture 把其文本块存储为 `"text":"{{system}}"`，提示词 sidecar 保留完整文本。该 pin 的 `request/header` 事件存储 `"tools":"{{tools}}"`，同时保留配置与原因，结构化 schema sidecar 保留完整目录。自身作用域组合出不同请求的 child Session 按 fixture 索引以 `pinsChildToolSchemas` 与 `pinsChildSystemPrompts` 单独声明。运行中改变请求 header 的场景声明 `expectedHeaderChanges`；运行中提示词发生变化的场景——替换节点 0，或在 `in-history` 路由上追加到已缓存历史之后——声明 `expectedPromptChanges`，每次变化在提示词 sidecar 中增加一个 `<!-- system/message change N -->` 小节。manifest 中对应字段为 `header.changes` 与 `header.promptChanges`。
 
 ### 平台与组合变体
 
-需要非 Windows 主机的场景声明 `posixOnly`，在 Windows 上跳过运行测试，但 fixture 保护仍在所有平台覆盖其已提交文件；组合需要可用 `pwsh` 的场景声明 `pwshOnly`。当临时目录授权自身待测时，`workspaceParent` 将生成子级 cwd 移出平台临时区域；场景签入的 `workspace/` 会先复制到该子级，随后 `prepareWorkspace` 在 agent 启动前针对生成 cwd 运行。默认生成的 workspace 在会话 fixture 中存储为 `{{cwd}}`，使平台临时根目录与随机 basename 不影响录制。
+需要非 Windows 主机的场景声明 `posixOnly`，在 Windows 上跳过运行测试，但 fixture 保护仍在所有平台覆盖其已提交文件；组合需要可用 `pwsh` 的场景声明 `pwshOnly`。当临时目录授权自身待测时，`workspaceParent` 将生成子级 cwd 移出平台临时区域；场景签入的 `workspace/` 会先复制到该子级，随后 `prepareWorkspace` 在 agent 启动前针对生成 cwd 运行。默认生成的 workspace 在会话 fixture 中存储为 `{{cwd}}`，使平台临时根目录与随机 basename 不影响录制。 Headless manifest 在测试 Session workspace 授权本身时使用 `workspace.parent: outside-temp`。适配器在父目录可写且位于系统临时授权之外时，于平台临时根目录旁分配目录，否则使用 home，并拒绝已被自动临时写授权覆盖的生成 cwd。
 
 ### 可能出什么问题
 
-- **fixture 保护拒绝已提交文件**——遗留场景目录、缺失文件、一个 header 类别包含多个 pin、重复的伴随文件内容、未擦除的 JSONL header 与格式错误的 pin header 都会在比较运行前使套件失败。
+- **子会话轮次等待失败**——即使首次日志收集就超过期限，`waitForSubagentTurnEnd` 也会指出子会话、目标轮次与等待期限，并通过错误的 cause 保留底层失败。
+- **fixture 保护拒绝已提交文件**——遗留场景目录、缺失文件、一个 header 类别包含多个 pin、重复的伴随文件内容、未擦除的提示文本或工具 schema、没有前置 `system/message` 的 `request/header`，以及格式错误的 pin header 都会在比较运行前使套件失败。
 - **会话收集需要原始 JSONL mode**——快照配置使用 JSONL 后端的 `compression: 'none'`；压缩 JSONL 没有快照收集路径。
 - **构建 mode 需要当前产物**——选择 `DSH_EXAMPLE_MODE=lib` 前先运行 `pnpm run build`；源 mode 仍是零构建路径。
 
@@ -104,7 +107,7 @@ defineAcpSnapshotSuite({
 
 ### 设计
 
-共享核心拥有 manifest、generation 限定角色选择、workspace 设置／比较、类型化身份映射、normalizer 与 fixture 不变式。ACP 适配器增加四个可组合层：launcher、场景 harness、normalizer 与 suite factory。`launchAcpTestAgent` 在 tsx 下启动源码 profile，或在普通 Node 下启动已构建 `lib` profile，通过原始字节 stdout tee 连接 SDK client，收集 Session update 与 stderr，默认拒绝未处理的权限请求，并负责关闭。`runScenario` 驱动 ACP JSON-RPC stdio，并收集每个 Session 目录中数值最高的持久原始 JSONL generation。纯 normalizer 把 cwd 路径与类型化身份变为稳定 token，将时间归零、展开物理来源区间，并清理 request header bulk。`defineAcpSnapshotSuite` 注册比较、generation 限定 fixture 回写与实时一致性保护。
+共享核心拥有 manifest、generation 限定角色选择、workspace 设置／比较、类型化身份映射、normalizer 与 fixture 不变式。ACP 适配器增加四个可组合层：launcher、场景 harness、normalizer 与 suite factory。`launchAcpTestAgent` 在 tsx 下启动源码 profile，或在普通 Node 下启动已构建 `lib` profile，通过原始字节 stdout tee 连接 SDK client，收集 Session update 与 stderr，默认拒绝未处理的权限请求，并负责关闭。`runScenario` 驱动 ACP JSON-RPC stdio，并收集每个 Session 目录中数值最高的持久原始 JSONL generation。纯 normalizer 把 cwd 路径与类型化身份变为稳定 token，将时间归零、展开物理来源区间，并擦除系统提示词文本与工具 schema bulk。`defineAcpSnapshotSuite` 注册比较、generation 限定 fixture 回写与实时一致性保护。
 
 ### 源码地图
 
@@ -119,11 +122,11 @@ defineAcpSnapshotSuite({
 | [`src/workspace.ts`](src/workspace.ts) | 场景 workspace 设置与完整预期状态比较 |
 | [`src/suite.ts`](src/suite.ts) | 场景表套件工厂、fixture 保护、录制/刷新回写 |
 | [`src/index.ts`](src/index.ts) | 再导出四个层的包入口 |
-| — | 不发布运行时不变式伴生入口；消费它的测试套件会检验该工具包。 |
+| — | 不发布运行时不变式伴生入口；该测试支持包不拥有任何生产事件流或可变数据；消费它的测试套件会检验该工具包。 |
 
 ### 数据流
 
-场景在启动器下运行 agent，通过 harness 向它喂入输入脚本，并捕获 stdout 与持久化日志。规范化器把捕获内容规范化——id 转为首次出现序列、生成 cwd 转为 `{{cwd}}`、header bulk 转为 `{{system}}`/`{{tools}}`——使已录制与本次运行可以结构化比较。随后工厂把规范化 stdout 与重新持久化日志同已提交 fixture 比较，或在录制/刷新模式下回写它们；其保护在任何比较结果被采信之前就拒绝畸形或漂移的 fixture。
+场景在启动器下运行 agent，通过 harness 向它喂入输入脚本，并捕获 stdout 与持久化日志。规范化器把捕获内容规范化——id 转为首次出现序列、生成 cwd 转为 `{{cwd}}`、`system/message` 文本转为 `{{system}}`、header 工具 schema 转为 `{{tools}}`——使已录制与本次运行可以结构化比较。随后工厂把规范化 stdout 与重新持久化日志同已提交 fixture 比较，或在录制/刷新模式下回写它们；其保护在任何比较结果被采信之前就拒绝畸形或漂移的 fixture。
 
 </details>