Merge branch 'master' into worktree/merge-compact-card

.agents/notes/implemented/architecture/2026-06-21-mandatory-app-attribution-headers.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 .agents/notes/implemented/architecture/2026-06-21-mandatory-app-attribution-headers.md
-2026-06-21-mandatory-app-attribution-headers.md: a8ffe91c431cdc7907626bbc3eaf8096035777de
-2026-06-21-mandatory-app-attribution-headers.zh.md: ac4affce583d5d81253f320ff022f670d4d66cc8
+2026-06-21-mandatory-app-attribution-headers.md: ad9d65805c8f0c96bd811b5036310d019760627e
+2026-06-21-mandatory-app-attribution-headers.zh.md: 3021c7fcca00f2e929d997625c303f9a27dbf673

.agents/notes/implemented/architecture/2026-06-21-mandatory-app-attribution-headers.md

@@ -32,7 +32,7 @@ The provider-neutral identity is owned by `dsh-llm` (`packages/llm/llm/src/attri
 
 - product token for `User-Agent`: `deepseek-harness` (continuity with the pre-Agent Note wire value and the repo/org identity)
 - version: read from the owning package's manifest via `createRequire`, never a hand-copied constant
-- app URL: `https://github.com/deepseek-ai/deepseek-harness-sdk` - the planned public home; a `FIXME` in `attribution.ts` blocks release until that repository actually exists
+- app URL: `https://github.com/deepseek-ai/deepseek-harness-sdk` - the planned public home, which must exist before release
 
 The default is mandatory and non-empty. White-label deployments pass their own `AppIdentity` to `attributionHeaders(identity)` - the override seam is the function parameter, with no deployment config plumbing until a consumer needs it - and omission falls back to the harness default rather than suppressing attribution. There is no per-request API for the model, user prompt, session id, cwd, user email, API key owner, or local machine identity to influence these fields.
 
@@ -77,7 +77,7 @@ The landed contract:
 
 **Providers see that traffic comes from the harness.** That is the point, but it means deployments that previously blended into generic SDK traffic become identifiable. Mitigation: send only static public product data and let forks/white-label deployments pass their own `AppIdentity`.
 
-**The app URL points at a repository that does not exist yet.** `deepseek-ai/deepseek-harness-sdk` is the planned public home; until it is created the URL is a dangling promise. The `FIXME` marker on the constant blocks a release from shipping with it unresolved (see `docs/development.md` marker semantics).
+**The app URL points at a repository that does not exist yet.** `deepseek-ai/deepseek-harness-sdk` is the planned public home; until it is created the URL is a dangling promise that blocks release.
 
 **Header support differs by client library.** The hand-rolled adapter sets headers directly; the pi-ai-backed adapter depends on pi-ai continuing to honor `StreamOptions.headers` (merged last over provider defaults). The wire-level mock-server tests are the guard: if a pi-ai upgrade stops delivering the header, the suite goes red. This is useful pressure on the abstraction: a provider adapter that cannot set mandatory headers cannot fully implement the harness LLM contract.
 

.agents/notes/implemented/architecture/2026-06-21-mandatory-app-attribution-headers.zh.md

@@ -32,7 +32,7 @@ LLM（大语言模型）提供方请求应当标识发出请求的产品。这
 
 - `User-Agent` 的产品 token：`deepseek-harness`（与 Agent Note 之前的线路值及仓库/组织身份保持连续性）
 - 版本：通过 `createRequire` 从所属包的 manifest（元数据清单）读取，绝不手动复制常量
-- 应用 URL：`https://github.com/deepseek-ai/deepseek-harness-sdk`——计划中的公开主页；`attribution.ts` 中的 `FIXME` 标记在该仓库实际存在之前阻塞发布
+- 应用 URL：`https://github.com/deepseek-ai/deepseek-harness-sdk`——计划中的公开主页，且必须在发布前实际存在
 
 默认值是强制的且非空。白标部署通过向 `attributionHeaders(identity)` 传入自己的 `AppIdentity` 来覆盖——覆盖 seam 就是函数参数，在有消费方需要之前不做部署配置管道——省略时回退到 harness 默认值而非抑制归属。没有逐请求 API 允许模型、用户提示词、会话 id、cwd、用户邮箱、API key 所有者或本地机器身份影响这些字段。
 
@@ -77,7 +77,7 @@ LLM（大语言模型）提供方请求应当标识发出请求的产品。这
 
 **提供方看到流量来自 harness。** 这正是目的，但意味着此前混在通用 SDK 流量中的部署变得可识别。缓解措施：仅发送静态公开产品数据，并允许 fork/白标部署传入自己的 `AppIdentity`。
 
-**应用 URL 指向一个尚不存在的仓库。** `deepseek-ai/deepseek-harness-sdk` 是计划中的公开主页；在它创建之前，该 URL 是一个悬空承诺。常量上的 `FIXME` 标记阻塞发布，不允许带着未解决的问题出门（见 `docs/development.md` 标记语义）。
+**应用 URL 指向一个尚不存在的仓库。** `deepseek-ai/deepseek-harness-sdk` 是计划中的公开主页；在它创建之前，该 URL 是一个阻塞发布的悬空承诺。
 
 **不同客户端库的头部支持有差异。** 手写适配器直接设置头部；基于 pi-ai 的适配器依赖 pi-ai 继续尊重 `StreamOptions.headers`（最后合并覆盖提供方默认值）。线路级 mock 服务器测试是守卫：如果 pi-ai 升级后不再投递该头部，套件会变红。这对抽象施加了有益的压力：一个无法设置强制头部的提供方适配器不能完整实现 harness 的 LLM 契约。
 

.agents/notes/implemented/architecture/2026-07-02-tool-render-intent-union.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 .agents/notes/implemented/architecture/2026-07-02-tool-render-intent-union.md
-2026-07-02-tool-render-intent-union.md: d82141f519bff66df000f1316093aacd38b8e42b
-2026-07-02-tool-render-intent-union.zh.md: e145908e019e71765a475b0ca9d22414e9b42d23
+2026-07-02-tool-render-intent-union.md: 67607b2848305439513503d7e03ad5e2a2e4020a
+2026-07-02-tool-render-intent-union.zh.md: 9cbab75ce87f6d313ca4b6ac8dda043733361f0d

.agents/notes/implemented/architecture/2026-07-02-tool-render-intent-union.md

@@ -14,7 +14,7 @@ A tool declares how its calls render in a UI (an editor's tool-call card) throug
 - Which combinations are *valid* is unwritten: a `terminal` call that also sets `content` means "description above the card"; a generic call that sets `terminal` is meaningless but representable. The type permits nonsense.
 - There is no way to express the one file-tool affordance an editor most wants — a **diff card** (`{path, oldText, newText}`, which Zed renders as an inline diff / new-file preview). `ToolCallPresentation.content` is the *LLM* `ContentBlock[]` vocabulary (text/image), so a tool literally cannot ask for a diff.
 
-The existing `FIXME(tool-presentation)` in `packages/core/tools/src/index.ts` named the fix: "redesign the type so a tool declares its render INTENT once (e.g. a tagged union over card kinds) rather than a bag of optional fields the bridge stitches together." An earlier rejected collapse-tool-owned-presentation proposal deferred it explicitly: rich rendering "should return later as a tagged render-intent union after there are at least two real tools and two real consumers to validate the vocabulary." That bar is met by multiple producer families plus the TUI and host/client-runtime (Web) consumers.
+An earlier rejected collapse-tool-owned-presentation proposal deferred rich rendering until it could "return later as a tagged render-intent union after there are at least two real tools and two real consumers to validate the vocabulary." That bar is met by multiple producer families plus the TUI and host/client-runtime (Web) consumers.
 
 ## Decision
 

.agents/notes/implemented/architecture/2026-07-02-tool-render-intent-union.zh.md

@@ -14,7 +14,7 @@ Status: implemented
 - 哪些组合是*合法的*没有文档说明：一个设置了 `content` 的 `terminal` 调用意味着「卡片上方的描述」；一个设置了 `terminal` 的 generic 调用毫无意义但类型上可表达。类型允许无意义的状态存在。
 - 无法表达编辑器最需要的文件工具能力：**diff 卡片**（`{path, oldText, newText}`，Zed 将其渲染为内联 diff / 新文件预览）。`ToolCallPresentation.content` 使用的是 *LLM（大语言模型）* 的 `ContentBlock[]` 词汇（text/image），工具根本无法请求 diff 展示。
 
-`packages/core/tools/src/index.ts` 中已有的 `FIXME(tool-presentation)` 指出了修复方向：「重新设计类型，让工具一次性声明其渲染意图（例如按卡片种类的带标签联合类型），而非一堆由 bridge 拼接的可选字段。」一个早先被否决的折叠工具自有呈现提案明确推迟了此事：富渲染「应当在至少有两个真实工具和两个真实消费方验证词汇之后，以带标签 render-intent 联合类型的形式回归。」该条件已由多个生产者族，加上 TUI 与宿主/客户端运行时（Web）这些消费方满足。
+一个早先被否决的折叠工具自有呈现提案把富渲染推迟到它能够「在至少有两个真实工具和两个真实消费方验证词汇之后，以带标签 render-intent 联合类型的形式回归」之时。该条件已由多个生产者族，加上 TUI 与宿主/客户端运行时（Web）这些消费方满足。
 
 ## 决策
 

.agents/notes/implemented/architecture/2026-07-05-prompt-variables-and-tool-guidance-ownership.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 .agents/notes/implemented/architecture/2026-07-05-prompt-variables-and-tool-guidance-ownership.md
-2026-07-05-prompt-variables-and-tool-guidance-ownership.md: 94f5fa409e7b539b48750d12576c7a342a30c9ba
-2026-07-05-prompt-variables-and-tool-guidance-ownership.zh.md: 341b3a89f423c9cc7d2fc56f1ea25a1985680d0d
+2026-07-05-prompt-variables-and-tool-guidance-ownership.md: a3b5021daf323971308760bde4f97651db8edbda
+2026-07-05-prompt-variables-and-tool-guidance-ownership.zh.md: 490d41302ea4f4e14e11e301acbf47170f95cace

.agents/notes/implemented/architecture/2026-07-05-prompt-variables-and-tool-guidance-ownership.md

@@ -10,7 +10,7 @@ The assembled system prompt had four defects, all of one family: facts the harne
 
 **The model could not know its own name.** `AgentOptions.model` drives every request, but no prompt text carried it — and nothing COULD carry it: sections in `dsh-system-prompt` were context-global while the model name is per-agent, and `assemble()` took no per-agent input at all.
 
-**Tool guidance was hand-written prose in leaf YAML.** The bash/subagent/todo_write usage guidance lived in the coding-agent and ACP persona strings — two drifting copies (the ACP one was already abridged) — while `dsh-tool-fs` and `dsh-tool-web` owned their guidance as `ctx.systemPrompt.section()` contributions. Loading or dropping a tool plugin meant editing every deployment's persona by hand; both YAMLs carried a `FIXME(config-comments)` apologizing for a symptom of the split, and the old terminal welcome banner hand-enumerated the tool set too.
+**Tool guidance was hand-written prose in leaf YAML.** The bash/subagent/todo_write usage guidance lived in the coding-agent and ACP persona strings — two drifting copies (the ACP one was already abridged) — while `dsh-tool-fs` and `dsh-tool-web` owned their guidance as `ctx.systemPrompt.section()` contributions. Loading or dropping a tool plugin meant editing every deployment's persona by hand, and the old terminal welcome banner hand-enumerated the tool set too.
 
 **The persona rendered after tool guidance.** The loop string-joined `agent.options.systemPrompt` AFTER the assembled sections, so the model read "Use the read tool…" before "You are a coding agent" — backwards relative to the identity-first convention (Claude Code, Codex) and a second composition path besides the section pipeline.
 

.agents/notes/implemented/architecture/2026-07-05-prompt-variables-and-tool-guidance-ownership.zh.md

@@ -10,7 +10,7 @@ Status: implemented
 
 **模型无法知道自己的名字。** `AgentOptions.model` 驱动每个请求，但没有任何提示词文本携带它——也不可能携带：`dsh-system-prompt` 中的 section 是上下文全局的，而模型名称是 per-agent 的，`assemble()` 根本不接受任何 per-agent 输入。
 
-**工具指导是 leaf YAML 中的手写行文。** bash/subagent/todo_write 的使用指导存放在 coding-agent 和 ACP persona 字符串里——两份漂移的副本（ACP 那份已经被删减）——而 `dsh-tool-fs` 和 `dsh-tool-web` 则通过 `ctx.systemPrompt.section()` 贡献各自的指导。加载或卸载一个工具插件意味着手动编辑每个部署的 persona；两份 YAML 都带着一条 `FIXME(config-comments)` 为这种分裂的症状道歉，旧终端欢迎横幅也手动枚举了工具集。
+**工具指导是 leaf YAML 中的手写行文。** bash/subagent/todo_write 的使用指导存放在 coding-agent 和 ACP persona 字符串里——两份漂移的副本（ACP 那份已经被删减）——而 `dsh-tool-fs` 和 `dsh-tool-web` 则通过 `ctx.systemPrompt.section()` 贡献各自的指导。加载或卸载一个工具插件意味着手动编辑每个部署的 persona，旧终端欢迎横幅也手动枚举了工具集。
 
 **Persona 渲染在工具指导之后。** agent loop（智能体循环）将 `agent.options.systemPrompt` 字符串拼接在已组装的 section 之后，于是模型先读到「Use the read tool…」再读到「You are a coding agent」——与 identity-first 约定（Claude Code、Codex）相反，且是 section 流水线之外的第二条组合路径。
 

.agents/notes/implemented/architecture/2026-07-05-reconstructable-requests.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 .agents/notes/implemented/architecture/2026-07-05-reconstructable-requests.md
-2026-07-05-reconstructable-requests.md: 2f559a3052b9fb84f788975a64799e4f020b0d3e
-2026-07-05-reconstructable-requests.zh.md: 26abdc024a166856e51ebf09f086c7868fc8236d
+2026-07-05-reconstructable-requests.md: ebca9b99cad791159302da9c2bbce9f4df147aab
+2026-07-05-reconstructable-requests.zh.md: 91ef3fd04502f2c2f092da60a8bd909cf7fa25df

.agents/notes/implemented/architecture/2026-07-05-reconstructable-requests.md

@@ -53,4 +53,3 @@ Like MiniCode, the conversation advances append-only and resets only when model-
 - Tool-result trimming (planned) needs no new mechanism: a logged single-entry surface replace (`start === end`) carrying a trimmed `tool/result` under the same `callId` — compaction-family, replay-correct, cache-bust batched by the same pressure logic.
 - Session logs grow one `request/header` snapshot per loop instance plus snapshots on real changes. This is larger than a delta codec but small beside chunk-heavy logs and retains one replay representation. `SESSION_FORMAT_VERSION` stays `0`; legacy delta events are rejected rather than migrated.
 - Snapshot expected outputs changed once (every transcript gains its header events); the fs-writing fixtures are stored in the normalized authored form with cwd-relative tool arguments, because replay only round-trips cwd-independent argument paths.
-- FIXME(call-config-shape): revisit `LlmCallConfig`'s exact field set — which fields are genuinely epoch-level for cache purposes (`model` certainly; the sampling scalars sit there out of caution), and where provider-specific extras (reasoning options, extra body params) belong when an adapter needs them.

.agents/notes/implemented/architecture/2026-07-05-reconstructable-requests.zh.md

@@ -53,4 +53,3 @@ Status: implemented
 - 工具结果裁剪（计划中）无需新机制：一个已记录的单条目 surface replace（`start === end`），携带同一 `callId` 下裁剪后的 `tool/result`——属压缩家族，回放正确，缓存击穿由相同的压力逻辑批量处理。
 - 会话日志每个循环实例增长一个 `request/header` 快照，并在真正变更时增加快照。它比 delta 编解码器更大，但相对分片密集型日志仍然很小，并只保留一种回放表示。`SESSION_FORMAT_VERSION` 保持 `0`；旧的 delta 事件被拒绝而非迁移。
 - 快照预期输出变更一次（每个 transcript（文本记录）增加其 header 事件）；写入文件系统的 fixture（测试前置数据）以规范化的撰写形式存储，工具参数使用 cwd 相对路径，因为回放只对 cwd 无关的参数路径做往返。
-- FIXME(call-config-shape)：重新审视 `LlmCallConfig` 的确切字段集——哪些字段对缓存而言真正属于 epoch 级别（`model` 毫无疑问；采样标量出于谨慎放在那里），以及当适配器需要时，提供方特定的额外项（推理（reasoning）选项、额外 body 参数）应归属何处。

.agents/notes/implemented/architecture/2026-08-03-pi-ai-declared-provider-catalog.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 .agents/notes/implemented/architecture/2026-08-03-pi-ai-declared-provider-catalog.md
-2026-08-03-pi-ai-declared-provider-catalog.md: d75b6bdb91d60026636bf320f8c6625590849a41
-2026-08-03-pi-ai-declared-provider-catalog.zh.md: f8dba9900b1a7a3abcb16c70a35cc18f0c44219f
+2026-08-03-pi-ai-declared-provider-catalog.md: f908eb6293b77680193fcd8f7be7a9089477855a
+2026-08-03-pi-ai-declared-provider-catalog.zh.md: ce91abd6dc71f790c72766cd3f819096d596182c

.agents/notes/implemented/architecture/2026-08-03-pi-ai-declared-provider-catalog.md

@@ -14,7 +14,7 @@ The adapter also streamed through `streamSimple` from `@earendil-works/pi-ai/com
 
 A provider route is a **declaration**, and the installed catalog is its default. `resolveProfiles` no longer checks route keys against `getBuiltinProviders()`. Instead each route resolves to a materialized model list plus the pi-ai `Provider` that serves it:
 
-- `catalog.ts` merges the installed catalog under the profile's own entries. A profile's `models` list *replaces* the route's catalog (an absent or empty list serves it unchanged), and each entry defaults its unset fields from the installed model of the same `id`. Only the fields the harness consumes are configurable — `id`, `name`, `contextWindow`, `maxTokens`. Pricing and input modalities are absent from the surface because nothing reads them: `replay.ts` zeroes pi-ai's cost metadata and `context.ts` keeps only text blocks. Reasoning is absent for a different reason: a bare capability flag would make pi-ai advertise effort levels with no `thinkingLevelMap` to spell them, so it rides the installed entry or is absent. Materialization spreads the installed entry and overrides those four fields, rather than enumerating the result: an enumerated rebuild silently drops every `Model` field this package does not model, which is how `headers` went missing from an nvidia route once already.
+- `catalog.ts` merges the installed catalog under the profile's own entries. A profile's `models` list *replaces* the route's catalog (an absent or empty list serves it unchanged), and each entry defaults its unset fields from the installed model of the same `id`. Only the fields the harness consumes are configurable — at this note's writing `id`, `name`, `contextWindow`, `maxTokens`; [[2026-08-08-pi-ai-per-model-reasoning-declarations]] later added `reasoningEfforts` and `compat`, which is also where the original "reasoning rides the installed entry or is absent" stance was revisited (a bare capability flag stays rejected; a full per-level declaration with wire spellings does not have its problem). Pricing and input modalities remain absent from the surface because nothing reads them: `replay.ts` zeroes pi-ai's cost metadata and `context.ts` keeps only text blocks. Materialization spreads the installed entry and overrides the configured fields, rather than enumerating the result: an enumerated rebuild silently drops every `Model` field this package does not model, which is how `headers` went missing from an nvidia route once already.
 - `provider.ts` builds the route's `Provider`. A catalog route that keeps its catalog protocol **reuses** the installed provider with `getModels()` replaced; every other route is built by `createProvider()` over a protocol table whose entries are the same `@earendil-works/pi-ai/api/*.lazy` factories pi-ai's own provider factories use. That table is narrower than pi-ai's full API set on purpose — it holds only protocols a profile can completely describe with a key, an endpoint, and headers, so Bedrock (SigV4 plus a region), Vertex (project, location, ADC), Azure (provider environment plus an api-version), and Codex (OAuth) are absent rather than offered as routes that cannot authenticate. Catalog routes still reach them through their own provider; only an explicit override is refused.
 - `adapter.ts` turns each resolution into an **immutable snapshot** — the profiles plus a `createModels()` collection holding those providers — and every operation captures a whole snapshot before its first `await`.
 - A model's **explicitly configured** `maxTokens` becomes the seam's `defaultMaxTokens`. The value inherited from the installed catalog does not: pi-ai requires `Model.maxTokens` as the model's output *capability*, while `defaultMaxTokens` is a cap the deployment chose to send on requests that name none, and materializing the former as the latter would start capping every request at a number nobody picked.
@@ -35,7 +35,7 @@ The configurable-provider directory is now the installed catalog **joined with**
 
 pi-ai reports a model with no reasoning metadata as supporting the single level `off`, and the adapter used to pass that straight through. It reaches the seam as a one-item effort list, which every surface renders as a picker holding one selectable control — and that control is a lie: `off` becomes an *omitted* reasoning option at dispatch, byte-for-byte the request that naming no effort already produces. A provider whose own default is to think keeps thinking while the surface shows `off` selected.
 
-`reasoningInfo` therefore omits the seam's `reasoning` field whenever `model.reasoning` is falsy. The condition is the model's own metadata, not where the model came from, so this covers every hand-declared model **and** the 251 installed-catalog models pi-ai marks as non-reasoning. Those previously offered the lone `off`; they now offer nothing, and the surface shows the provider default alone. Models that do carry reasoning metadata are untouched — their level list still crosses the seam unfiltered, `off` included, because there it selects between real alternatives.
+`reasoningInfo` therefore omits the seam's `reasoning` field whenever `model.reasoning` is falsy. The condition is the model's own metadata, not where the model came from, so this covers every hand-declared model whose entry declares no `reasoningEfforts` ([[2026-08-08-pi-ai-per-model-reasoning-declarations]] made declared efforts carry that metadata) **and** the 251 installed-catalog models pi-ai marks as non-reasoning. Those previously offered the lone `off`; they now offer nothing, and the surface shows the provider default alone. Models that do carry reasoning metadata are untouched — their level list still crosses the seam unfiltered, `off` included, because there it selects between real alternatives.
 
 ### Credentials stay outside pi-ai
 
@@ -50,7 +50,7 @@ A route's auth follows from that. A catalog route keeps the installed provider's
 - **Keep `createProvider()` but skip the `Models` collection**, streaming through `provider.streamSimple(model, ctx, {apiKey})`. Smallest diff and the credential path is untouched, but `createProvider`'s `auth` is a required field that this path never invokes — a required-by-signature implementation with no caller. It also leaves `refreshModels` needing a hand-built `RefreshModelsContext`, and keeps the adapter off the runtime pi-ai actually supports.
 - **Reuse the installed provider for catalog routes and `createProvider()` only for declared ones**, with no shared resolution. Zero risk to catalog behavior, but catalog materialization, endpoint override, and per-model configuration would each exist twice, and a catalog route that repoints its protocol would have to jump paths mid-resolution. The chosen split confines the asymmetry to provider construction, where it is forced by pi-ai not exposing a built provider's API implementations.
 - **Rebuild every route through `createProvider()`**, including catalog ones. Fully symmetric, but a built `Provider` does not expose its `api`, so the protocol table would become the ceiling on which providers work — Bedrock loads its Smithy module through a separate entry point and would silently stop working.
-- **Expose pi-ai's whole `Model` shape** (cost, input modalities, `thinkingLevelMap`, `compat`). Maximum configurability, but no current consumer reads those fields, so a configured price or modality would change nothing while reading as supported.
+- **Expose pi-ai's whole `Model` shape** (cost, input modalities, `thinkingLevelMap`, `compat`). Maximum configurability, but no current consumer read those fields then, so a configured price or modality would change nothing while reading as supported. The consumer-driven half of this arrived later: [[2026-08-08-pi-ai-per-model-reasoning-declarations]] opened reasoning (as `reasoningEfforts`, not a raw `thinkingLevelMap`) and the two reasoning-dispatch `compat` switches once selectors and dispatch actually consumed them; cost and modalities stay closed for the original reason.
 
 - **Keep one mutable `Models` collection and re-sync it.** Fewer allocations, and correct for every operation that resolves synchronously. It is exactly wrong for the one that does not: `stream()` awaits a credential between capturing its model and dispatching it.
 - **Simulate an atomic directory swap with dispose-then-register.** No seam change, and it works whenever the new set is valid — which is the case that never needed atomicity.

.agents/notes/implemented/architecture/2026-08-03-pi-ai-declared-provider-catalog.zh.md

@@ -14,7 +14,7 @@ Status: implemented
 
 提供方路由是一份**声明**，已安装 catalog 是它的默认值。`resolveProfiles` 不再拿路由键去核对 `getBuiltinProviders()`，而是把每条路由解析成一份物化模型列表，外加服务它的 pi-ai `Provider`：
 
-- `catalog.ts` 把已安装 catalog 合并到 profile 自身条目之下。profile 的 `models` 列表*替换*该路由的 catalog（列表缺席或为空则原样服务），每个条目从同 `id` 的已安装模型继承自身未设置的字段。只有 harness 会消费的字段可配置——`id`、`name`、`contextWindow`、`maxTokens`。定价与输入模态不出现在配置面，因为没有任何读取方：`replay.ts` 把 pi-ai 的成本元数据清零，`context.ts` 只保留文本块。推理缺席则是另一个理由：一个孤立的能力布尔量会让 pi-ai 公布出没有 `thinkingLevelMap` 可供拼写的档位，因此它沿用已安装条目或直接缺席。物化时以已安装条目铺底、再覆盖那四个字段，而不是逐字段枚举结果：枚举式重建会静默丢弃本包未建模的每一个 `Model` 字段——`headers` 就是这样从某条 nvidia 路由上消失过一次。
+- `catalog.ts` 把已安装 catalog 合并到 profile 自身条目之下。profile 的 `models` 列表*替换*该路由的 catalog（列表缺席或为空则原样服务），每个条目从同 `id` 的已安装模型继承自身未设置的字段。只有 harness 会消费的字段可配置——本 note 写就时为 `id`、`name`、`contextWindow`、`maxTokens`；[[2026-08-08-pi-ai-per-model-reasoning-declarations]] 之后加入了 `reasoningEfforts` 与 `compat`，当初「推理沿用已安装条目或直接缺席」的立场也在那里被重新审视（孤立的能力布尔量仍被拒绝；带 wire 拼写的逐档位完整声明没有它那个问题）。定价与输入模态仍不出现在配置面，因为没有任何读取方：`replay.ts` 把 pi-ai 的成本元数据清零，`context.ts` 只保留文本块。物化时以已安装条目铺底、再覆盖已配置的字段，而不是逐字段枚举结果：枚举式重建会静默丢弃本包未建模的每一个 `Model` 字段——`headers` 就是这样从某条 nvidia 路由上消失过一次。
 - `provider.ts` 构造路由的 `Provider`。保持 catalog 协议不变的 catalog 路由会**复用**已安装提供方，只替换 `getModels()`；其余路由都由 `createProvider()` 基于一张协议表构造，表中条目正是 pi-ai 自己的提供方工厂所用的 `@earendil-works/pi-ai/api/*.lazy` factory。该表刻意窄于 pi-ai 的完整 API 集合——只保留 profile 能用密钥、端点与标头完整描述的协议，因此 Bedrock（SigV4 加 region）、Vertex（project、location、ADC）、Azure（提供方环境加 api-version）与 Codex（OAuth）不在其中，而不是被当作无法认证的路由提供出去。catalog 路由仍可经自己的 provider 抵达它们；被拒的只有显式覆盖。
 - `adapter.ts` 把每次解析变成一份**不可变快照**——profiles 加上持有这些 provider 的 `createModels()` 集合——每个操作都在自己第一个 `await` 之前整体捕获一份。
 - 模型**显式配置**的 `maxTokens` 会成为 seam 的 `defaultMaxTokens`；从已安装 catalog 继承来的那份不会：pi-ai 要求 `Model.maxTokens` 表示模型的输出**能力**，而 `defaultMaxTokens` 是部署选定、发给未点名上限的请求的那个值，把前者物化成后者会让每个请求都被一个无人选择的数字封顶。
@@ -35,7 +35,7 @@ Status: implemented
 
 pi-ai 把没有推理元数据的模型报告为只支持 `off` 一档，而适配器此前原样透传。它抵达 seam 时是一个单元素的 effort 列表，任何界面都会把它渲染成一个只有一项可选控件的选择器——而这个控件在撒谎：`off` 在派发时变成被*省略*的 reasoning 选项，与「不点名任何档位」产出的请求逐字节相同。自身默认就在思考的提供方会继续思考，界面却显示 `off` 已选中。
 
-因此只要 `model.reasoning` 为假，`reasoningInfo` 就省略 seam 的 `reasoning` 字段。判据是模型自身的元数据，而非模型的来源，所以它覆盖每一个手工声明的模型**以及** pi-ai 标记为不具备推理能力的那 251 个已安装 catalog 模型。它们此前提供那个孤零零的 `off`，现在什么也不提供，界面只剩提供方默认。携带推理元数据的模型不受影响——其档位列表仍不经筛选地穿过 seam、`off` 也在内，因为在那里它是在真实备选之间做选择。
+因此只要 `model.reasoning` 为假，`reasoningInfo` 就省略 seam 的 `reasoning` 字段。判据是模型自身的元数据，而非模型的来源，所以它覆盖条目未声明 `reasoningEfforts` 的每一个手工声明模型（[[2026-08-08-pi-ai-per-model-reasoning-declarations]] 让声明的档位携带这份元数据）**以及** pi-ai 标记为不具备推理能力的那 251 个已安装 catalog 模型。它们此前提供那个孤零零的 `off`，现在什么也不提供，界面只剩提供方默认。携带推理元数据的模型不受影响——其档位列表仍不经筛选地穿过 seam、`off` 也在内，因为在那里它是在真实备选之间做选择。
 
 ### 凭据留在 pi-ai 之外
 
@@ -50,7 +50,7 @@ pi-ai 的 `Models` 自带一套凭据概念——按提供方 id 索引的 `Cred
 - **保留 `createProvider()` 但不建 `Models` 集合**，改由 `provider.streamSimple(model, ctx, {apiKey})` 发起。改动最小且凭据路径原封不动，但 `createProvider` 的 `auth` 是必填字段，这条路上它永远不会被调用——一份因签名而必填、却没有调用方的实现。它还让 `refreshModels` 需要手工构造 `RefreshModelsContext`，并使适配器始终不在 pi-ai 真正支持的运行时上。
 - **catalog 路由复用已安装提供方，只有声明式路由走 `createProvider()`**，且两者不共享解析。对 catalog 行为零风险，但 catalog 物化、端点覆盖与每模型配置这三件事都要各写两遍，而改指协议的 catalog 路由还得在解析中途跳到另一条路径。已采纳的拆法把不对称收敛在提供方构造这一处——那里的不对称是 pi-ai 不暴露已构造提供方的 API 实现所强加的。
 - **让每条路由都经 `createProvider()` 重建**，包括 catalog 路由。完全对称，但已构造的 `Provider` 不暴露自己的 `api`，于是协议表会成为「哪些提供方能用」的天花板——Bedrock 经独立入口加载其 Smithy 模块，会因此静默失效。
-- **完整暴露 pi-ai 的 `Model` 形状**（成本、输入模态、`thinkingLevelMap`、`compat`）。可配置性最大，但这些字段当前没有任何读取方，因此配了价格或模态什么也不会改变，却看起来像是受支持的。
+- **完整暴露 pi-ai 的 `Model` 形状**（成本、输入模态、`thinkingLevelMap`、`compat`）。可配置性最大，但这些字段当时没有任何读取方，因此配了价格或模态什么也不会改变，却看起来像是受支持的。这条否决里由消费方驱动的那一半后来兑现了：[[2026-08-08-pi-ai-per-model-reasoning-declarations]] 在选择器与分派真正消费之后开放了推理（以 `reasoningEfforts` 的形态，而非裸 `thinkingLevelMap`）和两个推理分派 `compat` 开关；成本与模态仍因原有理由保持关闭。
 
 - **保留单个可变 `Models` 集合并重新同步。** 分配更少，且对每个同步完成解析的操作都是正确的；唯独对那个不同步的操作恰恰是错的：`stream()` 会在捕获模型与派发模型之间 await 一次凭据。
 - **用「先 dispose 再注册」模拟目录原子替换。** 无需改 seam，且在新集合有效时确实可用——而那正是从不需要原子性的那种情形。

.agents/notes/implemented/bug-fix/2026-07-29-pnpm-setup-runner-isolation.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 .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
+2026-07-29-pnpm-setup-runner-isolation.md: 74b672b3f90ea445ad1a8e283a5904056059b2f8
+2026-07-29-pnpm-setup-runner-isolation.zh.md: 32c667dc09e561504e8e053bf2a338ed2190d9e8

.agents/notes/implemented/bug-fix/2026-07-29-pnpm-setup-runner-isolation.md

@@ -6,7 +6,7 @@ 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.
+`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 the reproducing run, 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
 

.agents/notes/implemented/bug-fix/2026-07-29-pnpm-setup-runner-isolation.zh.md

@@ -6,7 +6,7 @@ Status: implemented
 
 ## 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。
+`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。在复现运行中，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
 

.agents/notes/implemented/bug-fix/2026-08-06-api-key-format-validation.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 .agents/notes/implemented/bug-fix/2026-08-06-api-key-format-validation.md
-2026-08-06-api-key-format-validation.md: e9ca76ede06080f2b868f6436998d163e642adbc
-2026-08-06-api-key-format-validation.zh.md: 5666a884d4c9478291072375681d8d3526b2632a
+2026-08-06-api-key-format-validation.md: 2174cb466c6af72f15005ce1ba3dec8100de6f2f
+2026-08-06-api-key-format-validation.zh.md: f9b7d6518fedc42e5264c46beaa1a78619139c58

.agents/notes/implemented/bug-fix/2026-08-06-api-key-format-validation.md

@@ -14,13 +14,11 @@ Pasting a key containing an emoji, CJK text, or a full-width punctuation mark in
 
 Whitespace passed every check. `ProviderEditor` tested `keyDraft.length`, so a key of three spaces was stored and then authenticated as `Bearer` plus blanks. Neither adapter checked a credential- or environment-sourced key — the path the Models page writes, and therefore the path users actually take.
 
-Sources: deepseek-harness#1594 and #1595; dsh-external#247, #249, #266, and #210.
-
 ## Decision
 
 One rule defines a legal key: **after trimming, non-empty, and every character within `[\x21-\x7E]`** — printable ASCII, space excluded.
 
-This single predicate covers every input the sources list: empty, leading and trailing whitespace, interior whitespace, C0 control characters, emoji, CJK text, and full-width punctuation. It is also exactly the constraint that produced the ByteString failure, so the two issues close on one definition rather than on two coincidentally related fixes.
+This single predicate covers every reported input: empty, leading and trailing whitespace, interior whitespace, C0 control characters, emoji, CJK text, and full-width punctuation. It is also exactly the constraint that produced the ByteString failure, so the failures share one definition rather than two coincidentally related fixes.
 
 A second, narrower rule catches a pasted environment line: input matching `^[A-Z][A-Z0-9_]*=[^=]` or wrapped in matching quotes is refused. Restricting the prefix to upper-case keeps real keys clear of it — `sk-` forms break the identifier match at the hyphen — and requiring a non-`=` character after the separator keeps base64 padding clear of it too. It reports the same format failure as an illegal character rather than its own message: the reader's next move is identical either way, so a separate line would name a cause without changing what to do.
 
@@ -76,7 +74,7 @@ The client cannot import any of this: client packages reference only client pack
 
 **Running the shape heuristic in the resolvers too.** Symmetric, and it would stop a pasted environment line written directly into `.env`. Rejected for the lockout described above: a false positive in a resolver leaves the user no working path, while a false positive in the browser leaves the environment open.
 
-**Probing the provider at save time to prove the key works.** It would close the complaint the sources actually open with — a save that reports success and fails at the first turn. Rejected as out of scope and, on the code as it stood, unbuildable: `discoverModels` short-circuits to the installed catalog before any network call for exactly the providers pi-ai ships catalogs for, so it verified nothing about the key, and the DeepSeek card has no probe at all. A verifier's value is distinguishing "key rejected" from "cannot reach", which is the distinction this change makes reliable; building it first would have produced a verifier unable to tell its own outcomes apart. Comparable products also do not verify on save, so a blocking network call there would be an unexpected behavior rather than a missing one.
+**Probing the provider at save time to prove the key works.** It would close the original complaint — a save that reports success and fails at the first turn. Rejected as out of scope and, on the code as it stood, unbuildable: `discoverModels` short-circuits to the installed catalog before any network call for exactly the providers pi-ai ships catalogs for, so it verified nothing about the key, and the DeepSeek card has no probe at all. A verifier's value is distinguishing "key rejected" from "cannot reach", which is the distinction this change makes reliable; building it first would have produced a verifier unable to tell its own outcomes apart. Comparable products also do not verify on save, so a blocking network call there would be an unexpected behavior rather than a missing one.
 
 ## Consequences
 

.agents/notes/implemented/bug-fix/2026-08-06-api-key-format-validation.zh.md

@@ -14,13 +14,11 @@ Status: implemented
 
 空白字符能通过每一道检查。`ProviderEditor` 判的是 `keyDraft.length`，于是三个空格构成的 Key 会被存下，随后以 `Bearer` 加若干空格去认证。两个适配器都不检查来自凭据或环境的 Key——而那正是 Models 页写入的路径，也就是用户真正走的路径。
 
-来源：deepseek-harness#1594 与 #1595；dsh-external#247、#249、#266、#210。
-
 ## Decision
 
 一条规则定义什么是合法 Key：**trim 之后非空，且每个字符都落在 `[\x21-\x7E]`**——可打印 ASCII，不含空格。
 
-这一个断言覆盖了来源列出的全部输入：空值、首尾空白、中间空白、C0 控制字符、emoji、中文、全角标点。它同时正是造成 ByteString 失败的那条约束，所以两个 issue 收敛于同一个定义，而不是两个恰好相关的修复。
+这一个断言覆盖了所有已报告的输入：空值、首尾空白、中间空白、C0 控制字符、emoji、中文、全角标点。它同时正是造成 ByteString 失败的那条约束，所以这些故障收敛于同一个定义，而不是两个恰好相关的修复。
 
 第二条更窄的规则用于识别整行粘贴的环境变量：匹配 `^[A-Z][A-Z0-9_]*=[^=]` 或首尾成对引号的输入会被拒绝。把前缀限定为全大写可以让真实 Key 与之绝缘——`sk-` 这类形态会在连字符处中断标识符匹配——而要求分隔符之后必须是非 `=` 字符，则让 base64 的 padding 也与之绝缘。它报出的是与非法字符相同的那条格式失败，而不是自己的一句：读到它的人下一步动作完全一样，因此单列一句只会点出一个原因，却不改变该怎么做。
 
@@ -76,7 +74,7 @@ Status: implemented
 
 **让形状启发式也在 resolver 中运行。** 更对称，且能拦住直接写进 `.env` 的整行环境变量。因上文所述的锁死风险而否决：resolver 中的一次误判会让用户无路可走，浏览器中的一次误判则仍留有环境变量这条路。
 
-**在保存时探测 provider 以证明 Key 可用。** 它能关掉来源真正开篇抱怨的那件事——保存报成功、第一轮才失败。因超出范围而否决，且在当时的代码上无法建成：对 pi-ai 恰好自带 catalog 的那些 provider，`discoverModels` 会在任何网络调用之前短路到内置 catalog，因而对 Key 什么都验证不了；而 DeepSeek 卡片根本没有探测。验证器的价值在于分清「Key 被拒」与「无法连通」，而这正是本次改动让其变得可靠的区分；先建验证器只会得到一个分不清自身结果的验证器。同类产品也不在保存时验证，因此保存时的阻断式网络调用会是一个意外行为，而非一处缺失。
+**在保存时探测 provider 以证明 Key 可用。** 它能关掉最初报告的那件事——保存报成功、第一轮才失败。因超出范围而否决，且在当时的代码上无法建成：对 pi-ai 恰好自带 catalog 的那些 provider，`discoverModels` 会在任何网络调用之前短路到内置 catalog，因而对 Key 什么都验证不了；而 DeepSeek 卡片根本没有探测。验证器的价值在于分清「Key 被拒」与「无法连通」，而这正是本次改动让其变得可靠的区分；先建验证器只会得到一个分不清自身结果的验证器。同类产品也不在保存时验证，因此保存时的阻断式网络调用会是一个意外行为，而非一处缺失。
 
 ## Consequences
 

.agents/notes/implemented/feature/2026-06-18-compaction-capability-seam.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 .agents/notes/implemented/feature/2026-06-18-compaction-capability-seam.md
-2026-06-18-compaction-capability-seam.md: 26e6e2468c7bea661d85c8fb994adf8b109105ee
-2026-06-18-compaction-capability-seam.zh.md: 8f9cd1f6bc31e648a5b923e816cad75ebc1a0bd8
+2026-06-18-compaction-capability-seam.md: 8dcbe74429a620027a570124383442b969c12196
+2026-06-18-compaction-capability-seam.zh.md: 27b63f29c2e6f35637185b47c882ae42e5d41088

.agents/notes/implemented/feature/2026-06-18-compaction-capability-seam.md

@@ -131,4 +131,4 @@ The lifecycle boundary makes crash state unambiguous:
 - **Loop:** Tests pin pre-step after the preceding `step/end` and before the next `step/start`, actual `agent/request` routing, closed failed steps, fresh retry numbering, and complete thrown/in-band overflow → compaction → reconstructed retry composition.
 - **Manual:** Maintenance serialization, marker ordering, injection retention, live/stale orphan classification, cancellation, close/flush failures, command mapping, and the queued TUI journey are pinned without a model key.
 - **With-key e2e:** A real model and bash session with lowered limits triggers compaction, records a complete `compact/start…end` pair, shrinks the surface, and finishes the task.
-- **Snapshot gap:** Runaway-turn compaction cannot yet replay because the summarization call records no `assistant/chunk` events or `sessionId`; interleaved summarization-call replay remains follow-up work.
+- **Snapshot gap:** The summarization call is session-associated and logs `compact/summary`, but ordinary transcript replay does not derive its auxiliary response; keyless assembled coverage therefore needs an explicit replay override.

.agents/notes/implemented/feature/2026-06-18-compaction-capability-seam.zh.md

@@ -131,4 +131,4 @@ compact/end      → log-only. Releases the lock (carries `error` on a recoverab
 - **循环测试：** 测试固定 pre-step 发生在前一个 `step/end` 之后、下一个 `step/start` 之前，使用实际 `agent/request` 路由，关闭失败步骤，分配新的重试编号，并覆盖完整的抛出/带内溢出 → 压缩 → 重建重试组合。
 - **手动测试：** 无需模型密钥即可固定 maintenance 串行化、标记顺序、注入保留、活动／陈旧未匹配标记分类、取消、闭合／flush 失败、命令映射以及排队 TUI 流程。
 - **带密钥 e2e：** 真实模型和 bash 会话在降低的限制下触发压缩，记录完整的 `compact/start…end` 对，缩小 surface，并完成任务。
-- **快照缺口：** 失控轮次压缩尚无法回放，因为摘要调用未记录 `assistant/chunk` 事件或 `sessionId`；交错摘要调用的回放仍是后续工作。
+- **快照缺口：** 摘要调用与会话关联并记录 `compact/summary`，但普通 transcript（文本记录）回放不会派生其辅助响应；因此，要实现无密钥的组装态覆盖，就必须显式提供回放 override。

.agents/notes/implemented/feature/2026-07-26-todo-parallel-in-progress.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 .agents/notes/implemented/feature/2026-07-26-todo-parallel-in-progress.md
-2026-07-26-todo-parallel-in-progress.md: 2805ef894050d1b1cffe06fce59a4984d463f8d1
-2026-07-26-todo-parallel-in-progress.zh.md: 16b32daa05b10f24eacde4cec9622b2159cdef09
+2026-07-26-todo-parallel-in-progress.md: 8d047f33ab1aebd8c0de2a0a5e90e7efb1b28154
+2026-07-26-todo-parallel-in-progress.zh.md: 26b7081e7b4e3688a46a2857da91ba45328532a9

.agents/notes/implemented/feature/2026-07-26-todo-parallel-in-progress.md

@@ -37,7 +37,7 @@ The durable-log invariant deliberately does NOT follow the flag. A log written w
 
 ## The display surfaces are part of the change
 
-Lifting the cap makes a list shape reachable that no renderer had ever received, so this branch stacks on the [web todo display](2026-07-23-web-todo-display.md) rather than landing beside it: both change `tool-todo`, and the GUI is where a parallel plan becomes visible. Two web sites derived their one-line summary with `todos.find(t => t.status === 'in_progress')` — the collapsed plan-strip header and the `todo_write` row — and under the old cap that `find` was total, since at most one item could match. With several active it silently dropped every active item but the first: a four-item plan with three running tasks collapsed to the name of one, and the row read `1/4 已完成 · <one task>` while two others were in flight. The expanded list was always correct (it maps every item), which is why neither PR's tests caught it — only the collapsed header and the row lost information. The panel redesign in [#740](https://github.com/deepseek-harness/deepseek-harness/pull/740) has since replaced the collapsed header's named hint with `·`-joined per-status counts (localized, `1 completed · 2 in progress · 1 pending`, zero-count segments omitted), which reports parallel work correctly and needs no name to truncate; the row is the one site this branch still had to fix.
+Lifting the cap makes a list shape reachable that no renderer had ever received, so this branch stacks on the [web todo display](2026-07-23-web-todo-display.md) rather than landing beside it: both change `tool-todo`, and the GUI is where a parallel plan becomes visible. Two web sites derived their one-line summary with `todos.find(t => t.status === 'in_progress')` — the collapsed plan-strip header and the `todo_write` row — and under the old cap that `find` was total, since at most one item could match. With several active it silently dropped every active item but the first: a four-item plan with three running tasks collapsed to the name of one, and the row read `1/4 已完成 · <one task>` while two others were in flight. The expanded list was always correct (it maps every item), which is why neither PR's tests caught it — only the collapsed header and the row lost information. The panel redesign replaced the collapsed header's named hint with `·`-joined per-status counts (localized, `1 completed · 2 in progress · 1 pending`, zero-count segments omitted), which reports parallel work correctly and needs no name to truncate; the row is the one site this branch still had to fix.
 
 The row takes `planSummary` in `toolviews/plan-summary.ts`. It names the first active item and counts the rest, so the row reports how many tasks are running instead of implying one. Naming every active item was rejected: the row is a single line, and an unbounded join would overflow it — the count degrades predictably where a list does not. The derivation sits inside the toolviews domain rather than in `contract/`, the inter-domain face: the panel computes its own counts inline and shares nothing with the row, so a contract module would declare a sharing relationship that no longer exists.
 

.agents/notes/implemented/feature/2026-07-26-todo-parallel-in-progress.zh.md

@@ -37,7 +37,7 @@ Status: implemented
 
 ## 展示面是本次改动的一部分
 
-解除上限使一种此前任何渲染器都不曾收到的列表形状变得可达，因此本分支 stack（栈叠）在 [web todo 展示](2026-07-23-web-todo-display.md)之上，而不是与之并行落地：两者都改 `tool-todo`，而 GUI 正是并行计划变得可见的地方。web 有两处用 `todos.find(t => t.status === 'in_progress')` 推导单行摘要——折叠态的计划横条表头与 `todo_write` 工具行——在旧上限下这个 `find` 是完备的，因为最多只能有一个条目匹配。一旦有多个活跃项，它会静默丢掉除第一个之外的全部活跃条目：一个四条目、三个任务在跑的计划折叠后只显示其中一个的名字，工具行读作 `1/4 已完成 · <一个任务>`，而另外两个仍在进行。展开态的列表始终正确（它遍历每个条目），这也是两个 PR 的测试都没抓到它的原因——只有折叠表头与工具行丢失了信息。其后 [#740](https://github.com/deepseek-harness/deepseek-harness/pull/740) 的面板重做已把折叠表头的具名提示换成以 `·` 连接的各状态计数（本地化后形如 `1 已完成 · 2 进行中 · 1 待处理`，计数为零的段落省略），它能正确报告并行工作，且不需要任何可被截断的名字；工具行才是本分支仍需修的那一处。
+解除上限使一种此前任何渲染器都不曾收到的列表形状变得可达，因此本分支 stack（栈叠）在 [web todo 展示](2026-07-23-web-todo-display.md)之上，而不是与之并行落地：两者都改 `tool-todo`，而 GUI 正是并行计划变得可见的地方。web 有两处用 `todos.find(t => t.status === 'in_progress')` 推导单行摘要——折叠态的计划横条表头与 `todo_write` 工具行——在旧上限下这个 `find` 是完备的，因为最多只能有一个条目匹配。一旦有多个活跃项，它会静默丢掉除第一个之外的全部活跃条目：一个四条目、三个任务在跑的计划折叠后只显示其中一个的名字，工具行读作 `1/4 已完成 · <一个任务>`，而另外两个仍在进行。展开态的列表始终正确（它遍历每个条目），这也是两个 PR 的测试都没抓到它的原因——只有折叠表头与工具行丢失了信息。面板重做把折叠表头的具名提示换成以 `·` 连接的各状态计数（本地化后形如 `1 已完成 · 2 进行中 · 1 待处理`，计数为零的段落省略），它能正确报告并行工作，且不需要任何可被截断的名字；工具行才是本分支仍需修的那一处。
 
 工具行改用 `toolviews/plan-summary.ts` 中的 `planSummary`。它给出第一个活跃条目，并计数其余活跃项，因此工具行报告的是有多少任务在跑，而不是暗示只有一个。列出全部活跃条目被否决了：工具行是单行，无上界的拼接会溢出——在列表做不到的地方，计数能够可预测地降级。该推导放在 toolviews 域内而非 `contract/`（域间共享面）：面板自行内联计算其计数，与工具行不共享任何东西，因此放进 contract 会声明一种已不存在的共享关系。
 

.agents/notes/implemented/feature/2026-07-30-queued-manual-compaction.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 .agents/notes/implemented/feature/2026-07-30-queued-manual-compaction.md
-2026-07-30-queued-manual-compaction.md: 4b7a905712a01948146b8830dfc037185162eefc
-2026-07-30-queued-manual-compaction.zh.md: 15a42de3536f2da5304e77ce3cb282029856ba6d
+2026-07-30-queued-manual-compaction.md: 5100808ada4b7b284228113584e577d46ff91101
+2026-07-30-queued-manual-compaction.zh.md: 29c3b921d0d59527aa1d8af1c91d085b52045377

.agents/notes/implemented/feature/2026-07-30-queued-manual-compaction.md

@@ -75,7 +75,7 @@ Once a transaction has appended its start, every later failure makes one closing
 
 ### Reference implementation boundaries
 
-[PR #835](https://github.com/deepseek-harness/deepseek-harness/pull/835) was used as a reference implementation for the command, reservation, tests, and snapshot shape, but was not merged. Its process-local `WeakSet` lock and locked/unlocked method splits were considered and not adopted because the durable bracket is the single reachable lock.
+An unmerged reference implementation informed the command, reservation, tests, and snapshot shape. Its process-local `WeakSet` lock and locked/unlocked method splits were considered and not adopted because the durable bracket is the single reachable lock.
 
 That reference also carried client-side replacement-anchor machinery to preserve transcript placement. The log-ordered transcript projection already consumes compaction from event order and does not consult mutable surface positions, so those anchors were considered and not adopted.
 

.agents/notes/implemented/feature/2026-07-30-queued-manual-compaction.zh.md

@@ -75,7 +75,7 @@ DSH 有意在调用摘要器前记录 `compact/start`。缓慢或崩溃的尝试
 
 ### 参考实现边界
 
-[PR #835](https://github.com/deepseek-harness/deepseek-harness/pull/835) 用作命令、预留、测试与快照结构的参考实现，但未被合并。它的进程本地 `WeakSet` 锁与 locked／unlocked 方法拆分经过评估后未被采用，因为持久标记对是唯一可达的锁。
+一个未合并的参考实现为命令、预留、测试与快照结构提供了参考。它的进程本地 `WeakSet` 锁与 locked／unlocked 方法拆分经过评估后未被采用，因为持久标记对是唯一可达的锁。
 
 该参考实现还包含客户端侧替换锚点机制，用于保留 transcript（文本记录）位置。按日志顺序排列的 transcript 投影已经从事件顺序消费压缩，并且不会查询可变 surface 位置，因此这些锚点经过评估后未被采用。
 

.agents/notes/implemented/feature/2026-08-02-pwsh-tool-bash-parity.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 .agents/notes/implemented/feature/2026-08-02-pwsh-tool-bash-parity.md
-2026-08-02-pwsh-tool-bash-parity.md: 945d2d5243162fe8e7fb3f76cbc3bcf0b5c2fdee
-2026-08-02-pwsh-tool-bash-parity.zh.md: f537e313a0c895927c6e2319b11b98619a70d461
+2026-08-02-pwsh-tool-bash-parity.md: e35a903892d5d50a0d3ca12b23daa53f26d6aade
+2026-08-02-pwsh-tool-bash-parity.zh.md: d7a68f0cab5281b25c3321e7ffff61c358b21a0a

.agents/notes/implemented/feature/2026-08-02-pwsh-tool-bash-parity.md

@@ -14,7 +14,7 @@ The first Windows-native foundation shipped `dsh-tool-pwsh` as a deliberately mi
 
 - **Rendering adopts the bash story verbatim**: stdout, a marked `[stderr]` section, truncation notices with spill paths, `(no output)` for an empty body, and exit markers only for non-zero exits — a clean exit produces no marker. The description and the `tool:pwsh` prompt section state this precisely ("Non-zero exits are reported as `[exit code: N]` markers"), deliberately not copying the bash prompt's "every result" phrasing, which its own renderer contradicts.
 - **`run_in_background` is wired through the generic task runtime** exactly like the bash tool: preflight, owner registration, `task_output`/`task_kill` control, and the same outcome mapping. `pwsh-local`'s already-mirrored `start()` handle backs it.
-- **The `DSH_*` environment is shared, not duplicated**: `BashEnvRegistry` moved out of `dsh-tool-bash` into a new tool-independent `@deepseek-ai/dsh-bash-env` package (`ctx.bashEnv` + built-ins + the session-persistence contributor), and both shell tools inject it. Contributors apply to pwsh calls exactly as they do to bash calls, resolving the bash tool's `FIXME(bash-env-ownership)`.
+- **The `DSH_*` environment is shared, not duplicated**: `BashEnvRegistry` moved out of `dsh-tool-bash` into a new tool-independent `@deepseek-ai/dsh-bash-env` package (`ctx.bashEnv` + built-ins + the session-persistence contributor), and both shell tools inject it. Contributors apply to pwsh calls exactly as they do to bash calls; shared environment ownership therefore sits outside either model-facing shell tool.
 - **Windows reality is pinned where bash has no analog**: every command runs under a UTF-8 output preamble so the Windows PowerShell 5.1 fallback cannot garble non-ASCII output through the UTF-8-decoding collector, and the prompts teach that Windows forced termination settles as exit 1 without a signal marker.
 - **Out of scope, unchanged**: sandbox escalation (waits for a Windows-confining executor) and persistent PTY shells (backends are Linux/macOS-only; ConPTY is roadmap work). The pwsh-specific terminal card with an exit pill shipped separately in the [pwsh UI presentation matches bash](2026-08-05-pwsh-ui-bash-parity.md) decision.
 

.agents/notes/implemented/feature/2026-08-02-pwsh-tool-bash-parity.zh.md

@@ -14,7 +14,7 @@ Status: implemented
 
 - **渲染完全采用 bash 故事**：stdout、带标记的 `[stderr]` 段、带 spill 路径的截断通知、空体渲染 `(no output)`、退出 marker 仅限非零退出——干净退出不产生 marker。描述与 `tool:pwsh` prompt section 精确陈述这一点（"Non-zero exits are reported as `[exit code: N]` markers"），刻意不复制 bash prompt 中与其自身渲染矛盾的 "every result" 措辞。
 - **`run_in_background` 经通用任务运行时接线**，与 bash 工具完全一致：预检、owner 注册、`task_output`/`task_kill` 控制与相同的结果映射。其背后是 `pwsh-local` 早已镜像好的 `start()` 句柄。
-- **`DSH_*` 环境共享而非复制**：`BashEnvRegistry` 从 `dsh-tool-bash` 迁入新的工具无关包 `@deepseek-ai/dsh-bash-env`（`ctx.bashEnv` + 内置事实 + session-persistence contributor），两个 shell 工具都注入它。contributor 对 pwsh 调用与 bash 调用一视同仁，并消化了 bash 工具的 `FIXME(bash-env-ownership)`。
+- **`DSH_*` 环境共享而非复制**：`BashEnvRegistry` 从 `dsh-tool-bash` 迁入新的工具无关包 `@deepseek-ai/dsh-bash-env`（`ctx.bashEnv` + 内置事实 + session-persistence contributor），两个 shell 工具都注入它。contributor 对 pwsh 调用与 bash 调用一视同仁；因此，共享环境的所有权不属于任何一个面向模型的 shell 工具。
 - **Windows 现实在 bash 无对应处钉死**：每条命令都在 UTF-8 输出 preamble 下运行，使 Windows PowerShell 5.1 兜底无法经 UTF-8 解码的 collector 破坏非 ASCII 输出；prompt 教授 Windows 强制终止以无 signal 的 exit 1 结算。
 - **范围外，不变**：sandbox 升级（等待 Windows-confining 执行器）与持久 PTY shell（后端仅限 Linux/macOS；ConPTY 属路线图）。带退出 pill 的 pwsh 专属 terminal 卡已随 [pwsh UI 呈现与 bash 对齐](2026-08-05-pwsh-ui-bash-parity.md) 决策另行交付。
 

.agents/notes/implemented/feature/2026-08-05-pwsh-ui-bash-parity.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 .agents/notes/implemented/feature/2026-08-05-pwsh-ui-bash-parity.md
-2026-08-05-pwsh-ui-bash-parity.md: 6bbdb0e6bc69ef1af03a6a9146f83b84754cb2a6
-2026-08-05-pwsh-ui-bash-parity.zh.md: 75f3a3ddec002acaa1755c81114b0f122ab80593
+2026-08-05-pwsh-ui-bash-parity.md: 815b448b894e9c53b4c4a2076f6b94fdd316dd35
+2026-08-05-pwsh-ui-bash-parity.zh.md: 967c5a9e1409028043dc5028fdca640ddfeb1acc

.agents/notes/implemented/feature/2026-08-05-pwsh-ui-bash-parity.md

@@ -6,7 +6,7 @@ English | [中文](2026-08-05-pwsh-ui-bash-parity.zh.md)
 
 ## Problem
 
-The [pwsh tool bash parity decision](../../implemented/feature/2026-08-02-pwsh-tool-bash-parity.md) made `dsh-tool-pwsh` behaviorally interchangeable with `dsh-tool-bash` for execution, markers, and background tasks, but explicitly deferred the human-visible half: a completed pwsh foreground call presented as a generic `console`-fenced card while the bash tool's completed call presented as a terminal card with a parsed exit-status pill. The roadmap that owned this gap ([Windows defaults to pwsh](../../proposed/feature/2026-08-01-windows-pwsh-default.md)) named "pwsh TUI/GUI rendering" as stage 2 — but the TUI package was removed ([`ed30088adb`](https://github.com/deepseek-harness/deepseek-harness/commit/ed30088adb)), leaving the Web surface as the only UI the gap affects.
+The [pwsh tool bash parity decision](../../implemented/feature/2026-08-02-pwsh-tool-bash-parity.md) made `dsh-tool-pwsh` behaviorally interchangeable with `dsh-tool-bash` for execution, markers, and background tasks, but explicitly deferred the human-visible half: a completed pwsh foreground call presented as a generic `console`-fenced card while the bash tool's completed call presented as a terminal card with a parsed exit-status pill. The roadmap that owned this gap ([Windows defaults to pwsh](../../proposed/feature/2026-08-01-windows-pwsh-default.md)) named "pwsh TUI/GUI rendering" as stage 2, but the TUI package was removed, leaving the Web surface as the only UI the gap affects.
 
 ## Decision
 

.agents/notes/implemented/feature/2026-08-05-pwsh-ui-bash-parity.zh.md

@@ -6,7 +6,7 @@ Status: implemented
 
 ## Problem
 
-[pwsh 工具与 bash 对齐决策](../../implemented/feature/2026-08-02-pwsh-tool-bash-parity.md) 让 `dsh-tool-pwsh` 在执行、marker 与后台任务上行为可互换，但明确推迟了面向人类的一半：完成的 pwsh 前台调用呈现为通用 `console` 围栏卡片，而 bash 工具的完成调用呈现为带解析退出状态 pill 的 terminal 卡。拥有此缺口的路线图（[Windows 默认改用 pwsh](../../proposed/feature/2026-08-01-windows-pwsh-default.md)）把 "pwsh TUI/GUI 渲染" 列为阶段 2——但 TUI 包已被移除（[`ed30088adb`](https://github.com/deepseek-harness/deepseek-harness/commit/ed30088adb)），Web 表面成为该缺口唯一影响的 UI。
+[pwsh 工具与 bash 对齐决策](../../implemented/feature/2026-08-02-pwsh-tool-bash-parity.md) 让 `dsh-tool-pwsh` 在执行、marker 与后台任务上行为可互换，但明确推迟了面向人类的一半：完成的 pwsh 前台调用呈现为通用 `console` 围栏卡片，而 bash 工具的完成调用呈现为带解析退出状态 pill 的 terminal 卡。拥有此缺口的路线图（[Windows 默认改用 pwsh](../../proposed/feature/2026-08-01-windows-pwsh-default.md)）把 "pwsh TUI/GUI 渲染" 列为阶段 2，但 TUI 包已被移除，使 Web 表面成为该缺口唯一影响的 UI。
 
 ## Decision
 

.agents/notes/implemented/feature/2026-08-08-pi-ai-per-model-reasoning-declarations.i18n.yaml

@@ -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/2026-08-08-pi-ai-per-model-reasoning-declarations.md
+2026-08-08-pi-ai-per-model-reasoning-declarations.md: b6264feeb724e3693078fa3fc3e3fc16ed01aacb
+2026-08-08-pi-ai-per-model-reasoning-declarations.zh.md: 1b30f7e0c42974c777a535e133a47caa217e2e5e

.agents/notes/implemented/feature/2026-08-08-pi-ai-per-model-reasoning-declarations.md

@@ -0,0 +1,33 @@
+# Agent Note: Per-Model Reasoning Declarations in llm-pi-ai
+
+Status: implemented
+
+English | [中文](2026-08-08-pi-ai-per-model-reasoning-declarations.zh.md)
+
+## Problem
+
+Under the declared-provider catalog ([[2026-08-03-pi-ai-declared-provider-catalog]], which deliberately kept reasoning out of the configurable fields), a hand-declared pi-ai route's models materialized with `reasoning: false`, so `getSupportedThinkingLevels` short-circuited to `["off"]`: the composer offered no effort picker for them, and the route-level `reasoning` default — the only reasoning knob a profile had — made every request to such a model fail with `UNSUPPORTED_REASONING_EFFORT` before network I/O. The same route-level knob was also the wrong altitude for catalog routes: one provider's models disagree about which levels they accept (deepseek ships `[off, high, max]` beside catalog models with `xhigh`), so a single per-route level could not be set without breaking part of the route, which is why the Models page stopped writing it entirely (#1860) and left `settings.yaml` with no way to align efforts per model.
+
+Two adjacent gaps compounded this. pi-ai decides the reasoning *wire dialect* (`compat.thinkingFormat`, `compat.supportsReasoningEffort`) by recognizing the endpoint URL, and a private gateway's URL says nothing — a DeepSeek-dialect gateway was spoken to in the OpenAI dialect with no configuration that could correct it. And the only way to touch one catalog model was the `models` list, which *replaces* the served catalog: narrowing `gpt-5`'s levels meant restating all thirty-eight openai models or silently dropping thirty-seven.
+
+## Decision
+
+`PiAiModelProfile` gains `reasoningEfforts`: **each key is a level selectors offer, its value the spelling dispatch sends on the wire**. The declaration translates to pi-ai's `Model.reasoning` + `thinkingLevelMap` with all seven levels decided explicitly — declared levels carry their wire value, undeclared levels are pinned `null` — so the profile author never needs pi-ai's asymmetric defaulting rule (absent means "supported" for the five base levels but "unsupported" for `xhigh`/`max`). `off` is the one three-state key: left out, no Off is offered and an explicit Off request is refused (an effortless request still goes out bare, leaving the provider its default); declared valueless, Off is offered and dispatch sends nothing (the `deepseek` dialect sends `thinking: {type: "disabled"}`); declared with a value, that value goes on the wire. `false` declares a non-reasoning model; an empty declaration is refused rather than guessed at. The spelling for "disable" is `false` rather than `{}` because schemastery materializes an absent dict as `{}` — only a `z.union([z.const(false), dict])` keeps absent, disabled, and declared distinguishable, and a bare `reasoningEfforts:` (YAML null) slips through that union unvalidated, so resolution refuses it explicitly.
+
+`compat.thinkingFormat` and `compat.supportsReasoningEffort` become configurable at two levels — route (its models' default) and model (winning per field) — resolving model → route → installed catalog entry → pi-ai's URL guess. They exist only on `openai-completions` (pi-ai types them nowhere else): a model-level switch on another protocol fails resolution, a route-level default skips such models, and a route with no completions model at all is refused. The two `chat-template` formats stay withheld for want of `chatTemplateKwargs`. Both enums are pinned to pi-ai's types through `Record<UpstreamUnion, true>` drift gates, so a pi-ai upgrade that adds a format fails compilation until the new member is classified (verified against the published 0.84.1 tarball, whose `thinkingFormat` union adds `baseten` over the pinned 0.82.1).
+
+`modelOverrides` reshapes individual catalog models without replacing the served set: key = catalog model id, value = a `models` entry minus `id`, materialized by handing the override to the existing entry path so capacities, efforts, compat, and request-default semantics stay identical. Unlike Pi's own config layer, which ignores unknown ids, every override that lands nowhere is refused — beside a `models` list, on a hand-declared route, naming an unknown model, or smuggling an `id` in the value (the schema passes unknown keys through, and a smuggled id would quietly rename the model).
+
+## Alternatives considered
+
+- **Pass `reasoning` + `thinkingLevelMap` through verbatim** (pi-ai's own radius-config shape). Rejected by the user for operator confusion: the map's `null`-marks-unsupported convention plus the asymmetric absent-key rule mean the config's meaning depends on knowledge of pi-ai internals; the chosen shape makes the key set itself the offer.
+- **A bare level list** (`reasoningEfforts: [off, high]`). Cannot express wire renames, and the catalog's own maps prove renames are real: 66 of 1230 installed map entries are non-identity (`off→none`, `minimal→low`, `low→LOW`, `high→default`).
+- **`{}` as the disable spelling.** Unimplementable: schemastery materializes an absent dict as `{}`, so every model without the field would have been force-disabled.
+- **Folding this into the route-level `reasoning` knob.** That knob is a *default selection*, not a capability set; it stays, and a declared model's efforts now bound what it can select.
+
+## Consequences
+
+- The composer's effort pane works for hand-declared models with zero UI change — `resolveModelInfo` reports declared levels through the same seam catalog metadata uses (pinned by the `declared-reasoning` web scenario).
+- #1860's deferred gap — a route-level effort a model cannot take failing its requests — now has an operator remedy: align the model's `reasoningEfforts` or drop the route default.
+- There is deliberately no spelling for returning one map key or compat field to "whatever the catalog said": the declaration is the whole offer, so keeping a catalog value means restating it. The README documents this.
+- `verify-package-invariants` is untouched: the feature adds configuration resolution, no new events or mutable runtime relations.

.agents/notes/implemented/feature/2026-08-08-pi-ai-per-model-reasoning-declarations.zh.md

@@ -0,0 +1,33 @@
+# Agent Note: llm-pi-ai 的按模型推理声明
+
+Status: implemented
+
+[English](2026-08-08-pi-ai-per-model-reasoning-declarations.md) | 中文
+
+## 问题
+
+在声明式提供方 catalog（[[2026-08-03-pi-ai-declared-provider-catalog]]，它刻意把推理排除在可配置字段之外）之下，手工声明的 pi-ai 路由，其模型物化出来就带着 `reasoning: false`，于是 `getSupportedThinkingLevels` 短路成 `["off"]`：输入框不为它们提供档位选择器，而路由级的 `reasoning` 默认值——当时 profile 仅有的推理旋钮——让发往这类模型的每个请求都在网络 I/O 之前以 `UNSUPPORTED_REASONING_EFFORT` 失败。同一个路由级旋钮对 catalog 路由来说也放错了层级：同一提供方下各模型接受的档位并不一致（deepseek 自带 `[off, high, max]`，旁边就是带 `xhigh` 的 catalog 模型），单个路由级档位怎么设都会弄坏路由的一部分——这正是模型页彻底停写它的原因（#1860），而 `settings.yaml` 也因此没有了任何按模型对齐档位的办法。
+
+两个相邻的缺口让问题雪上加霜。pi-ai 靠识别端点 URL 来决定推理的*协议方言*（`compat.thinkingFormat`、`compat.supportsReasoningEffort`），而私有网关的 URL 什么也说明不了——说 DeepSeek 方言的网关只会收到 OpenAI 方言的请求，且没有任何配置能更正它。另外，想动单个 catalog 模型，唯一的手段是 `models` 列表，而它会*替换*所服务的 catalog：收窄 `gpt-5` 的档位，意味着要么重述全部三十八个 openai 模型，要么静默丢掉三十七个。
+
+## 决策
+
+`PiAiModelProfile` 新增 `reasoningEfforts`：**每个键是选择器提供的一个档位，其值是分派在协议中发送的拼写**。该声明会转换为 pi-ai 的 `Model.reasoning` + `thinkingLevelMap`，七个档位全部显式决定——已声明的档位携带自己的协议值，未声明的档位一律固定为 `null`——因此 profile 作者永远不需要了解 pi-ai 那条不对称的默认规则（键缺席对五个基础档位意味着「支持」，对 `xhigh`/`max` 却意味着「不支持」）。`off` 是唯一的三态键：不写，选择器不提供 Off，显式请求 Off 会被拒绝（不点名档位的请求仍会不带参数地发出，提供方保留自己的默认行为）；声明而不给值，则提供 Off，分派什么也不发送（`deepseek` 方言发送 `thinking: {type: "disabled"}`）；声明并给值，该值就在协议中发送。`false` 声明一个不具备推理能力的模型；空声明会被拒绝，而不是去猜。「禁用」的拼写取 `false` 而非 `{}`，因为 schemastery 会把缺席的字典物化成 `{}`——只有 `z.union([z.const(false), dict])` 才能让缺席、禁用与已声明三态保持可区分；而裸写的 `reasoningEfforts:`（YAML null）会不经校验地从该 union 溜过去，因此解析对它显式拒绝。
+
+`compat.thinkingFormat` 与 `compat.supportsReasoningEffort` 变为两级可配置——路由级（作为其模型的默认值）与模型级（逐字段胜出）——解析顺序为模型 → 路由 → 已安装 catalog 条目 → pi-ai 按 URL 得出的猜测。两者只存在于 `openai-completions` 上（pi-ai 也只在这一协议上为它们建了类型）：在其他协议的模型上设模型级开关会使解析失败，路由级默认值会跳过这类模型，而完全没有 completions 模型的路由则被拒绝。两个 `chat-template` 格式因缺 `chatTemplateKwargs` 而继续保持不开放。两个枚举都经 `Record<UpstreamUnion, true>` 漂移门禁钉在 pi-ai 的类型上，因此新增格式的 pi-ai 升级会编译失败，直到新成员被归类（对照已发布的 0.84.1 tarball 验证过：其 `thinkingFormat` 联合类型相对钉住的 0.82.1 新增了 `baseten`）。
+
+`modelOverrides` 就地重塑单个 catalog 模型而不替换所服务的集合：键 = catalog 模型 id，值 = 去掉 `id` 的 `models` 条目，物化时把覆盖交给既有的条目路径，因此容量、档位、compat 与请求默认值语义完全一致。与忽略未知 id 的 Pi 自有配置层不同，凡是落不到任何地方的覆盖都会被拒绝——与 `models` 列表并存、写在手工声明的路由上、点名未知模型，或在值里夹带 `id`（schema 会放行未知键，被夹带的 id 会悄悄把模型改名）。
+
+## 曾考虑的替代方案
+
+- **把 `reasoning` + `thinkingLevelMap` 原样透传**（pi-ai 自家 radius 配置的形状）。用户以运维人员困惑为由否决：map 用 `null` 标记「不支持」的约定，加上不对称的键缺席规则，意味着这份配置的含义取决于对 pi-ai 内部机制的了解；选定的形状则让键集合本身就是对外提供的全部。
+- **裸档位列表**（`reasoningEfforts: [off, high]`）。表达不了协议侧改名，而 catalog 自己的 map 证明改名真实存在：1230 条已安装 map 条目里有 66 条不是恒等映射（`off→none`、`minimal→low`、`low→LOW`、`high→default`）。
+- **用 `{}` 作为禁用拼写。** 无法实现：schemastery 会把缺席的字典物化成 `{}`，于是每个没写该字段的模型都会被强制禁用。
+- **把这件事并进路由级的 `reasoning` 旋钮。** 那个旋钮是*默认选择*，不是能力集合；它保留下来，而已声明模型的档位如今约束着它能选什么。
+
+## 后果
+
+- 输入框的档位面板对手工声明的模型直接可用，UI 零改动——`resolveModelInfo` 经 catalog 元数据所走的同一 seam 报告已声明档位（由 `declared-reasoning` web 场景钉住）。
+- #1860 暂缓的缺口——模型接不住的路由级档位会让发往它的请求失败——如今有了运维侧补救：对齐该模型的 `reasoningEfforts`，或去掉路由默认值。
+- 刻意不提供任何把单个 map 键或 compat 字段交还给「catalog 原本怎么说」的拼写：这份声明就是对外提供的全部，要保留某个 catalog 值就得重述它。README 记载了这一点。
+- `verify-package-invariants` 原封未动：该功能新增的是配置解析，没有新事件，也没有可变的运行时关系。

.agents/notes/implemented/feature/2026-08-08-user-explicit-skill-invocation.i18n.yaml

@@ -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/2026-08-08-user-explicit-skill-invocation.md
+2026-08-08-user-explicit-skill-invocation.md: d925938279923282170dc99934f4fa44d8ecf2b4
+2026-08-08-user-explicit-skill-invocation.zh.md: 64e23be0b42519fb9681adefcd0f05074d3aa35e

.agents/notes/implemented/feature/2026-08-08-user-explicit-skill-invocation.md

@@ -0,0 +1,36 @@
+# Agent Note: User-explicit skill invocation at the pre-step gesture boundary
+
+Status: implemented
+
+English | [中文](2026-08-08-user-explicit-skill-invocation.zh.md)
+
+## Problem
+
+A `disable-model-invocation: true` skill is user-only by design: it never enters the model-facing catalog and the `skill` tool refuses to load it. Its only legitimate entry point is an explicit user gesture — yet the web client had none. `skill.list` filtered to the model-and-user intersection (hiding user-only skills from the menu), an entered `/name` line rode into the default prompt sink as plain text, and the model it reached was forbidden to load the skill — so it degraded to `read`-ing the SKILL.md file or ignoring the gesture (issue #1470). Even for ordinary skills, the decision-21 plain-text reference made user invocation a collaboration cue the model could ignore, not a guarantee.
+
+## Decision
+
+User-explicit invocation is a host-side pre-step injection, uniform for every user-invocable skill and every front end:
+
+- `dsh-tool-skill` registers a second `agent/pre-step` listener (beside its catalog listener, the same seam `workspace-instructions` and the runtime-context snapshot ride): it scans the step's claimed messages for whitespace-bounded `/name` tokens — anywhere in the text, the same word-boundary shape the transcript chip decoration uses — collects first-seen-deduplicated names, loads each through `ctx.skills.get`, checks `isUserInvocable` on the loaded definition (the single lookup that produces what is injected), renders it with the shared `renderSkillContent`, and appends the injections after every other injection of the step: background first (workspace rules, runtime policy, catalog), the material the model must act on last, closest to its answer. Registration order pins the placement — the gesture listener registers before the catalog listener, so the waterfall hands it the catalog-bearing list to extend.
+- Precision is closed-set matching, exactly like slash commands: `/goal` resolves against the command registry, `/name` against the workspace's user-invocable skill directory; a miss stays ordinary prose, so nothing is ever guessed. Only `source.kind === 'user'` messages are scanned — external text cannot forge a gesture. Paths (`/usr/bin`), fractions (`5/8`), and prefixed tokens (`foo/name`) all break the boundary.
+- The client stays decision 21: a menu pick lands the literal `/name ` and the prompt ships it verbatim; ui-skill implements no adjudication hooks and no reference codec. `skill.list` (now the domain's only RPC) serves every user-invocable skill with `modelInvocable` so menus mark user-only entries. A name shared with a host command resolves to the command — adjudication claims the line client-side before it becomes a prompt.
+- The injection is a `user`-role message carrying the `skill-invocation` source (`{ name, form: 'instructions' }`), so `user/message` logging, the context-injection transcript row (labelled with the skill name), and replay all come free; `renderSkillContent` lives in the `dsh-skill` seam, shared verbatim with the `skill` tool result, and the catalog's closing sentence tells the model to follow an injected block instead of re-loading it.
+
+Peer-product survey (Pi, OpenCode, Claude Code, Kimi Code, Codex, DeepSeek-Reasonix — local checkouts) was unanimous that user-explicit triggering is programmatic injection with zero model participation; the final shape is closest to Codex's core-side `$name` mention scanning, which likewise frees every front end from implementing recognition.
+
+## Alternatives considered
+
+- **`skill.invoke` RPC (host injects, client claims)** — implemented first, in two iterations: a single mixed message (user text folded into the body), then a gesture prompt plus injection delivered through inbox primitives. Rejected after real-session testing: the mixed message polluted the injection with user prose; the two-message form depended on wake-ordering subtleties (`followup` claims the whole next-turn queue synchronously inside the first waking call, stranding any later message in the next turn — reproduced live), and the dedicated RPC duplicated a path `session.prompt` already provides while leaving TUI/ACP to reimplement recognition. The pre-step seam removes the RPC, the claim machinery, and the ordering hazard outright.
+- **`agent.inject()` from the RPC handler** — the inject queue (`next-step`, wake-free) is claimed ahead of the next-turn prompt, putting the injection above the gesture in the log; and pairing it with a waking `followup` reintroduces the same ordering coupling. The pre-step listener injects inside the step assembly, where ordering is explicit.
+- **A host `/skill <name>` command** (command registry, plan-mode precedent) — two-token UX, no name completion, and user-only skills stay undiscoverable in the menu; the per-cwd skill catalog also fits the static command registry poorly. Rejected.
+- **Client-side expansion** (fetch body, splice into the prompt) — authorization becomes bypassable client courtesy, the log loses the invocation semantics, and Codex deleted its equivalent mechanism (custom prompts) in favor of core injection. Rejected.
+- **Structured reference payload on the prompt wire** (Codex's `UserInput::Skill` analogue: the client ships `{skills: [...]}` beside the text and the boundary prefers it over scanning) — considered and deferred: the existing slash-command system is itself line-text on the wire, and closed-set directory matching already removes the guesswork; recorded as a ledger item should gesture precision ever need client intent.
+- **Per-injection preamble line** (Kimi's `User activated the skill …`) — dropped in favor of the one-time catalog sentence: same context, paid once, and the injected block stays byte-identical with the tool result.
+
+## Consequences
+
+- Decision 21's plain-text reference is now the whole client story: the draft carries plain text, chip visuals derive from the lexicon, and the sent text is judged by the host boundary — a hand-typed gesture, a menu pick, and a TUI prompt are indistinguishable and equally deterministic.
+- Every user-invocable skill invocation costs its full rendered body unconditionally — the price of determinism the peer survey showed everyone pays. Mentioning a known skill name mid-sentence loads it; that is the Codex mention semantic, accepted deliberately.
+- The `skill-invocation` source rides `user/message`, so Model-visible ⟺ logged holds with no new event type, and replay/UI read metadata rather than text markers.
+- Accepted residual of dropping the per-injection preamble: the no-reload framing rides only the catalog, and a workspace whose skills are all user-only never publishes a first catalog — an injection can arrive with no framing at all, and the model may redundantly try the `skill` tool once (the replacement catalog's empty arm carries the sentence; the never-published case does not). Publishing a catalog for framing alone was judged worse than that one recoverable error.

.agents/notes/implemented/feature/2026-08-08-user-explicit-skill-invocation.zh.md

@@ -0,0 +1,36 @@
+# Agent Note: pre-step 手势边界上的用户显式 skill 调用
+
+Status: implemented
+
+[English](2026-08-08-user-explicit-skill-invocation.md) | 中文
+
+## 问题
+
+`disable-model-invocation: true` 的 skill（技能）在设计上就是仅限用户的：它绝不进入面向模型的目录，`skill` 工具也拒绝加载它。它唯一正当的入口是一次显式的用户手势——而 web 客户端此前没有这个入口。`skill.list` 过滤到模型与用户的交集（把仅限用户的 skill 挡在菜单之外），回车提交的 `/name` 一行以纯文本落入默认提示词 sink，而这行文本到达的模型又被禁止加载该 skill——于是退化为模型去 `read` 那份 SKILL.md 文件，或者干脆无视这次手势（issue #1470）。即使对普通 skill，决策 21 的纯文本引用也让用户调用只是模型可以忽略的协作线索，而不是保证。
+
+## 决策
+
+用户显式调用是一次宿主侧的 pre-step 注入，对每一个用户可调用的 skill 和每一种前端一致：
+
+- `dsh-tool-skill` 注册第二个 `agent/pre-step` 监听器（与其目录监听器并列，也是 `workspace-instructions` 与运行时上下文快照搭乘的同一 seam）：它在该步骤已认领的消息中扫描以空白为界的 `/name` token——文本中任意位置均可，与 transcript（文本记录）chip 装饰所用的词边界形状相同——收集按首见去重的名称，逐个经 `ctx.skills.get` 加载，在已加载定义上检查 `isUserInvocable`（产生注入内容的正是这同一次查找），用共享的 `renderSkillContent` 渲染，并把注入追加在该步骤所有其他注入之后：背景在前（工作区规则、运行时策略、目录），模型必须着手处理的材料在最后、最贴近它的回答。注册顺序钉住了这一位置——手势监听器先于目录监听器注册，因此 waterfall 会把携带目录的列表交给它来扩展。
+- 精确性来自封闭集合匹配，与斜杠命令完全一致：`/goal` 对照命令注册表解析，`/name` 对照工作区的用户可调用 skill 目录解析；未命中即保持为普通行文，因此绝不猜测。只扫描 `source.kind === 'user'` 的消息——外部文本无法伪造手势。路径（`/usr/bin`）、分数（`5/8`）与带前缀的 token（`foo/name`）都会破坏该边界。
+- 客户端停留在决策 21：菜单 pick 落下字面文本 `/name `，提示词将其原样发出；ui-skill 不实现任何裁决钩子，也没有引用 codec。`skill.list`（现在是该领域唯一的 RPC）提供每一个用户可调用的 skill 并携带 `modelInvocable`，供菜单标出仅限用户的条目。与宿主命令同名的名称解析为命令——裁决在客户端把该行认领走，它尚未成为提示词。
+- 注入是一条携带 `skill-invocation` 来源（`{ name, form: 'instructions' }`）的 `user` 角色消息，因此 `user/message` 落账、上下文注入的 transcript 行（以 skill 名称标注）与回放全部免费获得；`renderSkillContent` 位于 `dsh-skill` seam，与 `skill` 工具结果逐字共享，目录的结尾一句会告诉模型遵循注入块而不是重新加载。
+
+同类产品调研（Pi、OpenCode、Claude Code、Kimi Code、Codex、DeepSeek-Reasonix——本地检出）一致表明：用户显式触发都是模型零参与的程序化注入；最终形态最接近 Codex 核心侧的 `$name` mention 扫描——它同样让每一种前端免于自行实现识别。
+
+## 考虑过的替代方案
+
+- **`skill.invoke` RPC（宿主注入、客户端认领）**——最先实现，共两轮迭代：先是单条混合消息（用户文本折进正文），后是经 inbox 原语投递的手势提示词加注入两条消息。经真实会话测试后否决：混合消息让用户行文污染了注入；两条消息的形态依赖唤醒顺序的微妙之处（`followup` 在第一个唤醒调用内同步认领整个 next-turn 队列，把之后的消息滞留到下一轮次——已实际复现），而专设 RPC 复制了 `session.prompt` 已提供的路径，还让 TUI/ACP 不得不各自重新实现识别。pre-step seam 把 RPC、认领机制与顺序隐患一并干净移除。
+- **从 RPC 处理器调用 `agent.inject()`**——inject 队列（`next-step`，不唤醒）会在 next-turn 提示词之前被认领，使注入在日志中排到手势之上；而与会唤醒的 `followup` 搭配又会重新引入同样的顺序耦合。pre-step 监听器在步骤组装内部注入，那里的顺序是显式的。
+- **宿主 `/skill <name>` 命令**（命令注册表，plan 模式先例）——两 token 的 UX、没有名称补全、仅限用户的 skill 在菜单里仍不可发现；按 cwd 的 skill 目录也与静态命令注册表格格不入。否决。
+- **客户端展开**（拉取正文、拼进提示词）——授权沦为可被绕过的客户端善意，日志失去调用语义，而且 Codex 已删除其等价机制（custom prompts）转向核心注入。否决。
+- **提示词协议上的结构化引用载荷**（Codex `UserInput::Skill` 的类似物：客户端在文本旁附带 `{skills: [...]}`，边界优先采用它而不是扫描）——考虑过并暂缓：现有斜杠命令体系在协议上本身就是行文本，封闭集合的目录匹配已经消除了猜测；已记为台账事项，以备手势精确性某天需要客户端意图。
+- **每次注入一条前导语**（Kimi 的 `User activated the skill …`）——弃用，改为一次性的目录句子：同样的上下文、只支付一次，且注入块与工具结果保持逐字节一致。
+
+## 后果
+
+- 决策 21 的纯文本引用如今就是客户端的全部故事：草稿承载纯文本，chip 视觉由 lexicon 派生，发出的文本由宿主边界评判——手动键入的手势、菜单 pick 与 TUI 提示词无从区分，也同等确定。
+- 每一次用户可调用 skill 的调用都无条件付出其完整渲染正文的成本——这是确定性的代价，同类调研表明所有产品都在支付。在句子中间提到一个已知 skill 名称也会加载它；这就是 Codex 的 mention 语义，属于有意接受。
+- `skill-invocation` 来源搭乘 `user/message`，因此「模型可见 ⟺ 已记录」在不新增事件类型的情况下继续成立，回放与 UI 读取的是元数据而非文本标记。
+- 放弃逐次注入前导语后被接受的残余：no-reload framing 只搭乘目录，而 skill 全部为仅用户的工作区永远不会发布首个目录——注入可能在完全没有 framing 的情况下到达，模型可能多余地调用一次 `skill` 工具（替换目录的空臂携带该句；从未发布的情形没有）。仅为 framing 而发布目录被判定比这一次可恢复的错误更糟。

.agents/notes/implemented/process/2026-07-13-documentation-site-projection.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 .agents/notes/implemented/process/2026-07-13-documentation-site-projection.md
-2026-07-13-documentation-site-projection.md: f19d9b309aa22821a75086dc07ee302097631ba0
-2026-07-13-documentation-site-projection.zh.md: cc5e94e709f0639fd35ad81165b199cc5c9effc0
+2026-07-13-documentation-site-projection.md: d9af915754fa6a1df51a27d18d412597472aaa73
+2026-07-13-documentation-site-projection.zh.md: 5b6de4b3425b6d20335a0f8055468ced414d4034

.agents/notes/implemented/process/2026-07-13-documentation-site-projection.md

@@ -18,7 +18,9 @@ Canonical Markdown remains in the repository tier that owns it. Product-facing g
 
 Locale home projections retain only the canonical YAML frontmatter. The repository-facing body can keep its H1 and bilingual source links, while the VitePress home theme owns the rendered hero and features and the site navigation owns locale switching.
 
-The projector parses Markdown links without reserializing the document. A link to another published source becomes a site-relative route; a link to an unpublished repository file becomes a GitHub source link; a repository image is copied into the generated tree and referenced from there ([why](2026-08-06-doc-site-carries-its-images.md)). Missing relative targets fail projection. Unit tests pin these transformations, and `docs:check` runs the projector tests plus a production VitePress build as part of `doc-sync` and the parallel documentation gates.
+The projector parses Markdown links without reserializing the document. A link to another published source becomes a site-relative route; a link to an unpublished repository file becomes a source link under the public `deepseek-ai/deepseek-harness-sdk` home; a repository image is copied into the generated tree and referenced from there ([why](2026-08-06-doc-site-carries-its-images.md)). Missing relative targets fail projection. Unit tests pin these transformations, and `docs:check` runs the projector tests plus a production VitePress build as part of `doc-sync` and the parallel documentation gates.
+
+`verify-public-repository-links` rejects internal repository remotes from tracked files. Public source links use the public home, while work tracking stays in repository metadata and source carries a TODO only when the local boundary matters to maintainers.
 
 `website/AGENTS.md` is the only maintained Markdown file in the website subtree. The projector test enumerates tracked and unignored files and rejects any other website Markdown, so site-specific locale, route, API, or generated source copies cannot bypass the publication manifest.
 

.agents/notes/implemented/process/2026-07-13-documentation-site-projection.zh.md

@@ -18,7 +18,9 @@ Status: implemented
 
 各 locale 的首页投影只保留权威 YAML frontmatter。面向仓库的正文可以保留其 H1 和双语源文件链接，而 VitePress 首页主题负责渲染 hero 与功能区，网站导航负责切换 locale。
 
-投影器解析 Markdown 链接，但不会重新序列化文档。指向另一个已发布源文件的链接会变成站内相对路由；指向未发布仓库文件的链接会变成 GitHub 源文件链接；仓库图片会被拷贝进生成树并从那里引用（[原因](2026-08-06-doc-site-carries-its-images.md)）。相对目标不存在时，投影会失败。单元测试会锁定这些转换行为，`docs:check` 则运行投影器测试和 VitePress 生产构建，并将二者纳入 `doc-sync` 和并行文档门禁。
+投影器解析 Markdown 链接，但不会重新序列化文档。指向另一个已发布源文件的链接会变成站内相对路由；指向未发布仓库文件的链接会变成公开 `deepseek-ai/deepseek-harness-sdk` 主页下的源文件链接；仓库图片会被拷贝进生成树并从那里引用（[原因](2026-08-06-doc-site-carries-its-images.md)）。相对目标不存在时，投影会失败。单元测试会锁定这些转换行为，`docs:check` 则运行投影器测试和 VitePress 生产构建，并将二者纳入 `doc-sync` 和并行文档门禁。
+
+`verify-public-repository-links` 会拒绝已跟踪文件中的内部仓库远程链接。公开源文件链接使用公开主页，而工作跟踪留在仓库元数据中；只有本地边界对维护者有意义时，源文件才保留 TODO。
 
 `website/AGENTS.md` 是网站子树中唯一维护的 Markdown 文件。投影器测试会枚举所有已跟踪文件和未被忽略的未跟踪文件，并拒绝网站中的任何其他 Markdown，因此网站专用的 locale、路由、API 或生成源文件副本无法绕过发布 manifest。
 

.agents/notes/implemented/process/2026-07-22-evidence-based-larger-hosted-runners.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 .agents/notes/implemented/process/2026-07-22-evidence-based-larger-hosted-runners.md
-2026-07-22-evidence-based-larger-hosted-runners.md: d46b8291ec05e997728da76354354f9e36bd2fb4
-2026-07-22-evidence-based-larger-hosted-runners.zh.md: f43712859d8351ffff45c7b6d5eb2b65015ee4c3
+2026-07-22-evidence-based-larger-hosted-runners.md: 53cc86efce9061c8f9836a17cb35ebb128085b7a
+2026-07-22-evidence-based-larger-hosted-runners.zh.md: 0484548c76cb7eed11dc4235ef024a700499ef6a

.agents/notes/implemented/process/2026-07-22-evidence-based-larger-hosted-runners.md

@@ -26,7 +26,7 @@ The artifact boundary remains explicit. `scripts/publint-all.ts` calls publint's
 
 Windows shares one 32-core setup across the blocking build and production site plus observational built-artifact contracts. Linux owns the duplicate lint, coverage, and snapshot inventories because running those observational copies on Windows extends the paid critical path without adding a blocking platform claim.
 
-An [exact-head all-size benchmark](https://github.com/deepseek-harness/deepseek-harness/actions/runs/29908491351) ran the complete unsharded primary Node aggregate on every Linux pool before the eager-build correction:
+An exact-head all-size benchmark ran the complete unsharded primary Node aggregate on every Linux pool before the eager-build correction:
 
 | Complete Linux primary | 4 cores | 8 cores | 16 cores | 32 cores | 64 cores | 96 cores |
 |---|---:|---:|---:|---:|---:|---:|
@@ -40,13 +40,13 @@ The same benchmark measured the required Windows build surfaces across every pro
 |---|---:|---:|---:|---:|---:|---:|
 | Active time | 152 s | 104 s | 104 s | 92 s | 103 s | 110 s |
 
-Repository work gains little above 16 Windows cores, but the 32-core pool can start the complete outer inventory together. A [retargeted production validation](https://github.com/deepseek-harness/deepseek-harness/actions/runs/29907581119/attempts/2) completed the full one-box Windows inventory in 173 seconds, including coverage and snapshot replay, so Windows remains consolidated.
+Repository work gains little above 16 Windows cores, but the 32-core pool can start the complete outer inventory together. A retargeted production validation completed the full one-box Windows inventory in 173 seconds, including coverage and snapshot replay, so Windows remains consolidated.
 
-The larger client package graph makes cache mechanics and scheduler pressure part of the measured workload. In [one exact-head candidate run](https://github.com/deepseek-harness/deepseek-harness/actions/runs/29912577681), Linux spent 39 seconds in repository gates but 69 seconds in the complete job, while Windows spent 117 seconds in repository gates and 228 seconds in the complete job. The Windows pnpm cache downloaded its 154 MB archive in about two seconds but spent 27 seconds extracting it, followed by a 23-second install and a 14-second post-job save. A [cacheless all-size trace](https://github.com/deepseek-harness/deepseek-harness/actions/runs/29913033155) completed the same 32-core Windows install in 27 seconds. A future larger-runner rollout therefore needs complete-job measurements rather than gate-only timing.
+The larger client package graph makes cache mechanics and scheduler pressure part of the measured workload. In one exact-head candidate run, Linux spent 39 seconds in repository gates but 69 seconds in the complete job, while Windows spent 117 seconds in repository gates and 228 seconds in the complete job. The Windows pnpm cache downloaded its 154 MB archive in about two seconds but spent 27 seconds extracting it, followed by a 23-second install and a 14-second post-job save. A cacheless all-size trace completed the same 32-core Windows install in 27 seconds. A future larger-runner rollout therefore needs complete-job measurements rather than gate-only timing.
 
 Host setup remains part of any comparison. A standard Node 26 job once spent 36 of its 67 seconds in `Set up job`, while `actions/setup-node` spent 46.56 seconds printing cached Windows environment details after finding Node in the hosted toolcache. A Linux candidate also spent 18 seconds registering a 50 KB Bubblewrap package because the hosted image scanned 202,507 package-database files. [`scripts/prepare-ci-bubblewrap.sh`](../../../../scripts/prepare-ci-bubblewrap.sh) instead verifies and extracts the pinned payload into the ephemeral runner directory, runs a functional confinement probe, and overlaps that preparation with dependency installation.
 
-Inner and outer worker limits are separate controls. An [exact-head 32-worker ESLint experiment](https://github.com/deepseek-harness/deepseek-harness/actions/runs/29918329463) slowed lint to 52.28 seconds and coverage to 42.71 seconds, where an adapter idle-timeout test failed. A later 8-gate trace reduced coverage to 35.17 seconds but delayed the production-site build until the aggregate reached 41.06 seconds. Core count therefore does not justify copying an equally large worker limit.
+Inner and outer worker limits are separate controls. An exact-head 32-worker ESLint experiment slowed lint to 52.28 seconds and coverage to 42.71 seconds, where an adapter idle-timeout test failed. A later 8-gate trace reduced coverage to 35.17 seconds but delayed the production-site build until the aggregate reached 41.06 seconds. Core count therefore does not justify copying an equally large worker limit.
 
 The process-bound coverage project contains exactly five suite files. Thirty-two forks crashed Node 24's CJS lexer twice, and a later 16-fork run reproduced the worker loss and invalid coverage result. The single Vitest invocation therefore uses threads for the broad inventory and reserves forks for suites that exercise process-global state, `process` APIs, or timing-sensitive process I/O. That narrow fork inventory includes the local bash process-plumbing suite and the pi-ai adapter suite because aggregate contention changed timing observations in both. These failures make deterministic coverage, not advertised cores, the upper bound on worker selection.
 

.agents/notes/implemented/process/2026-07-22-evidence-based-larger-hosted-runners.zh.md

@@ -26,7 +26,7 @@ Linux 主流程使用 3 个相互独立的 32 核作业。覆盖率单独运行
 
 Windows 以一次 32 核环境设置同时承载阻塞性构建、生产网站和观测性的构建产物契约。重复的 lint、覆盖率和快照清单由 Linux 承担，因为在 Windows 上运行这些观测性副本会延长付费关键路径，却不会新增任何阻塞性平台契约。
 
-一次[分支头精确的全规格基准测试](https://github.com/deepseek-harness/deepseek-harness/actions/runs/29908491351)在修正构建尽早启动逻辑前，对每种 Linux 池都运行了完整且未分片的主 Node 聚合流程：
+一次分支头精确的全规格基准测试在修正构建尽早启动逻辑前，对每种 Linux 池都运行了完整且未分片的主 Node 聚合流程：
 
 | Linux 完整主流程 | 4 核 | 8 核 | 16 核 | 32 核 | 64 核 | 96 核 |
 |---|---:|---:|---:|---:|---:|---:|
@@ -40,13 +40,13 @@ Windows 以一次 32 核环境设置同时承载阻塞性构建、生产网站
 |---|---:|---:|---:|---:|---:|---:|
 | 活动耗时 | 152 秒 | 104 秒 | 104 秒 | 92 秒 | 103 秒 | 110 秒 |
 
-Windows 仓库工作在超过 16 核后收益很小，但 32 核池可以让完整的外层清单同时启动。一次[重新定向的生产验证](https://github.com/deepseek-harness/deepseek-harness/actions/runs/29907581119/attempts/2)在 173 秒内完成了单机 Windows 完整清单，其中包括覆盖率和快照回放，因此 Windows 继续采用合并执行方式。
+Windows 仓库工作在超过 16 核后收益很小，但 32 核池可以让完整的外层清单同时启动。一次重新定向的生产验证在 173 秒内完成了单机 Windows 完整清单，其中包括覆盖率和快照回放，因此 Windows 继续采用合并执行方式。
 
-客户端包依赖图增大后，缓存机制和调度器压力也成为实测工作负载的一部分。在[一次分支头精确的候选运行](https://github.com/deepseek-harness/deepseek-harness/actions/runs/29912577681)中，Linux 的仓库门禁耗时 39 秒，完整作业耗时 69 秒；Windows 的仓库门禁耗时 117 秒，完整作业耗时 228 秒。Windows pnpm 缓存的 154 MB 归档下载耗时约 2 秒，但解压耗时 27 秒，随后安装耗时 23 秒，作业结束后的保存又耗时 14 秒。一次[无缓存的全规格运行轨迹](https://github.com/deepseek-harness/deepseek-harness/actions/runs/29913033155)在 27 秒内完成了同一台 32 核 Windows 运行器上的安装。因此，未来若要启用大型运行器，需要测量完整作业，而不能只测门禁耗时。
+客户端包依赖图增大后，缓存机制和调度器压力也成为实测工作负载的一部分。在一次分支头精确的候选运行中，Linux 的仓库门禁耗时 39 秒，完整作业耗时 69 秒；Windows 的仓库门禁耗时 117 秒，完整作业耗时 228 秒。Windows pnpm 缓存的 154 MB 归档下载耗时约 2 秒，但解压耗时 27 秒，随后安装耗时 23 秒，作业结束后的保存又耗时 14 秒。一次无缓存的全规格运行轨迹在 27 秒内完成了同一台 32 核 Windows 运行器上的安装。因此，未来若要启用大型运行器，需要测量完整作业，而不能只测门禁耗时。
 
 任何比较都必须计入主机设置。一个标准 Node 26 作业曾在总共 67 秒的耗时中，把 36 秒用在 `Set up job` 上；`actions/setup-node` 从托管 toolcache 找到 Node 后，仍花费 46.56 秒输出缓存的 Windows 环境详情。一个 Linux 候选作业还在注册 50 KB 的 Bubblewrap 包时耗时 18 秒，因为托管映像扫描了 202,507 个包数据库文件。[`scripts/prepare-ci-bubblewrap.sh`](../../../../scripts/prepare-ci-bubblewrap.sh) 改为验证固定版本的 payload 并将其解压到临时运行器目录，执行功能性隔离探针，并让这项准备工作与依赖安装重叠执行。
 
-内层与外层工作线程上限是相互独立的控制机制。一次[分支头精确、使用 32 个工作线程的 ESLint 实验](https://github.com/deepseek-harness/deepseek-harness/actions/runs/29918329463)使 lint 耗时增至 52.28 秒、覆盖率耗时增至 42.71 秒；同一次运行中，一项适配器空闲超时测试失败。后来一次同时运行 8 项门禁的运行轨迹将覆盖率耗时降至 35.17 秒，但生产网站构建被延后，直到聚合流程耗时达到 41.06 秒时才完成。因此，不能仅凭核心数照搬同等规模的工作线程上限。
+内层与外层工作线程上限是相互独立的控制机制。一次分支头精确、使用 32 个工作线程的 ESLint 实验使 lint 耗时增至 52.28 秒、覆盖率耗时增至 42.71 秒；同一次运行中，一项适配器空闲超时测试失败。后来一次同时运行 8 项门禁的运行轨迹将覆盖率耗时降至 35.17 秒，但生产网站构建被延后，直到聚合流程耗时达到 41.06 秒时才完成。因此，不能仅凭核心数照搬同等规模的工作线程上限。
 
 进程约束的覆盖率项目恰好包含 5 个套件文件。32 个 fork 曾两次导致 Node 24 的 CJS 词法分析器崩溃，后来一次使用 16 个 fork 的运行又复现了工作进程丢失和无效的覆盖率结果。因此，单次 Vitest 调用会对大范围测试清单使用线程，只为涉及进程全局状态、`process` API 或对时间敏感的进程 I/O 的套件保留 fork。这份有限的 fork 清单包括本地 bash 进程通路套件和 pi-ai 适配器套件，因为聚合争用改变了二者的时序观测结果。这些故障表明，选择工作线程数量时，上限取决于能否得到确定的覆盖率结果，而非标称核心数。
 

.agents/notes/implemented/process/2026-07-26-briefed-minimal-translation-updates.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 .agents/notes/implemented/process/2026-07-26-briefed-minimal-translation-updates.md
-2026-07-26-briefed-minimal-translation-updates.md: a47251376771165d0eb229aaa0fb7f63589d7d77
-2026-07-26-briefed-minimal-translation-updates.zh.md: c3c1b4e845b5a45acde55884b883b1dd1e570d77
+2026-07-26-briefed-minimal-translation-updates.md: afd990b7b63adfd0e66a4726975b678d044e7cad
+2026-07-26-briefed-minimal-translation-updates.zh.md: dcdb253746041a7928ab7a544ae427293bc3f2de

.agents/notes/implemented/process/2026-07-26-briefed-minimal-translation-updates.md

@@ -12,7 +12,7 @@ The [bilingual pairing contract](2026-07-02-bilingual-docs-and-pairing-gate.md)
 
 Pair updates run on a generated briefing instead of the guidance corpus; only new pairs still run the whole-document workflow, which is unchanged.
 
-- **`pnpm run gen-translation-brief [--apply] [pair...]`** ([scripts/gen-translation-brief.ts](../../../../scripts/gen-translation-brief.ts), assembly in [scripts/translation-brief.ts](../../../../scripts/translation-brief.ts)) prints, per out-of-sync pair, the authored side's diff from its recorded last-confirmed blob to the working tree plus the change mapped at the narrowest safely aligned granularity, deterministically widening on mapping failure: a change confined to the pair's byte-identical code fences is computed outright (`--apply` splices it into the counterpart and validates the result against the pairing gate's structural signature before writing); otherwise changed Markdown units (headings, paragraphs, table rows, list items, code fences, block quotes, HTML blocks, thematic breaks, link definitions — matched by container-scoped kind sequences) each carry their last-confirmed source, current source, and current counterpart text with line numbers; units that do not align fall back to depth-matched heading sections; and when sections do not align either, or both sides drifted, the briefing says so and withholds the mapping instead of guessing. Terminology rows are matched against the changed spans only (word-boundary English matching with plural inflections), and for Chinese targets the briefing tracks each relevant term's document-wide first occurrence — when an edit moves it, the vacated and receiving spans join the briefing with an explanatory note, since the 首次出现 annotation must move with it. The unit mapping, code splice, and first-occurrence mechanics adopt the planner design from the [incremental prompt-pipeline work](https://github.com/deepseek-harness/deepseek-harness/pull/684), whose provider-backed bake-off independently validated the same scope ladder for the automated pipeline. The briefing is the translator's whole working set; the full sources of truth remain the escalation path for decisions the briefing cannot answer.
+- **`pnpm run gen-translation-brief [--apply] [pair...]`** ([scripts/gen-translation-brief.ts](../../../../scripts/gen-translation-brief.ts), assembly in [scripts/translation-brief.ts](../../../../scripts/translation-brief.ts)) prints, per out-of-sync pair, the authored side's diff from its recorded last-confirmed blob to the working tree plus the change mapped at the narrowest safely aligned granularity, deterministically widening on mapping failure: a change confined to the pair's byte-identical code fences is computed outright (`--apply` splices it into the counterpart and validates the result against the pairing gate's structural signature before writing); otherwise changed Markdown units (headings, paragraphs, table rows, list items, code fences, block quotes, HTML blocks, thematic breaks, link definitions — matched by container-scoped kind sequences) each carry their last-confirmed source, current source, and current counterpart text with line numbers; units that do not align fall back to depth-matched heading sections; and when sections do not align either, or both sides drifted, the briefing says so and withholds the mapping instead of guessing. Terminology rows are matched against the changed spans only (word-boundary English matching with plural inflections), and for Chinese targets the briefing tracks each relevant term's document-wide first occurrence — when an edit moves it, the vacated and receiving spans join the briefing with an explanatory note, since the 首次出现 annotation must move with it. The unit mapping, code splice, and first-occurrence mechanics adopt the planner design from the incremental prompt-pipeline work; its provider-backed bake-off independently validated the same scope ladder for the automated pipeline. The briefing is the translator's whole working set; the full sources of truth remain the escalation path for decisions the briefing cannot answer.
 - **The update path in [dsh-translate-docs](../../../skills/dsh-translate-docs/SKILL.md)** consumes the briefing: mechanical (code-fence-only) changes are applied with `--apply`, no subagent; prose diffs go to a subagent whose prompt is the briefing, not the corpus; verification is clause-by-clause on the changed spans, not the whole document.
 - **The pairing gate takes pair arguments.** `verify-translation-pairing [pair...]` checks just the named pairs (any of a pair's three files, or the bare stem, names it); the corpus-wide sweep remains the no-argument form that `doc-sync` and CI run. `--write` now requires naming the confirmed pairs — bare `--write` refuses, and re-recording everything is an explicit `--write --all` — because the old bare form silently blessed every drifted pair in the tree, including ones the caller never looked at, and a prose-only drift would then stay green forever. Each record's comment names its own scoped command. Before recording, `--write` stores each side's exact bytes with `git hash-object -w --stdin` and pins the blob under a content-addressed local `refs/dsh/translation-pairing/snapshots/` ref; an uncommitted last-confirmed snapshot is therefore available to the briefing generator's later `git cat-file`, not merely named by a hash that Git cannot resolve or left vulnerable to garbage collection.
 

.agents/notes/implemented/process/2026-07-26-briefed-minimal-translation-updates.zh.md

@@ -12,7 +12,7 @@ Status: implemented
 
 配对更新基于生成的简报（briefing）运行，而非基于指导语料；只有新建配对仍走整篇文档工作流，后者保持不变。
 
-- **`pnpm run gen-translation-brief [--apply] [pair...]`**（[scripts/gen-translation-brief.ts](../../../../scripts/gen-translation-brief.ts)，组装逻辑在 [scripts/translation-brief.ts](../../../../scripts/translation-brief.ts)）针对每个失去同步的配对，打印被改一侧从其记录在案的上次确认 blob 到当前工作区的 diff，并附上以能安全对齐的最窄粒度映射的这次改动，映射失败时粒度确定性地逐级放宽：仅落在配对中逐字节一致的围栏代码块内的改动会直接算出（`--apply` 会把它拼接进对侧文件，并在写入前用配对门禁的结构签名校验所得结果）；否则，每个有改动的 Markdown 单元（标题、段落、表格行、列表项、围栏代码块、块引用、HTML 块、分隔线、链接定义；匹配依据是以容器为作用域的种类序列）都带上各自的上次确认源文、当前源文与当前对侧文本及行号；无法对齐的单元回退到按深度匹配的标题章节；当章节也无法对齐或两侧同时漂移时，简报会明说这一点并省略映射，而不是靠猜。术语表行只与改动块匹配（英文术语按词边界匹配，含复数变形）；当目标侧是中文时，简报还会跟踪每个相关术语在整篇文档中的首次出现：一旦某次编辑使其移位，腾出的与接收的两处区间就会附一条解释性说明加入简报，因为「首次出现」括注必须随之移动。单元映射、代码拼接与首次出现机制采纳了[增量提示词流水线工作](https://github.com/deepseek-harness/deepseek-harness/pull/684)中的规划器设计；该项工作中接入提供方的对比评测，已为自动流水线独立验证了同一套范围阶梯。简报就是译者的全部工作集；简报回答不了的决策，仍以完整的真源文档作为升级求证路径。
+- **`pnpm run gen-translation-brief [--apply] [pair...]`**（[scripts/gen-translation-brief.ts](../../../../scripts/gen-translation-brief.ts)，组装逻辑在 [scripts/translation-brief.ts](../../../../scripts/translation-brief.ts)）针对每个失去同步的配对，打印被改一侧从其记录在案的上次确认 blob 到当前工作区的 diff，并附上以能安全对齐的最窄粒度映射的这次改动，映射失败时粒度确定性地逐级放宽：仅落在配对中逐字节一致的围栏代码块内的改动会直接算出（`--apply` 会把它拼接进对侧文件，并在写入前用配对门禁的结构签名校验所得结果）；否则，每个有改动的 Markdown 单元（标题、段落、表格行、列表项、围栏代码块、块引用、HTML 块、分隔线、链接定义；匹配依据是以容器为作用域的种类序列）都带上各自的上次确认源文、当前源文与当前对侧文本及行号；无法对齐的单元回退到按深度匹配的标题章节；当章节也无法对齐或两侧同时漂移时，简报会明说这一点并省略映射，而不是靠猜。术语表行只与改动块匹配（英文术语按词边界匹配，含复数变形）；当目标侧是中文时，简报还会跟踪每个相关术语在整篇文档中的首次出现：一旦某次编辑使其移位，腾出的与接收的两处区间就会附一条解释性说明加入简报，因为「首次出现」括注必须随之移动。单元映射、代码拼接与首次出现机制采纳了增量提示词流水线工作的规划器设计；该项工作中接入提供方的对比评测，已为自动流水线独立验证了同一套范围阶梯。简报就是译者的全部工作集；简报回答不了的决策，仍以完整的真源文档作为升级求证路径。
 - **[dsh-translate-docs](../../../skills/dsh-translate-docs/SKILL.md) 中的更新路径**消费这份简报：机械类改动（只涉及围栏代码块）用 `--apply` 应用，不动用 subagent；行文类 diff 交给 subagent，其提示词就是简报本身，而非指导语料；核验只对改动块逐句进行，不覆盖整篇文档。
 - **配对门禁接受配对参数。**`verify-translation-pairing [pair...]` 只检查被点名的配对（配对三个文件中的任意一个，或其裸词干，都能指代该配对）；全语料扫描仍是 `doc-sync`（文档同步门禁）与 CI 运行的无参数形式。`--write` 现在要求点名已确认的配对：裸 `--write` 会拒绝执行，重新记录全部配对必须显式写 `--write --all`；原因是旧的裸形式会默默为树中每一个漂移的配对背书，包括调用者从未看过的那些，纯行文层面的漂移于是可以永远保持绿灯。每份记录的注释都写明针对该配对自身的按对命令。写下记录之前，`--write` 用 `git hash-object -w --stdin` 存入每一侧的精确字节，并在内容寻址的本地 `refs/dsh/translation-pairing/snapshots/` ref 下固定该 blob；未提交的上次确认快照因此能被简报生成器之后的 `git cat-file` 取回，而不只是留下一个 Git 无法解析的 hash 名称或暴露于垃圾回收。
 

.agents/notes/implemented/process/2026-07-27-wine-windows-gates-experiment.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 .agents/notes/implemented/process/2026-07-27-wine-windows-gates-experiment.md
-2026-07-27-wine-windows-gates-experiment.md: 640c8e455b1a35ea4ac83454227147b9979316dc
-2026-07-27-wine-windows-gates-experiment.zh.md: 67f59a93d1b1fad98e36e6f9c51bc77abb9e3d07
+2026-07-27-wine-windows-gates-experiment.md: 1b01fe00dc1588482442a3cedc35eefa7fdb0975
+2026-07-27-wine-windows-gates-experiment.zh.md: b50739f5b63c5836248c94927037e38900abc84f

.agents/notes/implemented/process/2026-07-27-wine-windows-gates-experiment.md

@@ -14,7 +14,7 @@ The question the experiment answered: can a plain Linux runner produce an equiva
 
 The required pull-request `windows` job in [ci.yml](../../../../.github/workflows/ci.yml) (`windows node 24 / wine blocking`) runs the blocking gate commands on `ubuntu-latest` under Wine with real Windows binaries: a checksum-verified win-x64 Node.js executes `tsc -b`, `tsdown`, and the VitePress production build, so the win32 branches of the toolchain — backslash path handling, `CreateProcess` spawn semantics, PE loading of `@esbuild/win32-x64`, and the rolldown/rollup MSVC `.node` addons — actually execute. The master `serial-windows` job is untouched: the complete native-kernel inventory, including the observational portability gates this lane does not run, still executes on real `windows-2025` on every master push.
 
-Dependencies install natively on Linux with `supportedArchitectures` extended to win32-x64, which materializes the Windows platform packages in the same store; the cmd-shim layer is bypassed by invoking each tool's JavaScript entrypoint directly, the same processes `run-gates` ultimately spawns. `nodeLinker: hoisted` is load-bearing, not stylistic: the independent prototype in [PR #689](https://github.com/deepseek-harness/deepseek-harness/pull/689) kept pnpm's default isolated layout — including a faithful offline Windows-pnpm re-install over a Linux-prefetched store — and Windows Node under Wine still could not resolve `@esbuild/win32-x64` or load the koffi prebuild through the isolated symlink chain, failing before any repository gate ran. A flat layout with real files is what makes the gates reachable at all; #689's checksum pinning is adopted, while its Windows-pnpm-installs-the-tree goal is explicitly given up (the install contract stays Linux-tested here).
+Dependencies install natively on Linux with `supportedArchitectures` extended to win32-x64, which materializes the Windows platform packages in the same store; the cmd-shim layer is bypassed by invoking each tool's JavaScript entrypoint directly, the same processes `run-gates` ultimately spawns. `nodeLinker: hoisted` is load-bearing, not stylistic: an independent prototype kept pnpm's default isolated layout — including a faithful offline Windows-pnpm re-install over a Linux-prefetched store — and Windows Node under Wine still could not resolve `@esbuild/win32-x64` or load the koffi prebuild through the isolated symlink chain, failing before any repository gate ran. A flat layout with real files is what makes the gates reachable at all; the prototype's checksum pinning is adopted, while its Windows-pnpm-installs-the-tree goal is explicitly given up (the install contract stays Linux-tested here).
 
 The lane holds the wall clock of the Linux CI jobs through four levers: the master-refreshed pnpm store cache (restore-only, same key as the Linux jobs), Wine provisioning (apt install, Windows Node download, `wineboot`) running concurrently with `pnpm install`, the two blocking surfaces running concurrently — the same shape `run-gates` gives them on native Windows — and an apt-archive cache keyed on the runner image, seeded from master by the `wine apt cache` job so every pull request restores from the default-branch scope.
 
@@ -32,7 +32,7 @@ Measured on 2026-07-27, warm caches, pull-request trigger, standard 2-core `ubun
 
 **A full Windows guest under QEMU/KVM inside the Linux runner.** Real NT kernel, so full fidelity including case-insensitive NTFS and ConPTY — but tens of minutes of image download and unattended install before the first gate runs (40m19s measured end-to-end on the sibling experiment branch `exp/kvm-windows-ci`). Promotable only with disk-image caching that pressures the Actions cache budget.
 
-**Windows pnpm performing the install under Wine ([PR #689](https://github.com/deepseek-harness/deepseek-harness/pull/689)).** The higher-fidelity variant of this same idea: MinGit and pnpm staged into the prefix, a Linux prefetch filling the store, then `pnpm install --offline` run by Windows Node so the install contract itself executes as win32. It reached the install but not the gates — Wine's networking could not reach the registry directly, and the isolated `node_modules` layout defeated resolution of the Windows platform packages even after a clean offline install. This lane trades that fidelity away (hoisted layout, Linux-side install) to reach the gates; the two records are complementary halves of the same verdict.
+**Windows pnpm performing the install under Wine.** The higher-fidelity variant of this same idea: MinGit and pnpm staged into the prefix, a Linux prefetch filling the store, then `pnpm install --offline` run by Windows Node so the install contract itself executes as win32. It reached the install but not the gates — Wine's networking could not reach the registry directly, and the isolated `node_modules` layout defeated resolution of the Windows platform packages even after a clean offline install. This lane trades that fidelity away (hoisted layout, Linux-side install) to reach the gates; the two records are complementary halves of the same verdict.
 
 **Filesystem-semantics lanes on Linux (casefolded ext4, filename lint).** Catches the highest-frequency Windows breakage class for near-zero cost but proves nothing about win32 binaries. Explored as the sibling experiment branch `exp/casefold-windows-ci`; complementary to, not competitive with, this lane.
 

.agents/notes/implemented/process/2026-07-27-wine-windows-gates-experiment.zh.md

@@ -14,7 +14,7 @@ Pull request 的 Windows 通道旨在验证两个阻断性 win32 表面，即 wo
 
 [ci.yml](../../../../.github/workflows/ci.yml) 中必需的 pull request `windows` 作业（`windows node 24 / wine blocking`）在 `ubuntu-latest` 上通过 Wine 用真实 Windows 二进制运行阻断门禁命令：校验和验证过的 win-x64 Node.js 执行 `tsc -b`、`tsdown` 与 VitePress 生产构建，因此工具链的 win32 分支——反斜杠路径处理、`CreateProcess` 派生语义、`@esbuild/win32-x64` 的 PE 加载、以及 rolldown/rollup 的 MSVC `.node` 插件——都真正执行。master 的 `serial-windows` 作业原封不动：完整的原生内核清单，包括本通道不运行的观察性可移植性门禁，仍在每次 master push 时于真实 `windows-2025` 上执行。
 
-依赖在 Linux 上原生安装，`supportedArchitectures` 扩展到 win32-x64，使 Windows 平台包物化进同一个 store；通过直接调用各工具的 JavaScript 入口绕开 cmd-shim 层，这正是 `run-gates` 最终派生的那些进程。`nodeLinker: hoisted` 是承重的，不是风格问题：[PR #689](https://github.com/deepseek-harness/deepseek-harness/pull/689) 的独立原型保留了 pnpm 默认的 isolated 布局——包括在 Linux 预取的 store 上忠实地用 Windows pnpm 离线重装——而 Wine 下的 Windows Node 依然无法穿过 isolated 符号链接链解析 `@esbuild/win32-x64` 或加载 koffi 预编译产物，在任何仓库门禁运行前就失败了。扁平的真实文件布局才让门禁变得可达；本通道采纳了 #689 的校验和固定，同时明确放弃其「Windows pnpm 安装依赖树」的目标（安装契约在此仍由 Linux 侧验证）。
+依赖在 Linux 上原生安装，`supportedArchitectures` 扩展到 win32-x64，使 Windows 平台包物化进同一个 store；通过直接调用各工具的 JavaScript 入口绕开 cmd-shim 层，这正是 `run-gates` 最终派生的那些进程。`nodeLinker: hoisted` 是承重的，不是风格问题：一个独立原型保留了 pnpm 默认的 isolated 布局——包括在 Linux 预取的 store 上忠实地用 Windows pnpm 离线重装——而 Wine 下的 Windows Node 依然无法穿过 isolated 符号链接链解析 `@esbuild/win32-x64` 或加载 koffi 预编译产物，在任何仓库门禁运行前就失败了。扁平的真实文件布局才让门禁变得可达；本通道采纳了该原型的校验和固定，同时明确放弃其「Windows pnpm 安装依赖树」的目标（安装契约在此仍由 Linux 侧验证）。
 
 该通道靠四个杠杆把墙钟时间保持在与 Linux CI 作业相当的水平：master 刷新的 pnpm store 缓存（只恢复，与 Linux 作业同键）、Wine 供给（apt 安装、Windows Node 下载、`wineboot`）与 `pnpm install` 并发运行、两个阻断表面并发运行——与 `run-gates` 在原生 Windows 上给它们的形状相同——以及按 runner 镜像为键的 apt 归档缓存，由 master 的 `wine apt cache` 作业播种，使每个 pull request 都能从默认分支作用域恢复。
 
@@ -32,7 +32,7 @@ Pull request 的 Windows 通道旨在验证两个阻断性 win32 表面，即 wo
 
 **在 Linux runner 内用 QEMU/KVM 跑完整 Windows 客户机。** 真实 NT 内核，保真度完整，包括大小写不敏感的 NTFS 与 ConPTY——但首个门禁运行前要花数十分钟下载镜像并做无人值守安装（兄弟实验分支 `exp/kvm-windows-ci` 实测端到端 40 分 19 秒）。只有配上会挤压 Actions 缓存预算的磁盘镜像缓存才可投入使用。
 
-**在 Wine 下由 Windows pnpm 执行安装（[PR #689](https://github.com/deepseek-harness/deepseek-harness/pull/689)）。** 同一想法的更高保真度变体：把 MinGit 与 pnpm 放进 prefix，用 Linux 预取填充 store，再由 Windows Node 运行 `pnpm install --offline`，让安装契约本身以 win32 身份执行。它到达了安装但没到达门禁——Wine 的网络无法直接访问 registry，且 isolated 的 `node_modules` 布局即便在干净的离线安装后也挫败了 Windows 平台包的解析。本通道牺牲这份保真度（hoisted 布局、Linux 侧安装）来换取门禁可达；两份记录是同一裁决互补的两半。
+**在 Wine 下由 Windows pnpm 执行安装。** 同一想法的更高保真度变体：把 MinGit 与 pnpm 放进 prefix，用 Linux 预取填充 store，再由 Windows Node 运行 `pnpm install --offline`，让安装契约本身以 win32 身份执行。它到达了安装但没到达门禁——Wine 的网络无法直接访问 registry，且 isolated 的 `node_modules` 布局即便在干净的离线安装后也挫败了 Windows 平台包的解析。本通道牺牲这份保真度（hoisted 布局、Linux 侧安装）来换取门禁可达；两份记录是同一裁决互补的两半。
 
 **Linux 上的文件系统语义通道（casefold ext4、文件名 lint）。** 以近零成本捕获最高频的 Windows 故障类别，但对 win32 二进制什么也证明不了。作为兄弟实验分支 `exp/casefold-windows-ci` 探索；与本通道互补而非竞争。
 

.agents/notes/implemented/process/2026-07-31-installer-adopts-existing-checkout.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 .agents/notes/implemented/process/2026-07-31-installer-adopts-existing-checkout.md
-2026-07-31-installer-adopts-existing-checkout.md: de3cd052f94a0d5256c7687e9a1a38ee69fd2caf
-2026-07-31-installer-adopts-existing-checkout.zh.md: 2e8be804b4af6151e77e36f8b109616aab3a18e9
+2026-07-31-installer-adopts-existing-checkout.md: 3a213a6232e57f305910240505983421dcd288ad
+2026-07-31-installer-adopts-existing-checkout.zh.md: 7cdcd5549fb2bca2a6cb11f0cf2867687565ae79

.agents/notes/implemented/process/2026-07-31-installer-adopts-existing-checkout.md

@@ -46,6 +46,6 @@ A container adopting an outside clone is also no longer self-contained: deleting
 
 ## Testing
 
-`scripts/install.sh` has no automated test, and this change does not add one: the user directed that `install.spec.ts` be left out of scope. That is a known gap on a shipped user-facing path, and the `/var` resolution defect above is exactly the class of bug a test would have caught first. The standing [`FIXME(install-ts)`](../../../../scripts/install.sh) asking for this workflow to move into a tested TypeScript entrypoint is correspondingly more pressing.
+`scripts/install.sh` now has a real-shell PTY regression suite in `apps/cli/tests/install-script.spec.ts`, covering adoption and curl-style paths with stubbed dependencies. Curl-style installs default to the public `deepseek-ai/deepseek-harness-sdk` source, while replacing the installer with pnpm/npx remains separate work.
 
 Verification was manual, through a throwaway harness driving the real script with a stubbed `pnpm`: adopting a standalone clone; adopting from a linked worktree into its existing container; an explicit `DSH_SOURCE` still opting back into cloning; a dirty tree adopting silently with no prompt or warning while its uncommitted file stays behind; a non-git checkout failing with guidance; and a `curl`-style clone install asserting the built layout, which is the regression that caught the unresolved-`REPO_ROOT` defect. The interactive path was exercised under tmux from a dirty checkout, confirming the run reaches the launcher with no adoption prompt and ends with `dsh` running from the new staging worktree while the original checkout keeps its branch and its uncommitted file.

.agents/notes/implemented/process/2026-07-31-installer-adopts-existing-checkout.zh.md

@@ -46,6 +46,6 @@ Status: implemented
 
 ## Testing
 
-`scripts/install.sh`没有自动化测试，本次变更也未添加：用户明确要求把`install.spec.ts`排除在范围之外。这是一条已交付的、面向用户的安装路径上的已知缺口，而上文那个`/var`解析缺陷，恰恰属于测试本应最先捕获的那类 bug。相应地，要求把这套流程迁移到有测试覆盖的 TypeScript 入口的既有[`FIXME(install-ts)`](../../../../scripts/install.sh)也变得更为紧迫。
+`scripts/install.sh` 现有一套位于 `apps/cli/tests/install-script.spec.ts` 的真实 shell PTY 回归测试，使用 stub 依赖覆盖接管路径和 curl 风格路径。curl 风格安装默认使用公开的 `deepseek-ai/deepseek-harness-sdk` 源，而以 pnpm/npx 替换安装器仍是另一项工作。
 
 验证是手工完成的，通过一个一次性测试装置以打桩的`pnpm`驱动真实脚本：接管独立克隆；从 linked worktree 接管进其已有容器；显式`DSH_SOURCE`仍回到克隆路径；工作树不干净时静默接管、既不提示也不警告，且其未提交文件留在原处；非 git 检出失败并给出指引；以及`curl`式克隆安装断言所构建的布局——正是这项回归测试捕获了`REPO_ROOT`未解析的缺陷。交互路径在 tmux 下从一个不干净的检出走通，确认整个过程不出现接管提示即可到达启动器，最终`dsh`从新的 staging worktree 运行，而原检出保持其分支不变、未提交文件仍在。

.agents/notes/implemented/process/2026-08-06-doc-site-carries-its-images.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 .agents/notes/implemented/process/2026-08-06-doc-site-carries-its-images.md
-2026-08-06-doc-site-carries-its-images.md: 9109808874579b79d85c2e22b0987110f41ddc42
-2026-08-06-doc-site-carries-its-images.zh.md: d601112e8870150c363d8533e85ef86e7f3f8ffc
+2026-08-06-doc-site-carries-its-images.md: 4078a9b6251cf67456590ae25602a9f288c88dc1
+2026-08-06-doc-site-carries-its-images.zh.md: a9afb138d45d1ab991963b997e408477cf88110b

.agents/notes/implemented/process/2026-08-06-doc-site-carries-its-images.md

@@ -20,7 +20,7 @@ Only a regular file whose real path stays inside the repository is copied; anyth
 
 `docsSourceFiles()` reports the placed images alongside the Markdown, so the dev server's watcher re-projects when a screenshot is replaced instead of serving the previous copy until something touches the page.
 
-`placeImage` is optional because `rewriteMarkdown` is also called directly by its spec, where no generated tree exists. Without it the old GitHub-raw behavior stands, which keeps that seam honest: the fallback is still the correct answer for a consumer that only rewrites text.
+`placeImage` is optional because `rewriteMarkdown` is also called directly by its spec, where no generated tree exists. Without it the GitHub-raw fallback points at the public source home, which keeps that seam honest for a consumer that only rewrites text.
 
 Canonical Markdown keeps writing ordinary repository-relative image paths, so the same file renders on GitHub and on the site. No document carries a site-absolute URL to satisfy VitePress.
 
@@ -36,7 +36,7 @@ Canonical Markdown keeps writing ordinary repository-relative image paths, so th
 
 Images in published documentation now work regardless of who is reading or whether the repository is public, and the site build has no runtime dependency on GitHub for them. The generated tree grows by one copy of each referenced image per locale — the four screenshots in the model-provider guide add roughly 270 KB per locale.
 
-Images referenced from *unpublished* documents are untouched: they still resolve to GitHub raw, and still fail for a private repository. Nothing consumes them today, and a document that is not on the site has no site build to carry its assets.
+Images referenced from *unpublished* documents are untouched. A text-only projection resolves them against the public source home; a document that is not on the site has no site build to carry its assets.
 
 ## Testing
 

.agents/notes/implemented/process/2026-08-06-doc-site-carries-its-images.zh.md

@@ -20,7 +20,7 @@ Status: implemented
 
 `docsSourceFiles()` 会连同被安置的图片一起上报，于是替换截图时开发服务器的 watcher 会重新投影，而不是一直服务旧副本直到有人碰一下页面。
 
-`placeImage` 之所以可选，是因为 `rewriteMarkdown` 也被它自己的 spec 直接调用，而那里并不存在生成树。不传它时保持原有的 GitHub raw 行为，这也让该接缝保持诚实：对只改写文本的消费方而言，这个回退仍是正确答案。
+`placeImage` 之所以可选，是因为 `rewriteMarkdown` 也被它自己的 spec 直接调用，而那里并不存在生成树。不传它时，GitHub raw 回退会指向公开源主页；这让该 seam 对只改写文本的消费方保持诚实。
 
 正本 Markdown 照旧写普通的仓库相对图片路径，因此同一份文件在 GitHub 上和站点上都能正常显示。没有任何文档为了迁就 VitePress 而写站内绝对 URL。
 
@@ -36,7 +36,7 @@ Status: implemented
 
 已发布文档中的图片，现在无论谁在阅读、无论仓库是否公开都能显示，站点构建也不再为图片依赖 GitHub 的运行时可达性。生成树会为每个 locale 各增加一份被引用图片的副本——配置模型指南里的四张截图，每个 locale 约 270 KB。
 
-**未发布**文档引用的图片不受影响：它们仍解析到 GitHub raw，对私有仓库仍然失败。今天没有任何消费方用到它们，而不在站点上的文档也没有站点构建可以承载其资源。
+**未发布**文档引用的图片不受影响。纯文本投影会相对于公开源主页解析它们；不在站点上的文档没有站点构建可以承载其资源。
 
 ## Testing
 

.agents/notes/implemented/testing/2026-06-19-acp-snapshot-tests.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 .agents/notes/implemented/testing/2026-06-19-acp-snapshot-tests.md
-2026-06-19-acp-snapshot-tests.md: e118ada58230fe31fbb2a6bffb83e5612757ab1f
-2026-06-19-acp-snapshot-tests.zh.md: e292dbf3bf3c4c5198dc77d122bb6b8c36014ebf
+2026-06-19-acp-snapshot-tests.md: c7b95bd68027705b99d850d596405e56eea0dfca
+2026-06-19-acp-snapshot-tests.zh.md: 43d43262684920cae5feedb5f2eb109db00f9f8c

.agents/notes/implemented/testing/2026-06-19-acp-snapshot-tests.md

@@ -80,6 +80,6 @@ Tool determinism comes from a generated cwd, scrubbed environment, fresh non-log
 
 ## Consequences
 
-The tier adds reviewed per-scenario input, session, stdout, optional override, and optional workspace fixtures, plus one file for each distinct pinned prompt and tool-schema sequence. Workspace seeds are copied into the generated cwd for both record and replay. In return the tier provides deterministic keyless coverage through the real Loader and tool composition. Most retained scenarios exercise the assembled backend rather than ACP; the [automation-only ACP decision](../simplification/2026-07-23-acp-automation-only-protocol.md#snapshot-boundary) keeps that corpus here and defers any move to a transport-neutral headless suite as an independent testing change (the suite-level FIXME marks it).
+The tier adds reviewed per-scenario input, session, stdout, optional override, and optional workspace fixtures, plus one file for each distinct pinned prompt and tool-schema sequence. Workspace seeds are copied into the generated cwd for both record and replay. In return the tier provides deterministic keyless coverage through the real Loader and tool composition. Most retained scenarios exercise the assembled backend rather than ACP; the [automation-only ACP decision](../simplification/2026-07-23-acp-automation-only-protocol.md#snapshot-boundary) keeps that corpus here until it can move to a transport-neutral headless suite without losing coverage.
 
 This Agent Note relates to but does not supersede the [proposed determinism Agent Note](../../proposed/testing/2026-06-11-deterministic-and-stress-testing.md): that proposal's "universal replay fixture" re-derives session *message history* after every test (an internal-consistency invariant), whereas these snapshots pin assembled behavior plus the external automation output. They are complementary until the backend corpus moves off ACP.

.agents/notes/implemented/testing/2026-06-19-acp-snapshot-tests.zh.md

@@ -80,6 +80,6 @@ Status: implemented
 
 ## 后果
 
-该测试层为每个场景增加经过评审的输入、会话、stdout、可选 override 和可选 workspace fixture，并为每个不同的已固定提示词序列、每个不同的已固定工具 schema 序列各增加一个文件。记录与回放都会把 workspace seed 复制到生成的 cwd。作为回报，该层通过真实 Loader 和工具组合提供确定性的无密钥覆盖。保留下来的大多数场景测试的是组装后的后端而非 ACP；[仅面向自动化的 ACP 决策](../simplification/2026-07-23-acp-automation-only-protocol.md#snapshot-boundary)将该语料保留在此处，并把向传输无关 headless 套件的任何迁移推迟为一项独立的测试变更（套件级 FIXME 标记了这一点）。
+该测试层为每个场景增加经过评审的输入、会话、stdout、可选 override 和可选 workspace fixture，并为每个不同的已固定提示词序列、每个不同的已固定工具 schema 序列各增加一个文件。记录与回放都会把 workspace seed 复制到生成的 cwd。作为回报，该层通过真实 Loader 和工具组合提供确定性的无密钥覆盖。保留下来的大多数场景测试的是组装后的后端而非 ACP；[仅面向自动化的 ACP 决策](../simplification/2026-07-23-acp-automation-only-protocol.md#snapshot-boundary)将该语料保留在此处，直至它能够在不损失覆盖的情况下迁移到传输无关的 headless 套件。
 
 本 Agent Note 与[拟议的确定性 Agent Note](../../proposed/testing/2026-06-11-deterministic-and-stress-testing.md)相关，但不取代它：该提案的“通用回放 fixture”在每次测试后重新派生会话*消息历史*（内部一致性不变量），而这些快照固定组装后的行为与外部自动化输出。在后端语料迁出 ACP 之前，两者相互补充。

.github/workflows/ci.yml

@@ -30,7 +30,7 @@ env:
 
 jobs:
 
-  # FIXME: Re-enable the three hosted serial reference jobs before cutting a release.
+  # TODO(hosted-serial-ci): Re-enable the three hosted serial reference jobs before release.
   # The self-hosted standby remains active on every master push.
 
   # Three enterprise jobs isolate coverage, static analysis, and the

apps/web/tests/declared-reasoning.e2e.ts

@@ -0,0 +1,95 @@
+// Web e2e scenario: a hand-declared model's `reasoningEfforts` reaches the
+// composer's effort pane — the levels a settings profile declares are exactly
+// what the picker offers, and picking one records it with the default route.
+// Zero model calls: declaring, describing, and switching are settings/llm
+// traffic only, so there is no fixture and a stray stream would fail loud.
+import { readFile } from 'node:fs/promises'
+import { fileURLToPath } from 'node:url'
+import { join } from 'node:path'
+import type { Browser, Page } from 'playwright'
+import { chromium } from 'playwright'
+import { afterAll, beforeAll, describe, expect, it, onTestFailed } from 'vitest'
+import { settingsNamespace } from '@deepseek-ai/dsh-settings'
+import {
+  assertFixtureInventory, captureStableAria, compareOrRefreshGolden,
+  launchWebScaffold, watchConsole, webSnapshotMode, type WebScaffold,
+} from './scaffold.ts'
+import { ZH_BROWSER_LOCALE, connectFreshWorkspaceZh, saveFailureShot } from './support.ts'
+
+/** Starts the shipped default on this scenario's declared reasoning model. */
+const OVERLAY = fileURLToPath(new URL('./declared-reasoning.overlay.yml', import.meta.url))
+const SNAPSHOT_DIR = fileURLToPath(new URL('./snapshots/declared-reasoning', import.meta.url))
+const UI_EXPECTED = fileURLToPath(new URL('./snapshots/declared-reasoning/ui.expected.md', import.meta.url))
+const MODE = webSnapshotMode()
+
+describe.skipIf(MODE === 'record')('web e2e: declared reasoning efforts reach the composer', () => {
+  let scaffold: WebScaffold
+  let browser: Browser
+  let page: Page
+  let tripwire: ReturnType<typeof watchConsole>
+
+  beforeAll(async () => {
+    scaffold = await launchWebScaffold({ extraOverlayPath: OVERLAY })
+    // The whole reasoning offer is the profile: key = selectable level, value
+    // = the wire spelling dispatch would send (`max: ultra` renames; the
+    // valueless `off` means "supported, send nothing"). The route sets no
+    // deployment default, so the pane leads with the provider-default entry.
+    await scaffold.ctx.settings.update(settingsNamespace('llm-pi-ai'), {
+      providers: {
+        'acme-gateway': {
+          displayName: 'Acme Gateway',
+          api: 'openai-completions',
+          baseURL: 'https://gateway.acme.example/v1',
+          models: [{
+            id: 'acme-think',
+            name: 'Acme Think',
+            reasoningEfforts: { off: null, high: 'high', max: 'ultra' },
+          }],
+        },
+      },
+    })
+    browser = await chromium.launch()
+    page = await browser.newPage({ viewport: { width: 1680, height: 1000 }, locale: ZH_BROWSER_LOCALE })
+    tripwire = watchConsole(page)
+    await page.goto(scaffold.baseUrl, { waitUntil: 'load' })
+    await page.waitForSelector('[class*="frame"]', { timeout: 30_000 })
+    await connectFreshWorkspaceZh(page, scaffold.workspaceCwd)
+  }, 120_000)
+
+  afterAll(async () => {
+    await browser?.close()
+    await scaffold?.close()
+  })
+
+  it('offers exactly the declared levels and records the picked one', async () => {
+    onTestFailed(() => saveFailureShot(page, 'web-e2e-declared-reasoning'))
+    const trigger = page.getByRole('button', { name: /^选择模型/ })
+    await trigger.waitFor({ timeout: 15_000 })
+    await trigger.click()
+    await page.getByRole('menuitem', { name: /推理等级/ }).click()
+
+    // Declared levels, nothing else: the provider-default entry (the route
+    // configures no `reasoning`), then Off/High/Max — minimal, low, medium,
+    // and xhigh were not declared and must not be offered.
+    const levels = page.getByRole('menuitemradio')
+    await expect.poll(async () => levels.allTextContents(), { timeout: 10_000 })
+      .toEqual(['Default', 'Off', 'High', 'Max'])
+    const snapshot = await captureStableAria(page, '[role="menu"]', scaffold.workspaceCwd)
+    await compareOrRefreshGolden(UI_EXPECTED, snapshot, MODE)
+
+    // Picking a level is the same gesture that saves the default target, so
+    // the effort lands in the gateway's settings section beside the route.
+    await page.getByRole('menuitemradio', { name: 'High' }).click()
+    await expect.poll(
+      async () => readFile(join(scaffold.harnessHome, 'settings.yaml'), 'utf8'),
+      { timeout: 10_000 },
+    ).toContain('reasoningEffort: high')
+    await expect.poll(() => trigger.getAttribute('aria-label'), { timeout: 10_000 })
+      .toBe('选择模型，当前 Acme Think，推理等级 High')
+    expect(tripwire.pageErrors).toEqual([])
+  }, 60_000)
+
+  it('keeps its snapshot inventory closed', async () => {
+    await assertFixtureInventory(SNAPSHOT_DIR, ['ui.expected.md'])
+  })
+})

apps/web/tests/declared-reasoning.overlay.yml

@@ -0,0 +1,8 @@
+# The fixture-less web scaffold registers no adapter, so the shipped
+# deepseek-official default would be a route nothing serves. This scenario
+# starts the default on its own declared reasoning model so the effort pane
+# describes that model from the first open.
+- id: api-gateway
+  config:
+    provider: acme-gateway
+    model: acme-think

apps/web/tests/skill-invocation-policy.e2e.ts

@@ -1,5 +1,6 @@
-// Web e2e scenario: the real host filters skill.list to the model-and-user
-// intersection before the browser slash source renders candidates. A real
+// Web e2e scenario: the real host serves every user-invocable skill to the
+// browser slash source — user-only (disable-model-invocation) entries appear
+// with their marker while user-disabled quadrants stay hidden. A real
 // chromium connects a fresh workspace seeded with all four policy quadrants;
 // no model call is issued, so a stray stream fails loud on the open LLM seam.
 import { mkdir, writeFile } from 'node:fs/promises'
@@ -92,7 +93,7 @@ describe('web e2e: skill invocation policy through the real host', () => {
     await scaffold?.close()
   })
 
-  it('renders only the model-and-user intersection in slash candidates', async () => {
+  it('renders every user-invocable skill and marks the user-only entry', async () => {
     onTestFailed(() => saveFailureShot(page, 'web-e2e-skill-invocation-policy'))
     const input = page.locator('textarea').first()
     await input.fill('/policy')
@@ -102,8 +103,10 @@ describe('web e2e: skill invocation policy through the real host', () => {
       { timeout: 10_000 },
     ).toBe(1)
 
+    // The user-only quadrant is invocable here — its only entry point — and
+    // wears the user-only marker; both user-disabled quadrants stay hidden.
+    expect(await menu.getByRole('option', { name: /policy-user-only user-only · / }).count()).toBe(1)
     expect(await menu.getByRole('option', { name: /policy-model-only/ }).count()).toBe(0)
-    expect(await menu.getByRole('option', { name: /policy-user-only/ }).count()).toBe(0)
     expect(await menu.getByRole('option', { name: /policy-trusted-only/ }).count()).toBe(0)
 
     const snapshot = await captureStableAria(page, '[role="listbox"]', scaffold.workspaceCwd)

apps/web/tests/skill-user-invoke.e2e.ts

@@ -0,0 +1,148 @@
+// Web e2e scenario: a user invokes a disable-model-invocation skill through
+// the composer (issue #1470). The entered `/name args` line claims into
+// skill.invoke: the real host forwards the gesture as an ordinary user
+// prompt, injects the rendered body as instructions context named after the
+// skill, and starts a turn answered by the replay seam. The transcript shows
+// the gesture bubble, the collapsed context-injection row, and the reply.
+import { mkdir, mkdtemp, rm, writeFile } from 'node:fs/promises'
+import { tmpdir } from 'node:os'
+import { fileURLToPath } from 'node:url'
+import { join } from 'node:path'
+import type { Browser, Page } from 'playwright'
+import { chromium } from 'playwright'
+import { afterAll, beforeAll, describe, expect, it, onTestFailed } from 'vitest'
+import type { ReplayOverrideDoc } from '@deepseek-ai/dsh-llm-replay'
+import {
+  assertFixtureInventory,
+  captureStableAria,
+  compareOrRefreshGolden,
+  launchWebScaffold,
+  watchConsole,
+  webSnapshotMode,
+  type WebScaffold,
+} from './scaffold.ts'
+import { connectFreshWorkspace, newEnglishPage, saveFailureShot } from './support.ts'
+
+const SNAPSHOT_DIR = fileURLToPath(new URL('./snapshots/skill-user-invoke', import.meta.url))
+const UI_EXPECTED = join(SNAPSHOT_DIR, 'ui.expected.md')
+const MODE = webSnapshotMode()
+
+const SKILL_NAME = 'user-invoke-demo'
+const ARGS_TEXT = 'and confirm the fixture wiring'
+const REPLY = 'USER_INVOKE_REPLY acknowledged; following the injected skill.'
+
+async function seedUserOnlySkill(workspaceCwd: string): Promise<void> {
+  const directory = join(workspaceCwd, 'workspace', '.agents', 'skills', SKILL_NAME)
+  await mkdir(directory, { recursive: true })
+  await writeFile(join(directory, 'SKILL.md'), [
+    '---',
+    `name: ${SKILL_NAME}`,
+    'description: Prove user-explicit invocation of a model-hidden skill',
+    'disable-model-invocation: true',
+    '---',
+    '',
+    'Reply with the fixture acknowledgement line.',
+    '',
+  ].join('\n'))
+}
+
+const REPLAY: ReplayOverrideDoc = [{
+  kind: 'chunks',
+  chunks: [
+    { type: 'block-start', index: 0, blockType: 'text' },
+    { type: 'text-delta', index: 0, text: REPLY },
+    { type: 'block-end', index: 0, block: { type: 'text', text: REPLY } },
+    { type: 'usage', usage: { inputTokens: 256, outputTokens: 16 } },
+    { type: 'finish', reason: { kind: 'stop' } },
+  ],
+}]
+
+describe.skipIf(MODE === 'record')('web e2e: user-explicit skill invocation through the composer', () => {
+  let scaffold: WebScaffold
+  let browser: Browser
+  let page: Page
+  let replayDir: string
+  let tripwire: ReturnType<typeof watchConsole>
+
+  beforeAll(async () => {
+    replayDir = await mkdtemp(join(tmpdir(), 'dsh-skill-user-invoke-replay-'))
+    const replayOverride = join(replayDir, 'replay.override.json')
+    await writeFile(replayOverride, JSON.stringify(REPLAY))
+    scaffold = await launchWebScaffold({
+      replayFixture: join(replayDir, 'override-only.jsonl'),
+      replayOverride,
+      // Paced replay keeps the timing-derived chrome (TTFT / tok/s) present
+      // deterministically; instant playback races it in and out of the golden.
+      paceMs: 10,
+    })
+    await seedUserOnlySkill(scaffold.workspaceCwd)
+    browser = await chromium.launch()
+    page = await newEnglishPage(browser)
+    tripwire = watchConsole(page)
+    await page.goto(scaffold.baseUrl, { waitUntil: 'load' })
+    await page.waitForSelector('[class*="frame"]', { timeout: 30_000 })
+    await connectFreshWorkspace(page, scaffold.workspaceCwd)
+  }, 120_000)
+
+  afterAll(async () => {
+    const failures: unknown[] = []
+    await browser?.close().catch((error: unknown) => failures.push(error))
+    await scaffold?.close().catch((error: unknown) => failures.push(error))
+    if (replayDir !== undefined) {
+      await rm(replayDir, { recursive: true, force: true })
+        .catch((error: unknown) => failures.push(error))
+    }
+    if (failures.length === 1) throw failures[0]
+    if (failures.length > 1) throw new AggregateError(failures, 'skill-user-invoke e2e cleanup failed')
+  })
+
+  it('claims /name args into a gesture bubble, an injection row, and a replayed answer', async () => {
+    onTestFailed(() => saveFailureShot(page, 'web-e2e-skill-user-invoke'))
+    const composer = page.locator('textarea:enabled').last()
+    await composer.waitFor({ timeout: 15_000 })
+
+    // The menu lists the user-only skill (its only entry point) before enter.
+    await composer.fill(`/${SKILL_NAME}`)
+    const menu = page.getByRole('listbox', { name: 'Trigger suggestions' })
+    await expect.poll(
+      () => menu.getByRole('option', { name: new RegExp(SKILL_NAME) }).count(),
+      { timeout: 10_000 },
+    ).toBe(1)
+
+    await composer.fill(`/${SKILL_NAME} ${ARGS_TEXT}`)
+    await composer.press('Enter')
+
+    // The gesture stays an ordinary user bubble (decorated /name token plus
+    // the trailing text), ahead of the injected context.
+    const bubble = page.locator('[data-ref-chip="skill"]').first()
+    await bubble.waitFor({ timeout: 15_000 })
+    expect(await bubble.textContent()).toBe(`/${SKILL_NAME}`)
+
+    // The rendered body arrives as a context-injection row named after the
+    // skill; expanding it reveals the canonical <skill_content> block, and
+    // the user's text is NOT folded into it.
+    const injectionRow = page.getByRole('button', { name: `Context injection ${SKILL_NAME}` })
+    await injectionRow.waitFor({ timeout: 15_000 })
+    await injectionRow.click()
+    const injectionBody = page
+      .locator('[data-context-injection-body]')
+      .filter({ hasText: `<skill_content name="${SKILL_NAME}">` })
+    await injectionBody.waitFor({ timeout: 10_000 })
+    const injected = await injectionBody.textContent()
+    expect(injected).toContain('Reply with the fixture acknowledgement line.')
+    expect(injected).not.toContain(ARGS_TEXT)
+    await injectionRow.click()
+
+    // The injection started a turn; the replay seam answers it.
+    await page.getByText('USER_INVOKE_REPLY', { exact: false }).first().waitFor({ timeout: 20_000 })
+
+    const snapshot = await captureStableAria(page, '[class*="centerCol"]', scaffold.workspaceCwd)
+    await compareOrRefreshGolden(UI_EXPECTED, snapshot, MODE)
+    expect(tripwire.pageErrors).toEqual([])
+    expect(tripwire.warnings).toEqual([])
+  }, 60_000)
+
+  it('keeps its snapshot inventory closed', async () => {
+    await assertFixtureInventory(SNAPSHOT_DIR, ['ui.expected.md'])
+  })
+})

apps/web/tests/snapshots/declared-reasoning/ui.expected.md

@@ -0,0 +1,7 @@
+- menu "模型与推理等级":
+  - menuitemradio "Default" [checked]:
+    - text: Default
+    - img
+  - menuitemradio "Off"
+  - menuitemradio "High"
+  - menuitemradio "Max"

apps/web/tests/snapshots/skill-invocation-policy/menu.expected.md

@@ -1,3 +1,4 @@
 - listbox "Trigger suggestions":
   - text: Skills
   - option "policy-shared Available to both model and user invocation" [selected]
+  - option "policy-user-only user-only · Available only to user invocation"

apps/web/tests/snapshots/skill-user-invoke/ui.expected.md

@@ -0,0 +1,33 @@
+- banner:
+  - navigation "Session hierarchy":
+    - button "/user-invoke-demo and confirm the fixtur" [disabled]
+  - tablist:
+    - tab "Chat" [selected]
+    - tab "Trajectory"
+- text: /user-invoke-demo and confirm the fixture wiring {{clock}}
+- button "Copy":
+  - img
+- button "Context injection @deepseek-ai/dsh-system-prompt":
+  - img
+  - img
+  - text: Context injection @deepseek-ai/dsh-system-prompt
+- button "Context injection user-invoke-demo":
+  - img
+  - img
+  - text: Context injection user-invoke-demo
+- paragraph: USER_INVOKE_REPLY acknowledged; following the injected skill.
+- button "Copy":
+  - img
+- button "Branch into a new conversation":
+  - img
+- text: {{clock}} Ran for {{duration}} TTFT {{duration}} {{throughput}} tok/s
+- textbox "Message the agent"
+- button "Commands":
+  - img
+- 'button "Access mode, current: Workspace Write"': Workspace Write
+- button "Select model, current DeepSeek-V4-Flash":
+  - text: DeepSeek-V4-Flash
+  - img
+- button "0% of context used"
+- button "Send message" [disabled]
+- text: 1 turns · 1 steps LLM {{duration}} TTFT avg {{duration}} · {{throughput}} tok/s Cache hit 0% Input 256 tok · Output 16 tok

apps/web/tsconfig.json

@@ -38,6 +38,7 @@
     "tests/settings-chrome.e2e.ts",
     "tests/models-settings.e2e.ts",
     "tests/default-model.e2e.ts",
+    "tests/declared-reasoning.e2e.ts",
     "tests/onboarding-deepseek-config.e2e.ts",
     "tests/remote-welcome.e2e.ts",
     "tests/workspace-management.e2e.ts",
@@ -57,6 +58,7 @@
     "tests/markdown-inline-code-links.e2e.ts",
     "tests/queue-actions.e2e.ts",
     "tests/skill-invocation-policy.e2e.ts",
+    "tests/skill-user-invoke.e2e.ts",
     "tests/permission-policy-context.e2e.ts",
     "tests/access-confirmation.e2e.ts",
     "tests/shipped-composition.e2e.ts",

docs/config-catalog.md

@@ -753,6 +753,22 @@ export interface PiAiProviderProfile {
    * unset fields from the installed model of the same id.
    */
   models?: PiAiModelProfile[]
+  /**
+   * Installed-catalog customizations by model id: each entry reshapes that
+   * one model with the same fields a {@link models} entry takes, while the
+   * rest of the catalog keeps serving untouched. Only meaningful on a catalog
+   * route with no `models` list — `models` already replaces the catalog, so
+   * an override beside it, on a route the catalog does not ship, or naming a
+   * model the catalog does not describe is refused rather than skipped.
+   */
+  modelOverrides?: Record<string, PiAiModelOverride>
+  /**
+   * Reasoning-dispatch switches for every `openai-completions` model on this
+   * route; each model's own `compat` overrides per field. What neither sets
+   * keeps the installed catalog entry's value, then pi-ai's baseURL-derived
+   * detection.
+   */
+  compat?: PiAiCompatProfile
   /**
    * Context capacity for a model this route lists that neither the entry nor
    * the installed catalog sizes (default 262,144). A guess by construction, so
@@ -800,12 +816,70 @@ export interface PiAiModelProfile {
    * default on its own.
    */
   maxTokens?: number
+  /**
+   * Selectable reasoning efforts. Absent inherits the installed catalog
+   * entry's capability (a hand-declared model has none and does not reason);
+   * `false` declares a non-reasoning model, which is how a profile strips
+   * reasoning from a catalog model its gateway cannot serve; a non-empty dict
+   * declares the offered levels and their wire spellings.
+   */
+  reasoningEfforts?: false | PiAiReasoningEfforts
+  /** Reasoning-dispatch switches for this model, winning over the route's. */
+  compat?: PiAiCompatProfile
 }
+
+/**
+ * Customization of one installed catalog model, keyed by its id in the
+ * route's `modelOverrides` dict — the same fields a `models` entry may set,
+ * with the id living in the key. Unlike a `models` list, overrides leave the
+ * rest of the catalog serving untouched, which is what makes "correct one
+ * model, keep the other thirty-seven" a three-line edit.
+ */
+export type PiAiModelOverride = Omit<PiAiModelProfile, 'id'>
+
+/**
+ * Reasoning-dispatch compatibility switches, set on the route (its models'
+ * default) or per model (winning over the route). Only the switches pi-ai's
+ * reasoning dispatch reads are offered; the rest of pi-ai's compat surface
+ * keeps its baseURL-derived auto-detection. pi-ai types both fields only on
+ * `OpenAICompletionsCompat` — the other wire protocols carry their reasoning
+ * shape in the protocol itself — so resolution rejects a model-level switch
+ * anywhere else, while a route-level default skips past models it cannot fit.
+ */
+export interface PiAiCompatProfile {
+  /** Reasoning parameter shape the endpoint expects; absent keeps the catalog entry's, then pi-ai's baseURL-derived guess. */
+  thinkingFormat?: PiAiThinkingFormat
+  /** Whether the endpoint accepts `reasoning_effort`; absent keeps the catalog entry's, then pi-ai's baseURL-derived guess. */
+  supportsReasoningEffort?: boolean
+}
+
+/**
+ * Selectable reasoning efforts for one model: each key is a level the model
+ * offers (and selectors show), and its value is the wire spelling dispatch
+ * sends for it. `off` alone may leave its value empty — "supported, send
+ * nothing" — because for most providers not thinking is the parameter's
+ * absence; every other declared level must name a wire value. A level absent
+ * from the dict is not offered.
+ */
+export type PiAiReasoningEfforts = Partial<Record<ModelThinkingLevel, string | null>>
+
+/** One reasoning-dispatch wire format a profile may name. */
+export type PiAiThinkingFormat = Exclude<PiThinkingFormat, WithheldThinkingFormat>
+
+/** The `compat.thinkingFormat` spellings pi-ai accepts on an `openai-completions` model. */
+type PiThinkingFormat = NonNullable<OpenAICompletionsCompat['thinkingFormat']>
+
+/**
+ * pi-ai thinking formats a profile cannot name: both drive the request through
+ * `chatTemplateKwargs`, which this configuration does not expose, so offering
+ * them would hand back a format with nothing to say.
+ */
+type WithheldThinkingFormat = 'chat-template' | 'qwen-chat-template'
 ```
 
-Depends on: `CacheRetention` (`@earendil-works/pi-ai`) · `ModelThinkingLevel` (`@earendil-works/pi-ai`) · [`RetryPolicyConfig`](../packages/llm/llm/src/index.ts) · `ThinkingBudgets` (`@earendil-works/pi-ai`) · `Transport` (`@earendil-works/pi-ai`)
+Depends on: `CacheRetention` (`@earendil-works/pi-ai`) · `ModelThinkingLevel` (`@earendil-works/pi-ai`) · `OpenAICompletionsCompat` (`@earendil-works/pi-ai`) · [`RetryPolicyConfig`](../packages/llm/llm/src/index.ts) · `ThinkingBudgets` (`@earendil-works/pi-ai`) · `Transport` (`@earendil-works/pi-ai`)
 
-Source: [`packages/llm/llm-pi-ai/src/config.ts:120`](../packages/llm/llm-pi-ai/src/config.ts)
+Source: [`packages/llm/llm-pi-ai/src/config.ts:142`](../packages/llm/llm-pi-ai/src/config.ts)
 
 ## `@deepseek-ai/dsh-llm-replay`
 
@@ -1457,7 +1531,7 @@ export interface Config {
 }
 ```
 
-Source: [`packages/skill/skill/src/index.ts:170`](../packages/skill/skill/src/index.ts)
+Source: [`packages/skill/skill/src/index.ts:262`](../packages/skill/skill/src/index.ts)
 
 ## `@deepseek-ai/dsh-skill-local`
 
@@ -2049,7 +2123,7 @@ export interface Config {
 }
 ```
 
-Source: [`packages/skill/tool-skill/src/index.ts:58`](../packages/skill/tool-skill/src/index.ts)
+Source: [`packages/skill/tool-skill/src/index.ts:61`](../packages/skill/tool-skill/src/index.ts)
 
 ## `@deepseek-ai/dsh-tool-str-replace-editor`
 

docs/cordis-catalog/events.md

@@ -677,7 +677,7 @@ A skill provider, runtime contribution, or provider-backed catalog may have chan
 'skills/change'(): void
 ```
 
-Source: [`packages/skill/skill/src/index.ts:188`](../../packages/skill/skill/src/index.ts)
+Source: [`packages/skill/skill/src/index.ts:280`](../../packages/skill/skill/src/index.ts)
 
 ## `subagent/*`
 

docs/cordis-catalog/services.md

@@ -1946,7 +1946,7 @@ async get(name: string, options: SkillLookupOptions = {}): Promise<SkillDefiniti
 
 Types: [SkillCatalogSnapshot](../core-data-structures/skills.md) · [SkillDefinition](../core-data-structures/skills.md) · [SkillLookupOptions](../core-data-structures/skills.md) · [SkillProvider](../core-data-structures/skills.md) · [SkillProviderControl](../core-data-structures/skills.md) · [SkillRegistration](../core-data-structures/skills.md) · [SkillSummary](../core-data-structures/skills.md)
 
-Source: [`packages/skill/skill/src/index.ts:209`](../../packages/skill/skill/src/index.ts)
+Source: [`packages/skill/skill/src/index.ts:301`](../../packages/skill/skill/src/index.ts)
 
 ## `ctx.spillStore` — `SpillStore` (abstract seam)
 

docs/cordis-tutorial/01-first-plugin.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 docs/cordis-tutorial/01-first-plugin.md
-01-first-plugin.md: c44e7f95fb11d5337ecfaf4251c8b2f2b9b14680
-01-first-plugin.zh.md: 9461884d312ad2e64af12fa42952a986e1ad5d8a
+01-first-plugin.md: 4359dfe4883f12e9cb242cf3009827fd7864768c
+01-first-plugin.zh.md: 62ccb7e5d37beb5b9636439e563cea6eaa1044a0

docs/cordis-tutorial/01-first-plugin.md

@@ -92,4 +92,4 @@ One caveat worth knowing early: a config entry whose module cannot be **resolved
 
 Next: [Lifecycle and effects](02-lifecycle-and-effects.md) — what happens when a plugin unloads.
 
-[![](https://img.shields.io/badge/powered_by-dsh-4D6BFE?style=flat-square&logo=deepseek&logoColor=white)](https://github.com/deepseek-harness/deepseek-harness)
+[![](https://img.shields.io/badge/powered_by-dsh-4D6BFE?style=flat-square&logo=deepseek&logoColor=white)](https://github.com/deepseek-ai/deepseek-harness-sdk)

docs/cordis-tutorial/01-first-plugin.zh.md

@@ -92,4 +92,4 @@ export function apply(ctx: Context) {
 
 下一章：[生命周期与 effect](02-lifecycle-and-effects.md)：插件卸载时会发生什么。
 
-[![](https://img.shields.io/badge/powered_by-dsh-4D6BFE?style=flat-square&logo=deepseek&logoColor=white)](https://github.com/deepseek-harness/deepseek-harness)
+[![](https://img.shields.io/badge/powered_by-dsh-4D6BFE?style=flat-square&logo=deepseek&logoColor=white)](https://github.com/deepseek-ai/deepseek-harness-sdk)

docs/cordis-tutorial/02-lifecycle-and-effects.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 docs/cordis-tutorial/02-lifecycle-and-effects.md
-02-lifecycle-and-effects.md: f1b39e06e9d25c51ab2d76503025e2b6ffe90c73
-02-lifecycle-and-effects.zh.md: 2e98e3af6d2f2b1b9cbb8ea38559bc1ffbf7e43b
+02-lifecycle-and-effects.md: 7b195b63a1e8730f27b9dd9af8af6a68a588cee9
+02-lifecycle-and-effects.zh.md: 4a3f83dedd5c95c7fcb5c1aebbbb8cb2e849b9cf

docs/cordis-tutorial/02-lifecycle-and-effects.md

@@ -95,4 +95,4 @@ One ordering caveat: disposers start in reverse registration order, but multiple
 
 Next: [Services](03-services.md) — how plugins share capabilities.
 
-[![](https://img.shields.io/badge/powered_by-dsh-4D6BFE?style=flat-square&logo=deepseek&logoColor=white)](https://github.com/deepseek-harness/deepseek-harness)
+[![](https://img.shields.io/badge/powered_by-dsh-4D6BFE?style=flat-square&logo=deepseek&logoColor=white)](https://github.com/deepseek-ai/deepseek-harness-sdk)

docs/cordis-tutorial/02-lifecycle-and-effects.zh.md

@@ -95,4 +95,4 @@ PENDING → LOADING → ACTIVE → UNLOADING → DISPOSED
 
 下一章：[服务](03-services.md)：插件如何共享功能。
 
-[![](https://img.shields.io/badge/powered_by-dsh-4D6BFE?style=flat-square&logo=deepseek&logoColor=white)](https://github.com/deepseek-harness/deepseek-harness)
+[![](https://img.shields.io/badge/powered_by-dsh-4D6BFE?style=flat-square&logo=deepseek&logoColor=white)](https://github.com/deepseek-ai/deepseek-harness-sdk)

docs/cordis-tutorial/03-services.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 docs/cordis-tutorial/03-services.md
-03-services.md: 5848132c6ad18338fa893954d45fc20005db6199
-03-services.zh.md: 0f599f082573364e6ad38278e1914d8faf67faa1
+03-services.md: 562b49ede0aa4cc1d58c4d6af7c7d5d1ebb2e4b1
+03-services.zh.md: 964f7e3654614d136b8765bb727f85a5a05587a8

docs/cordis-tutorial/03-services.md

@@ -95,4 +95,4 @@ Service names live in one flat namespace per application. Prefix or namespace yo
 
 Next: [Events](04-events.md) — communication without a shared service.
 
-[![](https://img.shields.io/badge/powered_by-dsh-4D6BFE?style=flat-square&logo=deepseek&logoColor=white)](https://github.com/deepseek-harness/deepseek-harness)
+[![](https://img.shields.io/badge/powered_by-dsh-4D6BFE?style=flat-square&logo=deepseek&logoColor=white)](https://github.com/deepseek-ai/deepseek-harness-sdk)

docs/cordis-tutorial/03-services.zh.md

@@ -95,4 +95,4 @@ export function apply(ctx: Context) {
 
 下一章：[事件](04-events.md)：无需共享服务即可通信。
 
-[![](https://img.shields.io/badge/powered_by-dsh-4D6BFE?style=flat-square&logo=deepseek&logoColor=white)](https://github.com/deepseek-harness/deepseek-harness)
+[![](https://img.shields.io/badge/powered_by-dsh-4D6BFE?style=flat-square&logo=deepseek&logoColor=white)](https://github.com/deepseek-ai/deepseek-harness-sdk)

docs/cordis-tutorial/04-events.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 docs/cordis-tutorial/04-events.md
-04-events.md: 18f39dc1b693e5fb7e1793ec4b7dcac9cf24db95
-04-events.zh.md: 3fdafb50303f49dca179bcaea32db211a66241f6
+04-events.md: 28ccb85d657afaabb5c6b4b1e9b10d6cf8710918
+04-events.zh.md: f78c971dcd9674d2a256c41000b627aecb2a572a

docs/cordis-tutorial/04-events.md

@@ -141,4 +141,4 @@ The harness uses waterfalls for decisions that cooperating plugins may wrap or a
 
 Next: [Configuration](05-config.md) — plugin options from `cordis.yml`.
 
-[![](https://img.shields.io/badge/powered_by-dsh-4D6BFE?style=flat-square&logo=deepseek&logoColor=white)](https://github.com/deepseek-harness/deepseek-harness)
+[![](https://img.shields.io/badge/powered_by-dsh-4D6BFE?style=flat-square&logo=deepseek&logoColor=white)](https://github.com/deepseek-ai/deepseek-harness-sdk)

docs/cordis-tutorial/04-events.zh.md

@@ -141,4 +141,4 @@ harness 使用 waterfall 处理协作插件可以包装或回答的决策：[`ag
 
 下一章：[配置](05-config.md)：来自 `cordis.yml` 的插件选项。
 
-[![](https://img.shields.io/badge/powered_by-dsh-4D6BFE?style=flat-square&logo=deepseek&logoColor=white)](https://github.com/deepseek-harness/deepseek-harness)
+[![](https://img.shields.io/badge/powered_by-dsh-4D6BFE?style=flat-square&logo=deepseek&logoColor=white)](https://github.com/deepseek-ai/deepseek-harness-sdk)

docs/cordis-tutorial/05-config.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 docs/cordis-tutorial/05-config.md
-05-config.md: 8d4043e33a58fc425d82d9846ff82473bcdef4c1
-05-config.zh.md: e9463bd34e9c72dbae7b1ceb9907e35edf7b773b
+05-config.md: 834bb140cc1ff976acc8f21c8f54a7fb02636eac
+05-config.zh.md: f5cc6ac1ca4fa02eba6a1b015b9f6ae3b1a925fc

docs/cordis-tutorial/05-config.md

@@ -81,4 +81,4 @@ The loader used in this repo supports a `!!js` tag for config values that must b
 
 Next: [Composition and HMR](06-composition-and-hmr.md) — treating `cordis.yml` as the application.
 
-[![](https://img.shields.io/badge/powered_by-dsh-4D6BFE?style=flat-square&logo=deepseek&logoColor=white)](https://github.com/deepseek-harness/deepseek-harness)
+[![](https://img.shields.io/badge/powered_by-dsh-4D6BFE?style=flat-square&logo=deepseek&logoColor=white)](https://github.com/deepseek-ai/deepseek-harness-sdk)

docs/cordis-tutorial/05-config.zh.md

@@ -81,4 +81,4 @@ ValidationError: invalid config:
 
 下一章：[组合与 HMR（热模块替换）](06-composition-and-hmr.md)：将 `cordis.yml` 视为应用。
 
-[![](https://img.shields.io/badge/powered_by-dsh-4D6BFE?style=flat-square&logo=deepseek&logoColor=white)](https://github.com/deepseek-harness/deepseek-harness)
+[![](https://img.shields.io/badge/powered_by-dsh-4D6BFE?style=flat-square&logo=deepseek&logoColor=white)](https://github.com/deepseek-ai/deepseek-harness-sdk)

docs/cordis-tutorial/06-composition-and-hmr.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 docs/cordis-tutorial/06-composition-and-hmr.md
-06-composition-and-hmr.md: 66d6a9d93fe39baa881940ba32388979e2678505
-06-composition-and-hmr.zh.md: 7c0a94b0abcc0f153f59391fd009f1e0b40500e5
+06-composition-and-hmr.md: f138918a2d217ed98fdfd4e56dffddc14e3397f0
+06-composition-and-hmr.zh.md: a678e86735c6d9e3b4f3cd0dfa46a2a64079c762

docs/cordis-tutorial/06-composition-and-hmr.md

@@ -110,4 +110,4 @@ needs-timer is PENDING — a required service is missing
 
 Next: [Into the harness](07-into-the-harness.md) — the same patterns against real harness services.
 
-[![](https://img.shields.io/badge/powered_by-dsh-4D6BFE?style=flat-square&logo=deepseek&logoColor=white)](https://github.com/deepseek-harness/deepseek-harness)
+[![](https://img.shields.io/badge/powered_by-dsh-4D6BFE?style=flat-square&logo=deepseek&logoColor=white)](https://github.com/deepseek-ai/deepseek-harness-sdk)

docs/cordis-tutorial/06-composition-and-hmr.zh.md

@@ -110,4 +110,4 @@ needs-timer is PENDING — a required service is missing
 
 下一章：[进入 harness](07-into-the-harness.md)：把相同模式用于真实的 harness 服务。
 
-[![](https://img.shields.io/badge/powered_by-dsh-4D6BFE?style=flat-square&logo=deepseek&logoColor=white)](https://github.com/deepseek-harness/deepseek-harness)
+[![](https://img.shields.io/badge/powered_by-dsh-4D6BFE?style=flat-square&logo=deepseek&logoColor=white)](https://github.com/deepseek-ai/deepseek-harness-sdk)

docs/cordis-tutorial/07-into-the-harness.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 docs/cordis-tutorial/07-into-the-harness.md
-07-into-the-harness.md: 6ec42c50fe5059955734fe7bc46117538dafaffc
-07-into-the-harness.zh.md: 903adb903aa4c4355b92eb34e89f218a0295767c
+07-into-the-harness.md: e02f8f8d55b3fbe9087d46f8f50baeecb592c1c6
+07-into-the-harness.zh.md: 5f770267e8f3db04cd9cb0e92b6a6278cf05d5e4

docs/cordis-tutorial/07-into-the-harness.md

@@ -104,4 +104,4 @@ Where to go next:
 - The generated [services](../cordis-catalog/services.md) and [events](../cordis-catalog/events.md) catalogs — everything you can inject and listen to.
 - [Architecture](../architecture.md) — the system map these plugins live in.
 
-[![](https://img.shields.io/badge/powered_by-dsh-4D6BFE?style=flat-square&logo=deepseek&logoColor=white)](https://github.com/deepseek-harness/deepseek-harness)
+[![](https://img.shields.io/badge/powered_by-dsh-4D6BFE?style=flat-square&logo=deepseek&logoColor=white)](https://github.com/deepseek-ai/deepseek-harness-sdk)

docs/cordis-tutorial/07-into-the-harness.zh.md

@@ -104,4 +104,4 @@ logger 会先触发：`tools/result` 在结果物化过程中发出，发生在
 - 生成的[服务](../cordis-catalog/services.md)与[事件](../cordis-catalog/events.md)目录：可以注入和监听的所有内容。
 - [架构](../architecture.md)：这些插件所处的系统地图。
 
-[![](https://img.shields.io/badge/powered_by-dsh-4D6BFE?style=flat-square&logo=deepseek&logoColor=white)](https://github.com/deepseek-harness/deepseek-harness)
+[![](https://img.shields.io/badge/powered_by-dsh-4D6BFE?style=flat-square&logo=deepseek&logoColor=white)](https://github.com/deepseek-ai/deepseek-harness-sdk)

docs/cordis-tutorial/index.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 docs/cordis-tutorial/index.md
-index.md: af622ad4e35829c6283c40f1b0019d7959dac973
-index.zh.md: 0b7684a9532a1efdcc3ea2d067da23852d146e2f
+index.md: a20976706f520416236ca759ee33649d1601eaa9
+index.zh.md: f6989521d4b7dffac6114867cc12371af4e4316f

docs/cordis-tutorial/index.md

@@ -13,7 +13,7 @@ If you want the condensed concept reference instead of a walkthrough, read the [
 You need a clone of this repository with dependencies installed — the [quick start](../user/guide/quickstart.md) covers prerequisites. No API key is needed for this tutorial; every example runs keylessly.
 
 ```sh
-git clone https://github.com/deepseek-harness/deepseek-harness.git
+git clone https://github.com/deepseek-ai/deepseek-harness-sdk.git
 cd deepseek-harness
 pnpm install
 ```
@@ -55,4 +55,4 @@ The examples use three TypeScript features beyond ordinary modern JavaScript:
 
 Chapter 5 also uses an `interface` to describe a configuration object's fields and a generic type such as `Schema<Config>` to say which object shape a schema validates. You can copy those declarations as shown; the surrounding text explains what each one connects.
 
-[![](https://img.shields.io/badge/powered_by-dsh-4D6BFE?style=flat-square&logo=deepseek&logoColor=white)](https://github.com/deepseek-harness/deepseek-harness)
+[![](https://img.shields.io/badge/powered_by-dsh-4D6BFE?style=flat-square&logo=deepseek&logoColor=white)](https://github.com/deepseek-ai/deepseek-harness-sdk)

docs/cordis-tutorial/index.zh.md

@@ -13,7 +13,7 @@ Cordis 是 DeepSeek Harness SDK 底层的插件框架：它是一个小型运行
 你需要克隆本仓库并安装依赖，具体前置条件见[快速入门](../user/guide/quickstart.md)。本教程不需要 API 密钥；所有示例均可在无密钥环境中运行。
 
 ```sh
-git clone https://github.com/deepseek-harness/deepseek-harness.git
+git clone https://github.com/deepseek-ai/deepseek-harness-sdk.git
 cd deepseek-harness
 pnpm install
 ```
@@ -55,4 +55,4 @@ node --import tsx ../../vendor/cordis/bin.js
 
 第 5 章还会使用 `interface` 描述配置对象的字段，并使用 `Schema<Config>` 这类泛型表示 schema 所校验的对象形状。你可以直接照写这些声明；周围的正文会解释每项声明连接了什么。
 
-[![](https://img.shields.io/badge/powered_by-dsh-4D6BFE?style=flat-square&logo=deepseek&logoColor=white)](https://github.com/deepseek-harness/deepseek-harness)
+[![](https://img.shields.io/badge/powered_by-dsh-4D6BFE?style=flat-square&logo=deepseek&logoColor=white)](https://github.com/deepseek-ai/deepseek-harness-sdk)

docs/core-data-structures/core.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 docs/core-data-structures/core.md
-core.md: f7cf288715a3aec2f7037f12fc983e3172a77cef
-core.zh.md: c17fd1335503c95e7f7f6f96cc286f567a8384e6
+core.md: dbd584f10b3daf873bc14210472efb6cd315717e
+core.zh.md: 1fe1616a0c96abb4e8b91417cc4eae292416e42a