dsh-vibe-math 2.8.3 → 2.8.5

package/README.en.md

@@ -168,14 +168,22 @@ Two installation methods, choose either one (they can also coexist):
 
 ### Method A: one-click install as a plugin package (recommended, installs all four presets at once)
 
-On the desktop app, install `dsh-vibe-math` from **Settings → Plugins**; on the command line, use a profile:
+On the desktop app, install `dsh-vibe-math` from **Settings → Plugins**; on the command line, use a profile (**always pin a version**):
 
 ```sh
-dsh plugin --profile <your profile> add dsh-vibe-math
+# Prerequisite: DSH forwards plugin installs to pnpm (it does not bundle pnpm) - check it first:
+pnpm --version
+
+# Install this package (with an explicit version; without one you may stay on an OLD version, see below):
+dsh plugin --profile <your profile> add dsh-vibe-math@<version>
 # Or install directly from GitHub:
 dsh plugin --profile <your profile> add github:ChongCyrus/Vibe-Mathematics
 ```
 
+> **Prerequisite**: `dsh plugin …` needs a working **`pnpm`** (DSH **forwards** installs to pnpm and does **not** bundle it). When it is missing the command ends with **`exit 1`**, while the **plugin market may show no reason** ⇒ attach `~/.dsh/profiles/<profile>/hub.log` (DSH's install log) so the failure can be located.
+>
+> **Why every example pins `@<version>`**: pnpm resolves a bare package name through the profile's `package.json` and `pnpm-lock.yaml` first, so **without a version you may stay on an OLD version**. Measured locally: `@latest`, `@^2`, `pnpm update --latest` and `pnpm add …@latest` **all four failed to move the install**; **only an explicit version** (e.g. `@2.8.5`) actually upgrades. Install and upgrade both use `add dsh-vibe-math@<version>` (use `add`, not `update`).
+
 Then start a new session and pick **Vibe Math V2** (v2, classic), **Vibe Math V3** (v3, classic), **Vibe Math V4** (v4, resident self-organization) or **Vibe Math V5** (v5, institute system) in the preset picker — the four architectures are peers, choose according to your actual needs (see "How to choose").
 **The two DSH generations land in different places, and this package adapts to both**:
 
@@ -199,7 +207,14 @@ Then start a new session and select **"Vibe Math V2"** / **"Vibe Math V3"** / **
 - **Supported range**: `0.1.2-alpha.4` … `0.2.0-rc.2` (11 versions declared `compatible`, tested target `0.2.0-rc.2`; `engines.node` is `^22.19.0 || >=24.0.0`) — per-version declarations, the `engines.dsh` range and `peerDependencies` live in `package.json` (`dsh.compatibility.dshReleases`).
 - **Delivery form**: DSH ≥ 0.1.7 uses **composition-row declarations** (the `agentPresets` service); DSH ≤ 0.1.6 uses the `~/.dsh/.agent-presets/<id>/` directory + preset picker — one bundle covers both lines.
 - **Hard constraint**: v5 keeps its institute state **only** in `State/<institute>.v5state.json` and **never** in the host session log — on an unknown event type DSH **refuses to load the whole session**, so it will not open on the next resume.
-- **Everything else** (host rows and required services, the startup self-check and capability gate, the live-resident cap, the three 2026 fixes, the `dsh.bundle.patch` constraint, the upgrade path and `peerDependencies`): see `docs/COMPAT-AUDIT-ROUND2.md`; to upgrade this package use `dsh plugin --profile <your profile> add dsh-vibe-math@latest` (use `add`, not `update`).
+- **Everything else** (host rows and required services, the startup self-check and capability gate, the live-resident cap, the three 2026 fixes, the `dsh.bundle.patch` constraint, the upgrade path and `peerDependencies`): see `docs/COMPAT-AUDIT-ROUND2.md` (install/upgrade: its §9); to upgrade this package use `dsh plugin --profile <your profile> add dsh-vibe-math@<new version>` (**pin the version**; use `add`, not `update` — measured locally: `@latest` does not upgrade).
+
+### Install/upgrade troubleshooting (pnpm and versions)
+
+- **Cannot install (`exit 1`)**: first check that `pnpm --version` runs — DSH's installer **forwards** installs to pnpm and does **not** bundle it; when it is missing the plugin market often shows only `exit 1` and no reason. Install pnpm, retry `add`, and attach `~/.dsh/profiles/<profile>/hub.log`.
+- **Installed, but an OLD version**: measured locally, **`@latest` does not upgrade either** (all four spellings failed), because the profile's `package.json` / `pnpm-lock.yaml` win over the `latest` tag. Use an **explicit version**: `dsh plugin --profile <your profile> add dsh-vibe-math@<new version>`; or set the profile dependency to `latest` and `install`.
+- **Check which version you actually have**: look at the profile's `package.json` (or run `pnpm list --depth 0` in that profile) instead of only checking that a preset appeared.
+- **After upgrading**: **restart DSH** (see above for when the managed content is replaced).
 
 ---
 
@@ -520,7 +535,7 @@ All four presets default to `finalPaper=true`: once the respective closure signa
 | `leanAsync` | `true` | all four | Lean compile mode: `true` (default) = background queue — `lean_run` / `lean_archive{run:true}` enqueue and return `async.jobId` immediately without blocking the member; `false` = synchronous await (the previous semantics, verbatim) |
 | `leanJobsMaxParallel` | `1` | all four | Background-compile concurrency cap (default 1 = serial; raise it to compile in parallel) |
 | `leanInitiative` | `'normal'` | all four | **Daily formalization eagerness**: `off` / `normal` (default, formalize along the way) / `eager` (more proactive). A **separate knob** from `formalVerify` (which only says how strong verification must be) |
-| `leanSearchPaths` | `[]` | all four | Extra Lean search roots (injected first; the automatic `<VibeMath root>` is added after them; nothing is injected when `leanArgs` already sets `--search-path`/`-R`/`--root`) |
+| `leanSearchPaths` | `[]` | all four | Extra Lean search roots (injected first; the automatic `<VibeMath root>` is added after them; nothing is injected when `leanArgs` already sets `-R`/`--root`) |
 | `finalPaper` | `true` | all four | **Final paper**: written automatically at closure (`false` disables only the automatic trigger; the manual command still works). Full contract: `docs/final-paper.md` |
 | `paperFormat` | `both` | all four | Which text versions to produce: `both` (md+tex) / `md` (skips compilation and must not warn about a missing tex) / `tex` |
 | `paperLanguage` | `zh` | all four | Paper language: `zh` (ctexart, engine prefers xelatex) / `en` (article, pdflatex first) |

package/README.md

@@ -168,14 +168,22 @@ flowchart TB
 
 ### 方式 A：作为插件包一键安装（推荐，同时装出四个预设）
 
-桌面版在「设置 → 插件」里安装 `dsh-vibe-math`；命令行 profile 用：
+桌面版在「设置 → 插件」里安装 `dsh-vibe-math`；命令行 profile 用（**务必带版本号**）：
 
 ```sh
-dsh plugin --profile <你的 profile> add dsh-vibe-math
+# 前置：DSH 的插件安装会调用 pnpm（它不自带 pnpm）——先确认它可用：
+pnpm --version
+
+# 安装本包（带显式版本；不带版本可能停在旧版，原因见下）：
+dsh plugin --profile <你的 profile> add dsh-vibe-math@<版本>
 # 或从 GitHub 直装：
 dsh plugin --profile <你的 profile> add github:ChongCyrus/Vibe-Mathematics
 ```
 
+> **前置**：`dsh plugin …` 需要一个可用的 **`pnpm`**（DSH 把安装**转交 pnpm**，且**不自带**它）。缺失时命令以 **`exit 1`** 结束，而**插件市场可能不显示原因** ⇒ 请附 `~/.dsh/profiles/<profile>/hub.log`（DSH 的安装日志）来定位。
+>
+> **为什么示例都带 `@<版本>`**：pnpm 优先按 profile 的 `package.json` 与 `pnpm-lock.yaml` 解析裸包名，**不带版本可能装到旧版**；本机实测 `@latest`、`@^2`、`pnpm update --latest`、`pnpm add …@latest` **四种写法都拿不到新版**，**只有显式版本**（如 `@2.8.5`）会真的升级。安装与升级都用 `add dsh-vibe-math@<版本>`（用 `add` 而非 `update`）。
+
 之后新建会话，预设选择器里选择 **Vibe Math V2**（v2，经典）、**Vibe Math V3**（v3，经典）、**Vibe Math V4**（v4，常驻自组织）或 **Vibe Math V5**（v5，研究所体系）即可——四个架构同级，按实际需求自选（见「怎么选」）。
 **两个 DSH 世代的落点不同，本包自动适配**：
 
@@ -199,7 +207,14 @@ dsh plugin --profile <你的 profile> add github:ChongCyrus/Vibe-Mathematics
 - **支持范围**：`0.1.2-alpha.4` … `0.2.0-rc.2`（11 个版本声明为 `compatible`，实测目标 `0.2.0-rc.2`；`engines.node` 为 `^22.19.0 || >=24.0.0`）——逐版本声明、`engines.dsh` 区间与 `peerDependencies` 见 `package.json`（`dsh.compatibility.dshReleases`）。
 - **交付形态**：DSH ≥ 0.1.7 走**组合行声明**（`agentPresets` 服务），DSH ≤ 0.1.6 走 `~/.dsh/.agent-presets/<id>/` 目录 + preset picker——一份 bundle 同时兼容两条线。
 - **硬约束**：v5 研究所状态**只**写 `State/<研究所>.v5state.json`，**绝不**写宿主会话日志——DSH 遇到日志里不认识的事件类型会**拒绝加载整个会话**，会话下次恢复将打不开。
-- **其余细节**（宿主插件行与必需服务、启动自检与能力门槛、常驻数量上限、2026 的三处修复、`dsh.bundle.patch` 约束、升级路径与 `peerDependencies`）：见 `docs/COMPAT-AUDIT-ROUND2.md`；升级本包用 `dsh plugin --profile <你的 profile> add dsh-vibe-math@latest`（用 `add` 而非 `update`）。
+- **其余细节**（宿主插件行与必需服务、启动自检与能力门槛、常驻数量上限、2026 的三处修复、`dsh.bundle.patch` 约束、升级路径与 `peerDependencies`）：见 `docs/COMPAT-AUDIT-ROUND2.md`（安装与升级见其 §9）；升级本包用 `dsh plugin --profile <你的 profile> add dsh-vibe-math@<新版本>`（**给显式版本**，用 `add` 而非 `update`——本机实测 `@latest` 不升级）。
+
+### 安装/升级排查（pnpm 与版本）
+
+- **装不上（`exit 1`）**：先确认 `pnpm --version` 能跑——DSH 的安装器把安装**转交 pnpm** 且**不自带**它；缺失时插件市场里往往只有 `exit 1`、没有原因。装好 pnpm 后重试 `add`，并附 `~/.dsh/profiles/<profile>/hub.log`。
+- **装上了但是旧版**：本机实测**带 `@latest` 也不升级**（四种写法都无效），因为 profile 的 `package.json` / `pnpm-lock.yaml` 优先于 `latest` 标签。改用**显式版本**：`dsh plugin --profile <你的 profile> add dsh-vibe-math@<新版本>`；或先把 profile 依赖改成 `latest` 再 `install`。
+- **确认实际版本**：看 profile 的 `package.json`（或在该 profile 下 `pnpm list --depth 0`），别只看预设选择器里有没有 preset。
+- **升级之后**：**重启 DSH**（受管内容的替换时机见上文）。
 
 ---
 
@@ -520,7 +535,7 @@ v5 的完整架构（成员生命周期、一轮时序、共识状态机、会
 | `leanAsync` | `true` | 四套 | Lean 编译模式：`true`（默认）= 后台队列，`lean_run` / `lean_archive{run:true}` 入队即返回 `async.jobId`，成员不阻塞；`false` = 同步 await（旧语义逐字保留） |
 | `leanJobsMaxParallel` | `1` | 四套 | 后台编译并发上限（默认 1 = 串行；调大即并行） |
 | `leanInitiative` | `'normal'` | 四套 | **日常形式化主动性**：`off` / `normal`（默认，顺手形式化）/ `eager`（更主动）。与 `formalVerify`（验证时的要求强度）**是两件事** |
-| `leanSearchPaths` | `[]` | 四套 | 额外 Lean 搜索根（先注入它们、再注入自动的 `<VibeMath 根>`；`leanArgs` 里已有 `--search-path`/`-R`/`--root` 时不注入） |
+| `leanSearchPaths` | `[]` | 四套 | 额外 Lean 搜索根（先注入它们、再注入自动的 `<VibeMath 根>`；`leanArgs` 里已有 `-R`/`--root` 时不注入） |
 | `finalPaper` | `true` | 四套 | **最终论文**：收口时自动撰写（`false` 只关自动触发，手动命令仍可用）。完整契约见 `docs/final-paper.md` |
 | `paperFormat` | `both` | 四套 | 论文产出格式：`both`（md+tex）/ `md`（跳过编译，且不报"缺 tex"）/ `tex` |
 | `paperLanguage` | `zh` | 四套 | 论文语言：`zh`（ctexart，引擎优先 xelatex）/ `en`（article，pdflatex 优先） |

package/cordis.patch.yml

@@ -84,7 +84,7 @@
 
                 - math_computation {op: probe|run|receipt|install, engine, mode: code|file|expr, …} — 先 probe 预检引擎/包/许可，再调引擎计算并把脚本与输出归档成可复核回执；缺引擎/缺包只报告与给安装指引/计划；shell 兜底不算归档。
                 - 替代必须声明（诚实性）：当替代方案改变了**精确性或结论强度**时（精确符号解 → 数值近似、闭式解 → 采样/求积、改了精度/容差/假设、换了算法类），结论里**必须写明**，不得读起来像得到了原本（精确/所要求的）结果；拿不到精确结果就直说。
-                - 归档→编辑→重跑：mode:'code' 的脚本原件在回执的 scriptPath（Computation/<id>/script.<ext>，**相对项目根**）；**成员的文件工具是按会话 cwd 解析的**，所以读它要用**绝对路径** `receipt.scriptAbs`，或把 `receipt.cwd` 与 `receipt.scriptPath` 拼起来（回执两个字段都有）。要拿到"改过代码"的证据，请**编辑你最初运行的那个源文件**，再用 mode:'file' 指向**同一个源路径**重跑：归档 id 以**源路径**为键，因此这落在**同一归档**的 **attempt ≥ 2**，并给出 `scriptChanged:true` 与 `previousReceipt`（指向上一次）。**指向归档副本本身**（`receipt.scriptAbs` 那个路径）按设计是**另一份新归档**：新 id、attempt 1、没有 `previousReceipt`、`scriptChanged:false`；旧 attempt 绝不会被覆盖，但它**不是**"同一归档的新 attempt"——工具会用 `fileIsArchivedScript` 与 `ARCHIVED_SCRIPT_RERUN` 警告明确说明。**旧回执对修改后的代码无效**——报告里必须引用与当前代码哈希一致的那份回执；工具会在 scriptChanged / scriptChangedDuringRun 为 true 时显式告警。
+                - 归档→编辑→重跑（**回执＝一次 math_computation 调用的 JSON 结果**：先 probe 或 run 一次，它有哪些字段就一目了然）：mode:'code' 的脚本原件在回执的 scriptPath（Computation/<id>/script.<ext>，**相对项目根**）；**成员的文件工具是按会话 cwd 解析的**，所以读它要用**绝对路径** `receipt.scriptAbs`，或把 `receipt.cwd` 与 `receipt.scriptPath` 拼起来（回执两个字段都有）。要拿到"改过代码"的证据，请**编辑你最初运行的那个源文件**，再用 mode:'file' 指向**同一个源路径**重跑：归档 id 以**源路径**为键，因此这落在**同一归档**的 **attempt ≥ 2**，并给出 `scriptChanged:true`（脚本内容相对上一份回执变过）与 `previousReceipt`（指向上一次）。**指向归档副本本身**（`receipt.scriptAbs` 那个路径）按设计是**另一份新归档**：新 id、attempt 1、没有 `previousReceipt`、`scriptChanged:false`；旧 attempt 绝不会被覆盖，但它**不是**"同一归档的新 attempt"——工具会用 `fileIsArchivedScript` 与 `ARCHIVED_SCRIPT_RERUN` 警告明确说明。**旧回执对修改后的代码无效**——报告里必须引用与当前代码哈希一致的那份回执；工具会在 scriptChanged / scriptChangedDuringRun 为 true 时显式告警。
 
                 Final paper: strict termination writes Paper/<project>/{paper.md,paper.tex,paper.pdf,paper.meta.json,paper.log.md}
                 automatically — a dedicated paper-writing child assembles the solved material (finalPaper=false disables only that
@@ -404,7 +404,7 @@
 
                 - math_computation {op: probe|run|receipt|install, engine, mode: code|file|expr, …} — 先 probe 预检引擎/包/许可，再调引擎计算并把脚本与输出归档成可复核回执；缺引擎/缺包只报告与给安装指引/计划；shell 兜底不算归档。
                 - 替代必须声明（诚实性）：当替代方案改变了**精确性或结论强度**时（精确符号解 → 数值近似、闭式解 → 采样/求积、改了精度/容差/假设、换了算法类），结论里**必须写明**，不得读起来像得到了原本（精确/所要求的）结果；拿不到精确结果就直说。
-                - 归档→编辑→重跑：mode:'code' 的脚本原件在回执的 scriptPath（Computation/<id>/script.<ext>，**相对项目根**）；**成员的文件工具是按会话 cwd 解析的**，所以读它要用**绝对路径** `receipt.scriptAbs`，或把 `receipt.cwd` 与 `receipt.scriptPath` 拼起来（回执两个字段都有）。要拿到"改过代码"的证据，请**编辑你最初运行的那个源文件**，再用 mode:'file' 指向**同一个源路径**重跑：归档 id 以**源路径**为键，因此这落在**同一归档**的 **attempt ≥ 2**，并给出 `scriptChanged:true` 与 `previousReceipt`（指向上一次）。**指向归档副本本身**（`receipt.scriptAbs` 那个路径）按设计是**另一份新归档**：新 id、attempt 1、没有 `previousReceipt`、`scriptChanged:false`；旧 attempt 绝不会被覆盖，但它**不是**"同一归档的新 attempt"——工具会用 `fileIsArchivedScript` 与 `ARCHIVED_SCRIPT_RERUN` 警告明确说明。**旧回执对修改后的代码无效**——报告里必须引用与当前代码哈希一致的那份回执；工具会在 scriptChanged / scriptChangedDuringRun 为 true 时显式告警。
+                - 归档→编辑→重跑（**回执＝一次 math_computation 调用的 JSON 结果**：先 probe 或 run 一次，它有哪些字段就一目了然）：mode:'code' 的脚本原件在回执的 scriptPath（Computation/<id>/script.<ext>，**相对项目根**）；**成员的文件工具是按会话 cwd 解析的**，所以读它要用**绝对路径** `receipt.scriptAbs`，或把 `receipt.cwd` 与 `receipt.scriptPath` 拼起来（回执两个字段都有）。要拿到"改过代码"的证据，请**编辑你最初运行的那个源文件**，再用 mode:'file' 指向**同一个源路径**重跑：归档 id 以**源路径**为键，因此这落在**同一归档**的 **attempt ≥ 2**，并给出 `scriptChanged:true`（脚本内容相对上一份回执变过）与 `previousReceipt`（指向上一次）。**指向归档副本本身**（`receipt.scriptAbs` 那个路径）按设计是**另一份新归档**：新 id、attempt 1、没有 `previousReceipt`、`scriptChanged:false`；旧 attempt 绝不会被覆盖，但它**不是**"同一归档的新 attempt"——工具会用 `fileIsArchivedScript` 与 `ARCHIVED_SCRIPT_RERUN` 警告明确说明。**旧回执对修改后的代码无效**——报告里必须引用与当前代码哈希一致的那份回执；工具会在 scriptChanged / scriptChangedDuringRun 为 true 时显式告警。
 
                 A /vibe slash command mirrors the main controls (/vibe start [override]|resume [override]|pause|abort|status|report|mode
                 <auto|manual>|setup|save|template [global|project]|add <id> <desc>|add-proposition <id> <概述>|list-propositions|methods|index|plan|lock|project [list|new <name>|<name>]|decisions|agents|paper [lang=zh|en] [format=both|md|tex] [force]).
@@ -722,8 +722,8 @@
                     too; they work in every mode.
                   - vibe_v4_formal_report — human-readable Lean formal-verification mirror (mode, Lean-passed
                   - math_computation {op: probe|run|receipt|install, engine, mode: code|file|expr, …} — 先 probe 预检引擎/包/许可，再调引擎计算并把脚本与输出归档成可复核回执；缺引擎/缺包只报告与给安装指引/计划；shell 兜底不算归档。
-                  - 归档→编辑→重跑：mode:'code' 的脚本原件在回执的 scriptPath（Computation/<id>/script.<ext>，**相对项目根**）；**成员的文件工具是按会话 cwd 解析的**，所以读它要用**绝对路径** `receipt.scriptAbs`，或把 `receipt.cwd` 与 `receipt.scriptPath` 拼起来（回执两个字段都有）。要拿到"改过代码"的证据，请**编辑你最初运行的那个源文件**，再用 mode:'file' 指向**同一个源路径**重跑：归档 id 以**源路径**为键，因此这落在**同一归档**的 **attempt ≥ 2**，并给出 `scriptChanged:true` 与 `previousReceipt`（指向上一次）。**指向归档副本本身**（`receipt.scriptAbs` 那个路径）按设计是**另一份新归档**：新 id、attempt 1、没有 `previousReceipt`、`scriptChanged:false`；旧 attempt 绝不会被覆盖，但它**不是**"同一归档的新 attempt"——工具会用 `fileIsArchivedScript` 与 `ARCHIVED_SCRIPT_RERUN` 警告明确说明。**旧回执对修改后的代码无效**——报告里必须引用与当前代码哈希一致的那份回执；工具会在 scriptChanged / scriptChangedDuringRun 为 true 时显式告警。
-                  - Archive -> edit -> re-run: for mode:'code' the script original is at the receipt's scriptPath (Computation/<id>/script.<ext>, **relative to the project root**); member file tools resolve paths against the SESSION CWD, so READ it via the ABSOLUTE `receipt.scriptAbs`, or join `receipt.cwd` with `receipt.scriptPath` (both are in the receipt). To produce evidence for EDITED code, edit the SOURCE FILE you originally ran and re-run mode:'file' pointing at THAT SAME PATH: the archive id is keyed by the SOURCE PATH, so this lands on the SAME archive as attempt >= 2, with `scriptChanged:true` and `previousReceipt` pointing at the previous attempt. Pointing mode:'file' at the ARCHIVED COPY itself (the `receipt.scriptAbs` path) is a DIFFERENT archive BY DESIGN - a new id, attempt 1, no `previousReceipt`, `scriptChanged:false`: the earlier attempt is never overwritten, but it is NOT "a new attempt of the same archive", and the tool says so via `fileIsArchivedScript` and the `ARCHIVED_SCRIPT_RERUN` warning. **An old receipt is NOT evidence for edited code** - cite the receipt whose scriptHash matches the current code; the tool warns explicitly via scriptChanged / scriptChangedDuringRun.
+                  - 归档→编辑→重跑（**回执＝一次 math_computation 调用的 JSON 结果**：先 probe 或 run 一次，它有哪些字段就一目了然）：mode:'code' 的脚本原件在回执的 scriptPath（Computation/<id>/script.<ext>，**相对项目根**）；**成员的文件工具是按会话 cwd 解析的**，所以读它要用**绝对路径** `receipt.scriptAbs`，或把 `receipt.cwd` 与 `receipt.scriptPath` 拼起来（回执两个字段都有）。要拿到"改过代码"的证据，请**编辑你最初运行的那个源文件**，再用 mode:'file' 指向**同一个源路径**重跑：归档 id 以**源路径**为键，因此这落在**同一归档**的 **attempt ≥ 2**，并给出 `scriptChanged:true`（脚本内容相对上一份回执变过）与 `previousReceipt`（指向上一次）。**指向归档副本本身**（`receipt.scriptAbs` 那个路径）按设计是**另一份新归档**：新 id、attempt 1、没有 `previousReceipt`、`scriptChanged:false`；旧 attempt 绝不会被覆盖，但它**不是**"同一归档的新 attempt"——工具会用 `fileIsArchivedScript` 与 `ARCHIVED_SCRIPT_RERUN` 警告明确说明。**旧回执对修改后的代码无效**——报告里必须引用与当前代码哈希一致的那份回执；工具会在 scriptChanged / scriptChangedDuringRun 为 true 时显式告警。
+                  - Archive -> edit -> re-run (a RECEIPT is the JSON result of ONE math_computation call - run `probe` or `run` once and the fields below are right there in front of you): for mode:'code' the script original is at the receipt's scriptPath (Computation/<id>/script.<ext>, **relative to the project root**); member file tools resolve paths against the SESSION CWD, so READ it via the ABSOLUTE `receipt.scriptAbs`, or join `receipt.cwd` with `receipt.scriptPath` (both are in the receipt). To produce evidence for EDITED code, edit the SOURCE FILE you originally ran and re-run mode:'file' pointing at THAT SAME PATH: the archive id is keyed by the SOURCE PATH, so this lands on the SAME archive as attempt >= 2, with `scriptChanged:true` (the script content differs from the previous receipt) and `previousReceipt` pointing at the previous attempt. Pointing mode:'file' at the ARCHIVED COPY itself (the `receipt.scriptAbs` path) is a DIFFERENT archive BY DESIGN - a new id, attempt 1, no `previousReceipt`, `scriptChanged:false`: the earlier attempt is never overwritten, but it is NOT "a new attempt of the same archive", and the tool says so via `fileIsArchivedScript` and the `ARCHIVED_SCRIPT_RERUN` warning. **An old receipt is NOT evidence for edited code** - cite the receipt whose scriptHash matches the current code; the tool warns explicitly via scriptChanged / scriptChangedDuringRun.
                   - 替代必须声明（诚实性）：当替代方案改变了**精确性或结论强度**时（精确符号解 → 数值近似、闭式解 → 采样/求积、改了精度/容差/假设、换了算法类），结论里**必须写明**，不得读起来像得到了原本（精确/所要求的）结果；拿不到精确结果就直说。
                   - Declare substitutions (honesty): when an alternative changes EXACTNESS or conclusion strength (exact symbolic solution -> numerical approximation, closed form -> sampling/quadrature, changed precision/tolerances/assumptions, a different algorithm class), the conclusion MUST say so explicitly and must not read as if the original (exact/requested) result had been obtained; if the exact result is unavailable, say so plainly.
                     objects, recorded blockers, formalization TODO, library paths).
@@ -768,7 +768,7 @@
                     `<project root>/Formal/<id>.lean`.
                   - Reuse FIRST: call vibe_v4_lean_lib (and vibe_v4_lean_read for the verbatim text) BEFORE writing a new
                     definition; reuse an archived file with `import Formal.Lib.<name>` / `import Formal.Proved.<name>`
-                    (the framework passes --search-path <VibeMath root>, so VibeMath is the module root).
+                    (the framework passes -R <VibeMath root>, so VibeMath is the module root).
                   - Choose WHAT to formalize with three criteria: (1) valuable or likely reusable, (2) rather key or
                     necessary, (3) you are confident in the statement — when you are NOT confident, record blocked
                     with the difficulty instead of hiding uncertainty behind a formalization.
@@ -1085,8 +1085,8 @@
                     verification (execute / archive / list the reuse library / read an archived file
                     verbatim / watch the background compile queue). Members use them too; they work in every mode.
                   - math_computation {op: probe|run|receipt|install, engine, mode: code|file|expr, …} — 先 probe 预检引擎/包/许可，再调引擎计算并把脚本与输出归档成可复核回执；缺引擎/缺包只报告与给安装指引/计划；shell 兜底不算归档。
-                  - 归档→编辑→重跑：mode:'code' 的脚本原件在回执的 scriptPath（Computation/<id>/script.<ext>，**相对项目根**）；**成员的文件工具是按会话 cwd 解析的**，所以读它要用**绝对路径** `receipt.scriptAbs`，或把 `receipt.cwd` 与 `receipt.scriptPath` 拼起来（回执两个字段都有）。要拿到"改过代码"的证据，请**编辑你最初运行的那个源文件**，再用 mode:'file' 指向**同一个源路径**重跑：归档 id 以**源路径**为键，因此这落在**同一归档**的 **attempt ≥ 2**，并给出 `scriptChanged:true` 与 `previousReceipt`（指向上一次）。**指向归档副本本身**（`receipt.scriptAbs` 那个路径）按设计是**另一份新归档**：新 id、attempt 1、没有 `previousReceipt`、`scriptChanged:false`；旧 attempt 绝不会被覆盖，但它**不是**"同一归档的新 attempt"——工具会用 `fileIsArchivedScript` 与 `ARCHIVED_SCRIPT_RERUN` 警告明确说明。**旧回执对修改后的代码无效**——报告里必须引用与当前代码哈希一致的那份回执；工具会在 scriptChanged / scriptChangedDuringRun 为 true 时显式告警。
-                  - Archive -> edit -> re-run: for mode:'code' the script original is at the receipt's scriptPath (Computation/<id>/script.<ext>, **relative to the project root**); member file tools resolve paths against the SESSION CWD, so READ it via the ABSOLUTE `receipt.scriptAbs`, or join `receipt.cwd` with `receipt.scriptPath` (both are in the receipt). To produce evidence for EDITED code, edit the SOURCE FILE you originally ran and re-run mode:'file' pointing at THAT SAME PATH: the archive id is keyed by the SOURCE PATH, so this lands on the SAME archive as attempt >= 2, with `scriptChanged:true` and `previousReceipt` pointing at the previous attempt. Pointing mode:'file' at the ARCHIVED COPY itself (the `receipt.scriptAbs` path) is a DIFFERENT archive BY DESIGN - a new id, attempt 1, no `previousReceipt`, `scriptChanged:false`: the earlier attempt is never overwritten, but it is NOT "a new attempt of the same archive", and the tool says so via `fileIsArchivedScript` and the `ARCHIVED_SCRIPT_RERUN` warning. **An old receipt is NOT evidence for edited code** - cite the receipt whose scriptHash matches the current code; the tool warns explicitly via scriptChanged / scriptChangedDuringRun.
+                  - 归档→编辑→重跑（**回执＝一次 math_computation 调用的 JSON 结果**：先 probe 或 run 一次，它有哪些字段就一目了然）：mode:'code' 的脚本原件在回执的 scriptPath（Computation/<id>/script.<ext>，**相对项目根**）；**成员的文件工具是按会话 cwd 解析的**，所以读它要用**绝对路径** `receipt.scriptAbs`，或把 `receipt.cwd` 与 `receipt.scriptPath` 拼起来（回执两个字段都有）。要拿到"改过代码"的证据，请**编辑你最初运行的那个源文件**，再用 mode:'file' 指向**同一个源路径**重跑：归档 id 以**源路径**为键，因此这落在**同一归档**的 **attempt ≥ 2**，并给出 `scriptChanged:true`（脚本内容相对上一份回执变过）与 `previousReceipt`（指向上一次）。**指向归档副本本身**（`receipt.scriptAbs` 那个路径）按设计是**另一份新归档**：新 id、attempt 1、没有 `previousReceipt`、`scriptChanged:false`；旧 attempt 绝不会被覆盖，但它**不是**"同一归档的新 attempt"——工具会用 `fileIsArchivedScript` 与 `ARCHIVED_SCRIPT_RERUN` 警告明确说明。**旧回执对修改后的代码无效**——报告里必须引用与当前代码哈希一致的那份回执；工具会在 scriptChanged / scriptChangedDuringRun 为 true 时显式告警。
+                  - Archive -> edit -> re-run (a RECEIPT is the JSON result of ONE math_computation call - run `probe` or `run` once and the fields below are right there in front of you): for mode:'code' the script original is at the receipt's scriptPath (Computation/<id>/script.<ext>, **relative to the project root**); member file tools resolve paths against the SESSION CWD, so READ it via the ABSOLUTE `receipt.scriptAbs`, or join `receipt.cwd` with `receipt.scriptPath` (both are in the receipt). To produce evidence for EDITED code, edit the SOURCE FILE you originally ran and re-run mode:'file' pointing at THAT SAME PATH: the archive id is keyed by the SOURCE PATH, so this lands on the SAME archive as attempt >= 2, with `scriptChanged:true` (the script content differs from the previous receipt) and `previousReceipt` pointing at the previous attempt. Pointing mode:'file' at the ARCHIVED COPY itself (the `receipt.scriptAbs` path) is a DIFFERENT archive BY DESIGN - a new id, attempt 1, no `previousReceipt`, `scriptChanged:false`: the earlier attempt is never overwritten, but it is NOT "a new attempt of the same archive", and the tool says so via `fileIsArchivedScript` and the `ARCHIVED_SCRIPT_RERUN` warning. **An old receipt is NOT evidence for edited code** - cite the receipt whose scriptHash matches the current code; the tool warns explicitly via scriptChanged / scriptChangedDuringRun.
                   - 替代必须声明（诚实性）：当替代方案改变了**精确性或结论强度**时（精确符号解 → 数值近似、闭式解 → 采样/求积、改了精度/容差/假设、换了算法类），结论里**必须写明**，不得读起来像得到了原本（精确/所要求的）结果；拿不到精确结果就直说。
                   - Declare substitutions (honesty): when an alternative changes EXACTNESS or conclusion strength (exact symbolic solution -> numerical approximation, closed form -> sampling/quadrature, changed precision/tolerances/assumptions, a different algorithm class), the conclusion MUST say so explicitly and must not read as if the original (exact/requested) result had been obtained; if the exact result is unavailable, say so plainly.
                 A /v5 slash command mirrors these (configure|start|resume|pause|stop|status|report|members|message|meeting|hire|fire|add|remove|set|paper).
@@ -1116,7 +1116,7 @@
                     small lemmas/propositions/definitions into the library even in ordinary rounds).
                   - Reuse first: check vibe_v5_lean_lib BEFORE writing a definition, reuse archived
                     content via `import Formal.Lib.<name>` / `import Formal.Proved.<name>` (the compile
-                    gets `--search-path <VibeMath root>`, preceded by any leanSearchPaths), or copy the
+                    gets `-R <VibeMath root>`, preceded by any leanSearchPaths), or copy the
                     exact text with vibe_v5_lean_read. Re-archiving identical content is de-duplicated.
                   - Compiles are ASYNC by default (leanAsync=true, leanJobsMaxParallel default 1):
                     vibe_v5_lean_run / vibe_v5_lean_archive enqueue and return at once; watch them with

package/docs/COMPAT-AUDIT-ROUND2.md

@@ -352,3 +352,30 @@ e2e-f1-agent-detach    CONFIRMED (drives the real AgentRegistry)
 
 - 本文件的 **`### 未修（需你决策，风险与取舍已说明）`** 一节（F-1…F-11；F-7 非缺陷、F-8 已知限制、F-9 已有 `tests/audit-corrupt-file-guard.test.mjs`、F-10/F-11 待定、**F-4c 已关闭**）是唯一权威的"未修 / 需决策"清单。
 - **按标题文本定位，不写行号**：行号会随编辑漂移，标题是稳定的；链接目标在本仓库内，且本文件随包发布（`files[]`），因此指针不会指向包外。
+
+## 9. 安装与升级（插件市场 · pnpm 前置 · 显式版本）
+
+**背景**：issue #10 报告"经 DSH 插件市场安装失败（`exit 1`）"。复核结论：**两条根因都不在本插件包内**（不在本包的代码、参数或提示词里），但任何一条都会让用户卡住，所以写进本文档。
+
+### 9.1 前置：安装需要一个可用的 `pnpm`
+
+DSH 的插件管理器把安装**转交 pnpm**，并且**不自带 pnpm**：`@deepseek-ai/dsh-plugin-manager/lib/index.js` 多处 `execa(options.command ?? "pnpm", …)`（L324/517/650/694/733）、`pnpmCommand` 默认 `"pnpm"`（L1354）、`bin.js:23` 注释 "forwarding to pnpm"；`dsh/node_modules` 中含 `pnpm` 的条目为空。报告者环境 `pnpm: unknown` / `npm: unknown` ⇒ 必然失败。
+
+- **复现（本机）**：把 PATH 里的 `pnpm` 剥掉后执行安装 ⇒ 同一条报错 ＋ `exit 1`；恢复 `pnpm` 后同一命令成功。
+- **表现与取证**：失败时插件市场 UI 可能**不显示原因**，只有一个 `exit 1`。定位请附 `~/.dsh/profiles/<profile>/hub.log`（DSH 的安装日志）。
+- **修复**：装好 `pnpm`（`pnpm --version` 能跑）后重试安装。
+
+### 9.2 安装/升级：必须给**显式版本**
+
+`install-spec.js` 只**校验** spec、**不注入版本**；裸包名交给 pnpm 后按 profile 的 `package.json`（首次 `add` 时由 pnpm 写下，例如 `"dsh-vibe-math": "^2.8.1"`）与 `pnpm-lock.yaml` 解析 ⇒ **可能停在旧版**。
+
+- **实测（本机）**：`@latest`、`@^2`、`pnpm update --latest`、`pnpm add …@latest` **四种写法全部无效**；**只有显式 `@2.8.4` 成功升级**。
+- **正确用法**：
+
+```sh
+pnpm --version                                              # 前置检查
+dsh plugin --profile <profile> add dsh-vibe-math@<版本>     # 安装与升级都写显式版本
+```
+
+- **其他排查**：怀疑装成旧版时，可先把 profile 依赖改成 `latest` 再 `install`；升级后**重启 DSH**（受管内容的替换时机见本文件其余章节与 README 安装段）。
+- **交叉引用**：README（中/英）安装段与 troubleshooting 小节给出同样的前置与显式版本要求，并由 `tests/audit-artifact-docs.mjs` 第 5 节静态守卫（含单点具名变异）。

package/docs/final-paper.md

@@ -125,6 +125,19 @@
   `consult = {messages: 0, meetings: 0}` 开始，**重启前的消息与会议不再计入**（这是设计，不是丢状态 ✗）。
   因此 `paperEditor="office"` 在 `force` 之后必须**重新**征询；`V5_PAPER_CONSULT_REQUIRED` 的响应会**具名**说明
   这一点（`restarted: true` ＋ `forceReason`），不再让人靠猜。
+- **会议何时会被放弃（看门狗）**：会议停滞超过 `2 × activityTimeoutMs`（默认 **240 s**）**且**没有任何被点名的成员
+  仍在进行中时，才由看门狗放弃 ✗；若仍有被点名的成员在飞，则宽限到 **3 倍**（默认 720 s）再放弃 —— 成员的**一整轮**
+  在真机上很容易超过软阈值，若此时就放弃，其发言只能退化成「会后补记」✗（实测一次真机运行出现 **1067** 次补记、
+  **863** 次定稿被拒 ✓）。放弃会写进纪要（`⚠ 本次会议因长时间无新发言而被放弃`）并在群聊具名一行 ✓；放弃**之后**
+  到达的发言不会被丢弃，而是作为「**会后补记**」追加进纪要 ✓（见 §7 上一条）。
+- **稿件（`paper_part`/`paper_review`）由论文阶段的专属轮次征集**：提示词只在这些轮次给出 schema（
+  `paperPartPrompt` / `paperReviewPrompt`），而**会议轮只征集 `input`（发言）**。采集端对轮次类型**不设限** ——
+  会议轮里附带 `paper_part` 框架也会收 ✓ —— 但**不要指望用会议轮交稿**：会议轮的提示词不会列出稿件字段 ✓
+  （实测某次真机复测里，成员在会议轮连续"空交"三次并空转 ✗ ⇒ 这是**阶段边界**，不是丢稿 ✗）。
+- **`consult` 的两个计数只在"真的发生"时增加**：所办消息在**实际送达**后计（`notePaperConsult('message')`），
+  会议在**实际开始**且由所办召集时计（`notePaperConsult('meeting')`）。**因此**：被挂起（`pendingMeeting`）或被拒绝的
+  会议请求**在开始之前不计入** ✓ —— 这是设计（"记录在案"才算征询 ✓），不是计数滞后 ✗；如果你在这一窗口内定稿，
+  会如实被 `V5_PAPER_CONSULT_REQUIRED` 挡住，直到那场会议真正开起来 ✓。
 
 ## 8. LaTeX 检测、编译与降级
 

package/docs/formal-verification.md

@@ -60,7 +60,7 @@
 | `leanAsync` | 布尔 | `true` | `true` = 编译走**后台队列**（`lean_run` / `lean_archive{run:true}` 入队后立即返回 `{async:{jobId,state:'queued'}}`）；`false` = **完全同步 await**（旧路径逐字保留）。只有 `settled(ok)` 才可能置 `passed`，见 §7.3 |
 | `leanJobsMaxParallel` | 正整数 | `1` | 后台编译并发上限（默认 1 = 串行；调大即并行，v2/v3 把合法范围夹在 1–8） |
 | `leanInitiative` | `'off'` \| `'normal'` \| `'eager'` | `'normal'` | **日常流程中的形式化主动性**：`off` 不主动（只在验证提示词里按 `formalVerify` 的要求做）/ `normal` 顺手把有价值且可能复用的东西形式化 / `eager` 更主动。**与 `formalVerify` 是两件事**：它不改变定论门禁 |
-| `leanSearchPaths` | 字符串数组 | `[]` | 额外编译搜索根：先注入它们、再注入自动的 `<VibeMath 根>`（去重、保持用户给的顺序）；`leanArgs` 里已显式给 `--search-path`/`-R`/`--root` 时**不注入任何东西**，见 §7.6 |
+| `leanSearchPaths` | 字符串数组 | `[]` | 额外编译搜索根：先注入它们、再注入自动的 `<VibeMath 根>`（去重、保持用户给的顺序）；`leanArgs` 里已显式给 `-R`/`--root` 时**不注入任何东西**，见 §7.6 |
 
 - 非法值一律**回退到默认**（`formalVerify` 非三档之一 → `'off'`；`leanInitiative` 非三档之一 → `'normal'`；
   `leanTimeoutMs` 非正 → 默认；`leanJobsMaxParallel` < 1 → 1；`leanSearchPaths` 非数组 → `[]` 并去空串/去重）。
@@ -438,18 +438,18 @@
 并把作业标为 `interrupted`、把对象停在 `attempted`——**绝不因为"进程已经不在"而当作通过**。
 `interrupted` 的作业可以重跑（重跑会 `attempts+1`）。
 
-### 7.6 `--search-path` 注入（四套一致）
+### 7.6 `-R` 注入（四套一致）
 
 编译器 argv 固定为：
 
 ```
-[exe, ...用户 leanArgs, --search-path <VibeMath 根>, <file>]
+[exe, ...用户 leanArgs, -R <VibeMath 根>, <file>]
 ```
 
 - 自动根**只有一个**：`<VibeMath 根>`（`import Formal.Lib.<name>` 的模块根），插在
-  **用户参数之后、文件名之前**——`lake env lean` 因此自然变成 `lake env lean --search-path <root> <file>`。
+  **用户参数之后、文件名之前**——`lake env lean` 因此自然变成 `lake env lean -R <root> <file>`。
 - `leanSearchPaths` 非空时，**先注入用户给的路径（按给定顺序），再注入自动根**，整体去重。
-- **用户显式给出搜索根就不注入**：`leanArgs` 里已有 `--search-path` / `-R` / `--root` 时完全尊重用户配置。
+- **用户显式给出搜索根就不注入**：`leanArgs` 里已有 `-R` / `--root` 时完全尊重用户配置。
 - **不使用环境变量**：宿主的 `spawn` 没有 env 槽位，所以"注入"只能是参数注入；
   也不依赖 `LEAN_PATH` 之类的约定。
 
@@ -589,7 +589,7 @@
     恢复扫描 `Formal/Jobs/*.json`——`queued` 重入队、`running` 标记中断（哈希匹配则重入队、`attempts+1`）、
     `settled` 只在**同一构建上下文**下补写结论。
 15. **去重与搜索路径注入**：同内容 + 同构建上下文再归档 → `deduped:true` 且不重写不重编译；
-    编译器 argv 必须是 `[exe, ...用户 leanArgs, --search-path <VibeMath 根>, <file>]`，
+    编译器 argv 必须是 `[exe, ...用户 leanArgs, -R <VibeMath 根>, <file>]`，
     用户已显式给出搜索根时**不注入**，`leanSearchPaths` 先于自动根且去重。
 
 ---

package/docs/release-notes/README.en.md

@@ -8,6 +8,8 @@ Chinese counterpart of this index is [`README.md`](./README.md) (中文索引).
 
 | Version | Key change | Notes |
 |---|---|---|
+| 2.8.5 | Patch: the Lean formalization path **works again** (the search-root flag now uses Lean 4's `-R`; previously the compiler failed at argument parsing ⇒ a real-host run compiles, archives into `Verified/Lean/`, and another member reuses it); TeX engines that are installed but not on PATH are now found **within bounds** (documented locations only, no full-disk scan, no automatic install, `triedPaths` named); the prompts no longer demand impossible actions; v5 gains a **cooperation / management / work-experience feedback library** (`self`/`team` self-adjust, `interpersonal` needs an assessment plus a written-back verification; switch `feedback`); the install/upgrade docs state the `pnpm` prerequisite and an **explicit version**. | [中文](./RELEASE-NOTES-2.8.5.md) · [English](./RELEASE-NOTES-2.8.5.en.md) |
+| 2.8.4 | Patch: v5's **meeting watchdog** no longer abandons a meeting while an ASKED member is still in flight (exempt within the soft bound; the 3× hard bound still closes it), so ordinary speeches return to the ordinary sections and "late notes" become an insurance policy again — before the fix one real run showed **1067** late notes and **863** refused finalisations. | [中文](./RELEASE-NOTES-2.8.4.md) · [English](./RELEASE-NOTES-2.8.4.en.md) |
 | 2.8.3 | Patch: a speech arriving **after the meeting closed** is no longer dropped silently (it becomes a named "late note" in the minutes, which stay append-only); an office finalisation refused after a `force` restart now **explains why** (`restarted: true` + `forceReason`, stating that the counters restarted with the new paper and that a fresh consultation is required); the docs now state "`force` = a new paper". | [中文](./RELEASE-NOTES-2.8.3.md) · [English](./RELEASE-NOTES-2.8.3.en.md) |
 | 2.8.2 | Patch: an explicit `paperLatexCommand` is now probed **alone**, so `triedPaths` no longer reports impossible paths of the form "known TeX root + absolute path"; v5's delivered `paper.md` / `paper.tex` now also **each state** the compile outcome (no engine detected / compilation failed), exactly once and without duplication on re-entry — matching v4 since 2.8.1. Both have real-host reproductions and named-red guards. | [中文](./RELEASE-NOTES-2.8.2.md) · [English](./RELEASE-NOTES-2.8.2.en.md) |
 | 2.8.1 | Patch: v4's delivered `paper.md` and `paper.tex` now each state the compile outcome (no engine detected → tex+md only, or PDF compilation failed with tex/md kept), **exactly once** and idempotently — resolving known limitation #3 of the 2.8.0 notes; added `scripts/release-check.mjs`, which runs the pre-release checks as one command (version triple, section order of both notes, manifest registration, declared-but-missing = 0, repository-only files not shipped, no CRLF in the package). | [中文](./RELEASE-NOTES-2.8.1.md) · [English](./RELEASE-NOTES-2.8.1.en.md) |

package/docs/release-notes/README.md

@@ -7,6 +7,8 @@
 
 | 版本 | 关键变化 | 说明 |
 |---|---|---|
+| 2.8.5 | 补丁：Lean 形式化通道**恢复可用**（搜索根旗标改用 Lean 4 的 `-R`；此前编译器在参数解析阶段即失败 ⇒ 真机实测编译通过、归档 `Verified/Lean/` 并被另一成员复用）；"装了但不在 PATH"的 TeX 引擎现在能**有界**发现（只查文档化位置、不扫全盘、不自动安装，`triedPaths` 具名）；提示词不再要求成员做做不到的事；v5 新增**合作/管理/工作经验反馈库**（`self`/`team` 自我调整、`interpersonal` 需评估＋事后回填，开关 `feedback`）；安装/升级写明 `pnpm` 前置与**显式版本**。 | [中文](./RELEASE-NOTES-2.8.5.md) · [English](./RELEASE-NOTES-2.8.5.en.md) |
+| 2.8.4 | 补丁：v5 **会议看门狗**不再在"被点名的成员仍在进行中"时放弃会议（软界内豁免、硬界 3 倍预算仍会放弃）⇒ 正常发言回到正常小节，「会后补记」退回为保险；修复前真机实测一次运行 **1067** 次补记、**863** 次定稿被拒。 | [中文](./RELEASE-NOTES-2.8.4.md) · [English](./RELEASE-NOTES-2.8.4.en.md) |
 | 2.8.3 | 补丁：会议**收束后到达的发言**不再静默丢弃（作为具名「会后补记」写入纪要、只增不覆盖）；`force` 重启后 `paperEditor="office"` 的定稿拒绝**说明原因**（新增 `restarted:true` ＋ `forceReason`，并写明"计数随新论文重置、需重新征询"）；文档明确「`force` ＝ 一张新论文」。 | [中文](./RELEASE-NOTES-2.8.3.md) · [English](./RELEASE-NOTES-2.8.3.en.md) |
 | 2.8.2 | 补丁：显式 `paperLatexCommand` 改为**单独探测** ⇒ `triedPaths` 不再出现「已知 TeX 根 ＋ 绝对路径」这类**不可能路径**；v5 交付的 `paper.md`／`paper.tex` 也**各自写明**编译结果（未检测到引擎／编译失败），恰好一次且重入不重复 ⇒ 与 2.8.1 的 v4 对齐。两处均有真机复现读数与具名红守卫。 | [中文](./RELEASE-NOTES-2.8.2.md) · [English](./RELEASE-NOTES-2.8.2.en.md) |
 | 2.8.1 | 补丁：v4 交付的 `paper.md`／`paper.tex` 现在各自写明编译结果（未检测到引擎 ⇒ 只产出 tex+md，或 PDF 编译失败并保留 tex/md），**恰好一次**且幂等 ⇒ 2.8.0 说明里的「已知限制」第 3 条已解决；新增 `scripts/release-check.mjs` 把发布前逐项核对做成一条命令（版本三处一致／两份说明章节顺序／清单登记／声明缺失=0／开发专用文件不随包／包内无 CRLF）。 | [中文](./RELEASE-NOTES-2.8.1.md) · [English](./RELEASE-NOTES-2.8.1.en.md) |

package/docs/release-notes/RELEASE-NOTES-2.8.4.en.md

@@ -0,0 +1,48 @@
+# dsh-vibe-math 2.8.4 — Release Notes
+
+> Previous release: 2.8.3. This is a **patch release**: it fixes v5's **meeting watchdog**, which used to abandon a meeting while an ASKED member was still working — turning ordinary speeches into "late notes" and making `paperEditor="office"` finalisations fail again and again. Parameter names and defaults, data formats and the directory layout are **unchanged**, the DSH support window is unchanged, and **no migration is needed**.
+
+## Overview
+
+- A meeting is **no longer** abandoned before an ASKED member still in flight hands its speech in: such members may hold the meeting open up to **3×** the stall budget, after which the hard bound still closes it (a genuinely wedged turn must not wedge the whole meeting).
+- The practical effect is that **ordinary speeches land in the ordinary sections again**: one real run before the fix showed **1067** "late notes" and **863** refused finalisations; with the fix, meetings actually collect their speakers and the "late note" path goes back to being an **insurance policy** rather than the norm.
+- The `paperEditor="office"` gate therefore becomes satisfiable in practice (it requires at least one office message and one **recorded** meeting for this paper).
+- The documentation now states when a meeting is abandoned, how the abandonment is named, and what happens to speeches that arrive **after** it.
+
+## Added
+
+- None. (Patch release: no new parameters, fields or tools.)
+
+## Changed
+
+- Meeting stall decision: within the soft bound (`2 × activityTimeoutMs`, default 240 s) the meeting is kept open as long as an **asked member is still in flight**; past the **3×** hard bound (default 720 s) it is still abandoned.
+- `docs/final-paper.md` §7 now documents "when a meeting is abandoned" (soft/hard bounds and their relation to the "late note" path).
+
+## Fixed
+
+- Symptom: after members (including the **convener**) spoke in a meeting, the minutes only gained "late note" entries — one real run reached **1067** of them, and `paperEditor="office"` finalisations were refused by `V5_PAPER_CONSULT_REQUIRED` **863** times. Cause: the stall check measured only the time since the last INPUT and **ignored whether an asked member was still busy**; a member's turn on a real host frequently outlives the 240 s budget, so the meeting was abandoned by the watchdog and later speeches could only take the late-note path. Members that were asked and are still busy now hold the meeting open until the hard bound (3× the budget), so ordinary speeches return to the ordinary sections. Guards: `tests/e2e-v5-round2.test.mjs` (`★ [real1004-stall]`, a **source-level invariant** — this branch cannot be triggered deterministically inside the shipped suites, so the criterion is "source invariant + real-host re-test"), `tests/v5-institute-fixes.mutants.mjs` (**19/19** named reds).
+
+## Compatibility & Migration
+
+- Breaking changes: **none**.
+- Additive only: no new fields (behaviour and documentation only).
+- Migration: none needed.
+- The DSH support window is unchanged (see `engines.dsh` in `package.json`).
+
+## Known Limitations
+
+- The **real-host** behaviour of the meeting watchdog can only be confirmed by a real-host re-test (inside the shipped suites the heartbeat delay is not controllable, so the branch cannot be triggered deterministically; the guard is therefore a **source-level invariant**, as stated in the test comments and the commit message).
+- **TeX installed in a non-standard location is still not auto-discovered**: the default probe covers PATH and the documented common install roots.
+- The shipped suites **do not cover host-level behaviour** (for example "the caller is dropped after an interrupt").
+- **A root-agent question blocks the whole institute** (operational note, product unchanged): state the problem and acceptance criteria up front for unattended runs. See [`slv-playbook.md`](./slv-playbook.md) §5.
+
+## Verification
+
+- `node tests/run-tests.mjs` — all green (**106** items: 44 suites + 62 probes).
+- `node scripts/release-check.mjs` — exits 0; after publishing, add `--registry` to compare the registry shasum with the local tarball.
+- This release's fix: `node tests/e2e-v5-round2.test.mjs` (`★ [real1004-stall]`, measured `passed=436 failed=0`) and `node tests/v5-institute-fixes.mutants.mjs` (**19/19** named reds).
+- Real-host re-check recipe: install the exact published version (`-Pin <version>`) and run a **trimmed** meeting + paper task (no research), then count "late notes" (1067 before the fix) and refused finalisations (863 before the fix) — both should drop sharply.
+
+## Dependencies
+
+- No new runtime dependencies; the DSH support window is unchanged.

package/docs/release-notes/RELEASE-NOTES-2.8.4.md

@@ -0,0 +1,48 @@
+# dsh-vibe-math 2.8.4 — 发布说明
+
+> 上一版本：2.8.3。本版是**补丁版**：修正 v5 的**会议看门狗**——它此前会在"已被点名的成员仍在进行中"时就把会议放弃，导致成员的正常发言退化成「会后补记」，并让 `paperEditor="office"` 的定稿被反复拒绝。参数名与默认值、数据格式与目录结构**未变**，DSH 支撑窗口不变，**无需迁移**。
+
+## 概述
+
+- 会议**不再**在"被点名、且仍在进行中"的成员交回发言之前被放弃：这类成员最多把会议续命到停滞预算的 **3 倍**，超过硬界才放弃（真正卡死的一轮仍不能拖住整场会议）。
+- 结果是**正常发言回到正常小节**：真机实测修复前一次运行出现 **1067** 次「会后补记」、**863** 次定稿被拒；修复后会议能真正收齐发言，「会后补记」退回为**保险**而非常态。
+- `paperEditor="office"` 的定稿闸门因此更容易被真正满足（它要求本张论文至少有 1 条所办消息与 1 次**记录在案**的会议）。
+- 文档写明会议何时会被放弃、放弃如何具名、以及放弃**之后**到达的发言如何处置。
+
+## 新增
+
+- 无。（本版为修复版：没有新增参数、字段或工具。）
+
+## 变更
+
+- 会议停滞判定：软界（`2 × activityTimeoutMs`，默认 240 s）内，只要仍有**被点名的成员在进行中**就不放弃；超过 **3 倍**（默认 720 s）的硬界则照旧放弃。
+- `docs/final-paper.md` §7 增补"会议何时会被放弃"的口径（含软界／硬界与「会后补记」的关系）。
+
+## 修复
+
+- 症状：成员（含**召集人**）在会议里发言后，纪要里却只留下「会后补记」条目；一次真机运行累计 **1067** 次补记，并且 `paperEditor="office"` 的定稿被 `V5_PAPER_CONSULT_REQUIRED` 拒绝 **863** 次。原因：会议停滞只按"最后一次发言"计时，**不看被点名的成员是否仍在忙**；真机上成员的一整轮常常超过 240 s 的预算，于是会议在看门狗手里被放弃，随后到达的发言只能走补记通道。现在"被点名且仍在忙"的成员会把放弃推迟到硬界（3 倍预算），正常发言因此回到会议的正常小节。守卫：`tests/e2e-v5-round2.test.mjs`（`★ [real1004-stall]`，**源码级不变量**：该分支的真机行为无法在随包套件里确定性触发，故判据是"源码不变量 ＋ 真机复测"）、`tests/v5-institute-fixes.mutants.mjs`（**19/19** 具名红）。
+
+## 兼容性与迁移
+
+- 破坏性变更：**无**。
+- 纯新增：无新增字段（仅行为与文档变化）。
+- 迁移：无需迁移。
+- DSH 支撑窗口不变（见 `package.json` 的 `engines.dsh`）。
+
+## 已知限制
+
+- 会议看门狗分支的**真机**行为只能由真机复测确认（随包套件里心跳延时不可控 ⇒ 该分支不能确定性触发；因此守卫是**源码不变量**，这一点已在测试注释与提交信息里写明）。
+- **非标准位置的 TeX 仍不会被自动发现**：默认探测只覆盖 PATH 与文档化的常见安装根。
+- 随包套件**不覆盖宿主层行为**（例如"中断后调用方被丢弃"）。
+- **根代理提问会阻塞整所流程**（运维注意项，不改产品）：长跑时请把问题与验收条件一次说清。详见 [`slv-playbook.md`](./slv-playbook.md) §5。
+
+## 验证方式
+
+- `node tests/run-tests.mjs` —— 全部通过（**106** 项：44 套件 + 62 探针）。
+- `node scripts/release-check.mjs` —— 退出 0；发布后加 `--registry` 比对线上 shasum 与本地 tarball。
+- 本版修复：`node tests/e2e-v5-round2.test.mjs`（`★ [real1004-stall]`，实测 `passed=436 failed=0`）与 `node tests/v5-institute-fixes.mutants.mjs`（**19/19** 具名红）。
+- 真机复测配方：用 registry 上的精确版本（`-Pin <version>`）跑一个**精简**会议＋论文任务（跳过研究），统计「会后补记」次数（修复前 1067 次）与定稿被拒次数（修复前 863 次）应显著下降。
+
+## 依赖
+
+- 无新增运行时依赖；DSH 支撑窗口不变。

package/docs/release-notes/RELEASE-NOTES-2.8.5.en.md

@@ -0,0 +1,75 @@
+# dsh-vibe-math 2.8.5 — Release Notes
+
+> Previous release: 2.8.4. This is a **patch release** (the version stays on the 2.8.x line): it is mostly **fixes** — the Lean formalization path could not compile by default, TeX engines that are installed but not on PATH are now found, plus several corrections to the member-facing prompts; it also carries one **purely additive** capability (the feedback library, on by default and changing no existing behaviour). Parameter names and defaults, data formats and the directory layout are **unchanged**, the DSH support window is unchanged, and **no migration is needed**.
+
+## Overview
+
+- **The Lean formalization path works again**: the injected search-root flag was spelled for Lean 3, so the compiler failed at **argument parsing** (the file was never read) and formalization could never pass; the presets now inject `-R <root>`, and on a real host a minimal proof compiles, is archived, and can be read and reused by another member.
+- **TeX engines that are installed but not on PATH are now found**: new **bounded**, documented candidate roots (still documented locations only, no full-disk scan, never an automatic install, never a write outside the workspace); when nothing is detected, `triedPaths` still **names** every path that was probed.
+- **The prompts no longer demand things a member cannot do**: a missing engine is now reported to the office (or the group chat) instead of "asking the user"; the LaTeX guidance states **when it applies**; the "receipt" concept is given background; the empty engine table no longer contradicts itself; the member-facing notice frame is now **【研究所提示】**; and agent-facing text no longer contains drive-letter literals.
+- **New cooperation / management / work-experience feedback library** (v5): record how the institute works and close the loop — the `self`/`team` routes are **self-adjustment with no approval**, while the `interpersonal` route requires an assessment and a **written-back verification** afterwards; switch `feedback` (default `on`).
+- **The two install/upgrade traps are now documented**: `dsh plugin …` needs a working **`pnpm`**, and install/upgrade should **pin an explicit version** (a bare name can stay on an OLD version).
+- Maintainer-facing: a set of static invariants and named-mutation self-checks across all four presets; the gate is **decoupled** from whether the host has TeX installed, so the suites give the same result on any machine.
+
+## Added
+
+- Tool **`vibe_v5_feedback`** (v5, member-usable): `op` is `'add' | 'update' | 'list' | 'summary'`; `list` defaults to the **not-yet-closed** entries (filter by `category`/`route`; `all: true` shows everything); `summary` counts by category and route.
+- Carrier: `Shared/Feedback/<category>.md` (`cooperation`｜`management`｜`process`｜`obstacle`｜`conflict`) plus a `Shared/Feedback/index.json` index; the authoritative data lives in the institute's **durable state**, so a restart does not lose it.
+- Parameter **`feedback`** (v5): `'on'` (default) | `'off'`. With `'off'` the prompt section is **not injected** and **every** operation of the tool is **refused by name** (`V5_FEEDBACK_DISABLED`) — never a silent write.
+- **Three routes** (stated in the prompt and the docs): `self` = your own way of working ⇒ **adjust it yourself, nobody has to adopt it**; `team` = organisation / workflow / how the team runs ⇒ **no approval either**, change it as soon as it looks wrong; `interpersonal` = **caused by someone else and only they can fix it** ⇒ the **only** route that requires an assessment first and a **written-back verification** afterwards (`outcome` must be filled in before it can close).
+- Permissions: members and temp workers may `add`, and may `update` only the entries **they started**; the office may `update` any entry (mark it resolved / write the result back).
+- Observability: entry count / not-yet-closed count / per-category / per-route counts appear in `vibe_v5_feedback {op:'summary'}`, `vibe_v5_report` and `vibe_v5_overview`.
+
+## Changed
+
+- Prompts (the four affected presets): handling a missing engine changed from "ask the user" to "**report what happened to the office** (or the group chat)", and the office confirms with the user — members have no channel to the user.
+- Prompts: the LaTeX guidance now says it **only applies while you are writing or compiling the paper**, instead of appearing as unexplained noise in every wake.
+- Prompts: in `math_computation`'s "archive → edit → re-run" section, the **first** mention of a receipt now explains "**a receipt = the JSON result of one `math_computation` call** — run `probe` or `run` once and its fields are right there", with minimal glosses on the key fields.
+- Prompts: the empty engine table changed from `math_computation：本机可用 （无）` to **`math_computation：可用引擎 无`** (English `available engines none`).
+- Prompts: the member-facing notice frame changed from `【框架提示】` to **`【研究所提示】`**.
+- Documentation: the install section and its troubleshooting note state the `pnpm` prerequisite and the "explicit version" rule; the README's old "upgrade with `@latest`" wording is corrected to an explicit version.
+- Maintainer-facing: the shipped prompt corpus was regenerated together with the text above.
+
+## Fixed
+
+- **Symptom**: Lean formalization almost never passed in v2/v3/v4/v5 — the log showed many Lean invocations while `formal.passed` stayed empty. **Cause**: the injected search-root flag was Lean 3's `--search-path`, while Lean 4 only accepts `-R` / `--root`, so the compiler exited with `rc=1` at **argument parsing** and the file was never read. **Fix**: all four presets inject `-R <root>`; if the user already set a search root in `leanArgs`, no second one is injected (semantics unchanged). **Verified**: on a real host the compiler is invoked correctly and reads the file; the minimal proof `theorem probe_true : True := trivial` compiles and is archived into `Verified/Lean/`, after which **another member reads that archived lemma and uses its statement as the premise of a new theorem** (reuse works). Guards: `tests/formal-verify-v2.test.mjs` / `tests/formal-verify-v3.test.mjs` / `tests/formal-verify-v4.test.mjs` / `tests/formal-verify-v5.test.mjs`, `tests/audit-prompt-invariants.mjs` (flag spelling plus "no preset text may teach `--search-path` again"), `tests/v5-institute-fixes.mutants.mjs` (named reds).
+- **Symptom**: when TeX Live was installed at a location such as `texlive\<year>\bin\windows` (installed but not on PATH), all four presets reported "no LaTeX engine detected" and delivered tex+md only. **Cause**: the default probe covered PATH and MiKTeX's common roots only. **Fix**: new **bounded** candidate roots (Windows `texlive\<year>\bin\windows` under a drive root, `/usr/local/texlive` and `/opt/texlive` on Unix-like systems, `/Library/TeX/texbin` on macOS) with a **cap** on the total; still **documented locations only, no full-disk scan, never an automatic install, never a write outside the workspace**; when nothing is detected, `triedPaths` still **names** every path that was probed. **Verified**: on a real host an engine that is installed but not on PATH is found (the PATH lookup fails, a documented root hits); with those roots sandboxed, every candidate probe stays inside the allowed tree and the total stays under the cap. Guards: `tests/audit-prompt-invariants.mjs` (documented-root coverage / the cap / no bare drive root / every probe site must go through the sandboxable seam), `tests/e2e-v5-round2.test.mjs`, `tests/audit-math-computation-parity.mjs`.
+- **Symptom**: members were asked to do something they cannot do ("ask the user"). **Fix**: changed to "report what happened to the office (or the group chat)". Guards: `tests/audit-prompt-invariants.mjs` (member-visible text may not demand an impossible action), `tests/e2e-v5-round2.test.mjs`.
+- **Symptom**: agent-facing text contained host absolute paths (drive-letter literals), which are wrong guidance on another machine. **Fix**: rewritten portably (`texlive\<year>\bin\windows`, `/usr/local/texlive`, `/opt/texlive`, `/Library/TeX/texbin`). Guards: `tests/audit-prompt-invariants.mjs` (source-level drive-letter criterion plus the path criterion over the four prompt corpora), `tests/audit-path-discipline.mjs`.
+- **Symptom**: the first time a member saw "the receipt fields `scriptPath` / `scriptAbs`" it had no idea what a receipt was and probed around for it. **Fix**: the "archive → edit → re-run" section now gives that background at its **first** mention. Guards: `tests/e2e-v5-round2.test.mjs` (module-level behaviour plus the member prompt end to end), `tests/v5-institute-fixes.mutants.mjs` (removing that sentence reddens by name).
+- **Symptom**: with no engine, the prompt rendered `math_computation：本机可用 （无）`, which contradicts itself. **Fix**: it now reads "可用引擎 无" (English `available engines none`). Guards: `tests/e2e-v5-round2.test.mjs`, `tests/v5-institute-fixes.mutants.mjs` (restoring the old wording reddens by name).
+- **Symptom**: member-facing text called the plugin "框架", which is ambiguous from an agent's point of view (the plugin? the host? the institute?). **Fix**: the notice frame is now **【研究所提示】**. Guards: `tests/e2e-v5-round2.test.mjs` (source level: `【框架提示】` no longer appears in agent-facing text), `tests/v5-institute-fixes.mutants.mjs`.
+
+## Compatibility & Migration
+
+- Breaking changes: **none**.
+- Additive only: the `vibe_v5_feedback` tool, the `feedback` parameter (default `'on'`), and the `Shared/Feedback/` directory with its index.
+- Migration: **none needed**; existing parameter names and defaults, data formats and the directory layout are unchanged.
+- The DSH support window is unchanged (see `engines.dsh` in `package.json`).
+- **Upgrade**:
+
+```
+dsh plugin --profile <your profile> add dsh-vibe-math@2.8.5
+```
+
+  - If you are still on an **OLD version** after upgrading: first check that `pnpm` works — the plugin management behind `dsh plugin …` needs a working `pnpm`; without it the install fails, and the **plugin market may show no reason** (attach `~/.dsh/profiles/<profile>/hub.log` to locate it).
+  - Then check whether that profile's `package.json` / `pnpm-lock.yaml` still points at the old version: they **win** over the `latest` tag, so **a bare name can stay on an OLD version**; retry with an explicit version.
+
+## Known Limitations
+
+- The LaTeX engine probe still covers **documented locations only**: an engine in a non-standard directory will not be discovered; use the `paperLatexCommand` parameter to name the engine command by absolute path (the prompt tells you to probe again after setting it).
+- The few behaviours that need a real engine/host can only be confirmed by a **real-host re-test**; the shipped suites protect them with **source-level invariants plus a sandboxable seam**, and do not claim to cover host-level behaviour (for example "the caller is dropped after an interrupt").
+- **The feedback library produces no research evidence**: it records methodology and collaboration only (how work is organised, where it is stuck, friction between people); research conclusions still go to `Progress/` and `Verified/`.
+- Install/upgrade: on this machine's toolchain `@latest`, `@^2`, `pnpm update --latest` and `pnpm add …@latest` all failed to guarantee an upgrade, so an **explicit version** is still recommended (see "Upgrade").
+- A root-agent question blocks the whole institute (operational note, product unchanged): state the problem and the acceptance criteria up front for long runs.
+
+## Verification
+
+- `node tests/run-tests.mjs` — all green (**106** items: 44 suites + 62 probes).
+- `node scripts/release-check.mjs` — exits 0; after publishing, add `--registry` to compare the registry shasum with the local tarball.
+- This release's suites: `node tests/e2e-v5-round2.test.mjs` (measured `passed=484 failed=0`), `node tests/v5-institute-fixes.mutants.mjs` (**33/33** named reds), `node tests/audit-prompt-invariants.mjs` (**211/0**; `--self-probe` **26/26**), `node tests/audit-artifact-docs.mjs` (**24/0**; its mutant harness **8/8**), `node tests/audit-math-computation-parity.mjs` (**108/0**).
+- Real-host re-check recipe: ① **Lean** — on a host with Lean installed, send one object through `vibe_v5_lean_archive` (`kind:'proof'`); it should compile, the file should appear under `Verified/Lean/`, and another member should be able to `import` it; ② **TeX** — install the engine into a documented root but keep it off PATH; the paper flow should no longer report "no engine detected", and `triedPaths` should list the paths that were probed.
+
+## Dependencies
+
+- No new runtime dependencies; the DSH support window is unchanged.