diff --git a/.agent/HANDOFF.md b/.agent/HANDOFF.md
index b431e7c..d096c68 100644
--- a/.agent/HANDOFF.md
+++ b/.agent/HANDOFF.md
@@ -2,187 +2,107 @@
 
 ## 项目概览
 
-CMP 是使用 mcpp 构建的 C++23 Modules 协程运行时库，公开模块为 `mcpplibs.cmp`。
-v1 当前包含：
+CMP 是使用 mcpp 构建的 C++23 Modules 协程运行时库，公开根模块为 `mcpplibs.cmp`。
+当前 v1 能力包括 `Task<T>`、`RunLoop`、定时与协作取消、`when_all()`、`TaskGroup`、
+异步事件与互斥、固定大小 `ThreadPool`、`run_blocking()`，以及单 driver `IoContext`、
+数值地址 `TcpStream`/`TcpListener`。
 
-- 懒启动、单消费者 `Task<T>`；
-- 变参/vector `when_all()` 与静止点 `TaskGroup`；
-- `OneShotEvent`、`AsyncManualResetEvent`、RAII `AsyncMutex`；
-- 带定时和 `std::stop_token` 取消的调用线程 `RunLoop`；
-- 固定大小 `ThreadPool` 与结构化同步调用隔离 `run_blocking()`；
-- 单私有 I/O driver 的 `IoContext`、数值地址 TCP client `TcpStream`，以及数值地址
-  TCP server `TcpListener`。
-
-项目固定 mcpp `2026.8.28.1`、LLVM `22.1.8`、Asio `1.38.1` 与 gtest `1.15.2`。
-三平台 CI 引导 xlings 固定为 `2026.8.27.5`。`examples/basic` 是独立 path-dependency
-consumer。
+包版本保持 `0.1.0`，项目固定 mcpp `2026.8.28.1`、LLVM `22.1.8`、Asio `1.38.1`、
+gtest `1.15.2`。DNS、TLS、UDP、detached task、自动取消传播、work stealing 与通用 I/O
+不属于当前 v1 契约。
 
 ## 当前目标与状态
 
-PR #10 已合并，`main` 已同步到 squash commit `9492075`。当前分支为
-`feature/phase6c-tcp-server-v1`。Phase 6C TCP Server v1 已通过全部本地门禁，核心实现提交
-为 `e0248ca`（`实现协程 TCP 服务端`），当前分支已推送至 `origin`。
-
-Phase 7 是最后一个已规划的 v1 阶段，只做发布验收，不增加运行时 API。它的 Design/Plan
-已完成复核，并已执行不依赖远程权限的早期包转换预检；正式实施仍等待 Phase 6C 合并后
-切换到独立分支。
-
-用户已授权当前持续目标内的 Git commit 与 push。Phase 6C PR #11 已创建，Linux、macOS、
-Windows CI 均通过；合并仍需用户另行明确授权。
-
-## 已完成工作
-
-- 在现有 `mcpplibs.cmp:tcp` 分区增加 move-only `TcpListener`：
-  - `bind(IoContext&, Scheduler, numericAddress, port, stop_token)`；
-  - `accept(Scheduler, stop_token)`；
-  - `close()`、`is_open()`、`local_port()`。
-- bind/accept 保持懒执行；每个成功或错误都尝试通过调用方选择的返回 Scheduler 发布。
-- 支持数值 IPv4/IPv6、端口零、系统最大 backlog、一个 pending accept、取消后复用、
-  线程安全幂等 close，以及 listener 与已接受 `TcpStream` 的独立所有权。
-- `IoContext` 增加私有 listener 弱引用 registry；shutdown 先关闭 listener，再关闭
-  stream，并排空 native handler 后 join driver。
-- accept 候选 socket 在 native initiation 前登记，并在返回 Scheduler 后复核待发布资源；
-  自审同时纠正 bind/connect/accept 的 context shutdown 发布窗口，关闭获胜时返回
-  `OperationCancelled`，不会发布已关闭对象。
-- read/write/accept 的 pending 标志是唯一重叠门禁；driver 可覆盖已完成 operation 的弱引用，
-  不再错误依赖上一 Task 帧已经销毁。
-- 增加 21 项 Phase 6C TCP 测试；`tcp_test` 现为 45 项，完整套件为 10 个二进制、161 项。
-- readiness 的正常 TCP server 已从阻塞 POSIX socket 迁移到
-  `TcpListener`/`TcpStream`；POSIX 只保留未监听失败端口构造。
-- 三种 README、三种架构文档、Phase 6C Design/Plan 和 readiness 报告已同步。
-- 新增并复核 Phase 7 v1 发布验收 Design/Plan，明确三平台双配置 CI、独立 consumer、
-  源码归档、mcpp-index Form A、GitHub Release 与各远程授权边界；Phase 7 后不自动规划
-  Phase 8。
-- 预检纠正两项 mcpp `2026.8.28.1` 行为假设：生成的描述符不能直接作为 index 文件，
-  `--allow-dirty` 也不会把未提交工作树内容装入源码归档。
-
-## 重要决策
-
-- 只提供数值地址 bind；不隐藏阻塞 DNS。
-- port 0 交由系统选择，`local_port()` 返回实际端口；close 后保留最后绑定端口，
-  moved-from 返回 0。
-- v1 使用系统最大 backlog，不隐式设置 `reuse_address`，不开放任意 socket option。
-- 同一 listener 只允许一个 accepted-but-not-yet-published accept；连续服务由普通 accept
-  循环与结构化 handler 组合。
-- pre/active accept cancellation 保持 listener 可用；listener close 或 context shutdown
-  获胜时 pending accept 恰好以 `OperationCancelled` 完成一次。
-- 继续使用一个私有 I/O driver 和两类线性弱引用 registry；只有基准证明瓶颈后才扩展。
-- 不增加 DNS、TLS、UDP、endpoint 类型、超时、连接配额、detached handler、通用 I/O
-  抽象或新依赖。
-
-## 路线完整性审计
-
-- 从 bootstrap、RunLoop、timer、cancellation、`when_all`、TaskGroup、events、AsyncMutex，
-  到 Phase 4–7 及 CI 冷缓存修复，共有 16 组逻辑 Design/Plan；已逐一核对，全部成对存在。
-- Phase 4 已完成 v1 基础契约和 readiness；Phase 5 已完成 ThreadPool；Phase 6A 已完成
-  blocking offload；Phase 6B 已完成 TCP client，并由 PR #10 三平台验证。
-- Phase 6C TCP server 的本地实现、161 项门禁和 PR #11 三平台 CI 均已完成，只待授权合并。
-- Phase 7 是最后一个计划阶段，仅做发布资格、三平台双配置、干净归档、GitHub Release 和
-  mcpp-index 验收。没有遗漏的已规划 Phase 8；未来能力必须由真实需求启动独立 Design。
-
-## 修改 / 重要文件
-
-- 实现：`src/tcp.cppm`
-- 测试：`tests/tcp_test.cpp`
-- readiness consumer：`benchmarks/v1-readiness/src/main.cpp`
-- readiness 数据：`docs/benchmarks/2026-08-29-cmp-v1-readiness.md`
-- Design：
-  `docs/superpowers/specs/2026-08-30-cmp-phase6c-tcp-server-v1-design.md`
-- Plan：
-  `docs/superpowers/plans/2026-08-30-cmp-phase6c-tcp-server-v1.md`
+Phase 6C PR #11 已在 Linux x86_64、macOS arm64、Windows x86_64 全部通过后授权合并；
+`main` 的对应 squash commit 为 `02cb6b9`。本地已从该提交创建
+`release/phase7-v1-release-qualification`。
+
+Phase 7 是最后一个已规划的 v1 阶段，只做发布资格验收，不增加运行时 API。当前已完成
+本地实现、测试、readiness、早期源码归档和 Form A 描述符预检；Phase 7 项目 PR、远程
+三平台 CI、干净合并候选归档和公开发布仍未完成。验收门禁提交为 `941e42c`
+（`完善 v1 发布验收门禁`），当前分支已推送并跟踪
+`origin/release/phase7-v1-release-qualification`。
+
+Phase 7 PR #12 已获授权并创建。首轮 Windows CI 在真正编译测试前因 mcpp 已知 depfile
+降级失败；根因修正提交 `b2c0911`（`修正 Windows 发布验收门禁`）已通过 Linux、macOS、
+Windows 三平台。PR 描述已获授权并修正为严格探针的真实语义，当前保持 OPEN、CLEAN、
+MERGEABLE，尚未获得合并授权。
+
+用户已授权当前持续目标内的 Git commit 与 push。PR #12 的修改、PR 合并、tag、GitHub
+Release、mcpp-index PR 和 index 合并仍必须按各自远程门禁单独授权。
+
+## 本阶段修改
+
+- 仅加强现有 `.github/workflows/ci-linux.yml`、`ci-macos.yml`、`ci-windows.yml`：
+  - Linux/macOS 根包运行严格、关闭缓存的 Dev/Release build 和 test；
+  - Windows 先运行严格探针，只允许已知 depfile 降级，再执行无 `--strict` 的同配置无缓存
+    完整 build/test；
+  - 独立 example 运行一次，并增加对应平台的无缓存 Release build。
+- 三种 README 与三种架构文档同步相同的本地验证命令和真实发布状态。
+- Phase 7 Design/Plan 记录已完成的本地证据和仍待完成的远程门禁。
+- 未修改运行时 API、源码、测试、版本、依赖或打包实现。
+
+## 本地验证
+
+- 本机 Linux 根包 Dev：`mcpp build --profile dev --strict --cache=off` 通过；
+  `mcpp test --profile dev --strict --cache=off` 为 10 个二进制、161/161 通过。
+- 根包 Release：`mcpp build --profile release --strict --cache=off` 通过；
+  `mcpp test --profile release --strict --cache=off` 为 10 个二进制、161/161 通过。
+- `examples/basic`：`mcpp run` 输出预期十行并退出 0；严格无缓存 Release build 通过。
+- readiness Release 连续五轮全部 PASS，每轮硬计数均为：
+  - compute：49,000 成功 / 1,000 预期失败 / 0 非预期失败；
+  - file I/O：1,000 / 100 / 0；
+  - network：20,000 / 100 / 0。
+- 五轮中位数：compute `38.475 ms / 1,299,546.0 ops/s`，file I/O
+  `217.363 ms / 5,060.7 ops/s`，network `2,622.017 ms / 7,665.9 ops/s`。
+- 三个 workflow 均通过 PyYAML 解析，并包含一致的发布资格流程。
+- `mcpp publish --dry-run --allow-dirty` 基于已提交的 Phase 6C 合并基线生成早期归档：
+  `cmp-0.1.0.tar.gz`，206,341 bytes，临时 SHA-256
+  `fe36976fe76225f6121cce81ca22a1b2392a833795795fb9a00ed52d98dba4de`。
+- 临时描述符规范化为 namespace `mcpplibs`、原子包名 `cmp`、无内联 `mcpp` 段的 Form A
+  后，`mcpp xpkg parse` 报告三平台 `0.1.0` 和 `parse OK`。
+- 解压后的早期源码归档使用已安装的 mcpp `2026.8.28.1` 完成 Dev/Release 严格无缓存
+  构建和 161/161 测试；归档内 example 运行和 Release build 也通过。
+- 本阶段没有源码或测试变更，按 Phase 7 规则无需新增 analyzer 检查。
+- 最终本地自审确认仅三个现有 workflow 与发布状态文档发生变化；三平台命令一致性、
+  YAML 结构、CRLF 和 `git diff --check` 均通过。
+- Windows 严格探针通过 Bash 语法检查；在本机严格构建成功路径退出 0，合成日志验证精确
+  已知降级会放行、增加任意 warning 会失败，且临时日志由 trap 清理。
+- PR #12 的 `b2c0911` 三平台检查全部通过：Linux 2m53s、macOS 3m08s、Windows 5m22s。
+  完整日志确认每个平台的 Dev/Release 各为 10 个二进制、161/161，example 均输出预期十行
+  并完成 Release build；Windows 探针只出现已知 depfile warning 与对应 strict 汇总。
+- PR #12 描述已通过 GitHub REST 精确更新并回读比对，明确 Linux/macOS 严格门禁、Windows
+  精确严格探针与无缓存完整门禁；未修改标题、分支或合并状态。
+
+## 已知限制
+
+- 早期归档只包含已提交的 `02cb6b9`，不包含当前未提交的 Phase 7 workflow/文档变更；
+  正式 SHA 必须在 Phase 7 合并并同步干净 `main` 后重新生成。
+- mcpp `2026.8.28.1` 自动描述符的内联 `mcpp` 段缺少 `sources`，不能直接解析；index
+  候选必须使用已验证的 Form A，且无真实 CN 镜像时保持普通字符串 URL。
+- 本机 `~/.mcpp` 的 vendored xlings `2026.8.11.2` 较旧且缺少 `subos_info`，因此显示
+  runtime facts inconclusive 警告；构建、测试、示例和 readiness 均实际通过，未修改用户
+  全局环境。离开项目环境做归档验证时使用已安装的 mcpp `2026.8.28.1` 绝对路径。
+- readiness 性能仅是本机诊断数据，不是 CI 阈值或 SLA。
+
+## 重要文件
+
 - Phase 7 Design：
   `docs/superpowers/specs/2026-08-30-cmp-phase7-v1-release-qualification-design.md`
 - Phase 7 Plan：
   `docs/superpowers/plans/2026-08-30-cmp-phase7-v1-release-qualification.md`
+- CI：`.github/workflows/ci-linux.yml`、`ci-macos.yml`、`ci-windows.yml`
 - 用户文档：`README.md`、`README.zh.md`、`README.zh.hant.md`
 - 架构文档：`docs/architecture.md`、`docs/architecture.zh.md`、
   `docs/architecture.zh.hant.md`
-- 当前状态：`.agent/HANDOFF.md`
-
-## 验证情况
-
-- `mcpp build --profile dev --strict --cache=off`：通过，LLVM `22.1.8`。
-- `mcpp test --profile dev --strict --cache=off`：10 个二进制、161/161 通过。
-- `mcpp build --profile release --strict --cache=off`：通过。
-- `mcpp test --profile release --strict --cache=off`：10 个二进制、161/161 通过。
-- `mcpp build --profile dist --strict --cache=off`：通过。
-- focused Release 发布窗口门禁连续 20 轮通过：bind、connect、accept 在返回 Scheduler
-  阻塞期间遇到 context shutdown 均返回取消，未发布已关闭资源。
-- focused Release 原生 operation 复用门禁连续 5 轮、25 项执行全部通过，覆盖
-  read/write/accept 的重叠拒绝与取消后复用。
-- Release `tcp_test` 使用随机顺序 seed `60830` 运行，45/45 通过，未发现用例顺序依赖。
-- 静态可移植性扫描确认 `src/`、`tests/`、`examples/` 未引入平台 socket 头或全局
-  POSIX socket 调用；此类调用只存在于明确标注 POSIX-only 的 readiness consumer。
-- 基于 mcpp 生成的 26 条模块编译数据库运行 `clang-analyzer-*` 检查
-  `src/tcp.cppm` 与 `tests/tcp_test.cpp`：两者退出码均为 0、无诊断。
-- stream pending read 与 listener pending accept 的并发 close 门禁在 Release 连续 100 轮
-  通过：合计 800 次并发 `close()` 均保持幂等，两类操作每轮都只取消完成一次。
-- focused Release listener 竞态连续 5 轮通过：共覆盖 100 次 close/accept 与 500 次
-  stop/accept，均恰好产生一个合法结果。
-- focused Release 生命周期门禁连续 20 轮、共 100 项执行全部通过，覆盖 ThreadPool 返回、
-  失效返回 Scheduler、主动取消后复用、context shutdown 与 32 客户端负载。
-- 额外尝试的 Release TCP 全套 50 轮压力并未作为通过项：前 11 轮 43/43 全通过，随后本机
-  仅 `60700-61000` 的临时端口范围被短连接 TIME_WAIT 占满，port-0 bind 开始返回
-  `Address already in use`；现场有 2,333 个 TIME_WAIT。未修改系统参数，也未让库隐式开启
-  `reuse_address` 来掩盖资源耗尽。
-- TIME_WAIT 自然回落到 2 后，正常单轮 Release TCP 套件立即恢复为 43/43 通过，确认该失败
-  来自宿主端口资源压力，而非持续 socket 泄漏或实现状态损坏。
-- `examples/basic` 的 `mcpp run`：通过，既有十行输出正确，退出码 0。
-- `examples/basic` 的 `dist` 构建及生成二进制运行：通过，同样输出正确十行，退出码 0。
-- readiness Release 严格无缓存构建通过；最终 5 轮共 15 行场景全部 PASS，所有轮次硬计数均为：
-  - compute：49,000 成功 / 1,000 预期失败 / 0 非预期失败；
-  - file I/O：1,000 成功 / 100 预期失败 / 0 非预期失败；
-  - network：20,000 成功 / 100 预期失败 / 0 非预期失败。
-- 最终中位数：compute `25.443 ms / 1,965,140.8 ops/s`，file I/O
-  `147.661 ms / 7,449.5 ops/s`，native async TCP client/server
-  `2,239.147 ms / 8,976.6 ops/s`。
-- Phase 6C 当前工作树再次执行 Dev/Release 严格无缓存构建与测试，均为 10 个二进制、
-  161/161 通过；`src/tcp.cppm` 与 `tests/tcp_test.cpp` 的 `clang-analyzer-*` 再次退出 0、
-  无诊断；独立 example 再次正确输出十行并退出 0。
-- 当前 readiness 二进制重新连续运行 5 轮，所有轮次仍保持 compute
-  `49,000/1,000/0`、file I/O `1,000/100/0`、network `20,000/100/0` 的
-  成功/预期失败/非预期失败硬计数。
-- `mcpp publish --dry-run --allow-dirty` 成功生成早期归档
-  `cmp-0.1.0.tar.gz`（184,046 bytes，SHA-256
-  `08e467dfe45cfdddde3664be7223995b56c7e172a6c6e568ce7ae9f3da4db062`）与描述符。
-  该 SHA 只属于当前已提交基线，Phase 6C 合并后的最终归档必须重新生成。
-- 生成描述符直接解析会因内联 `mcpp` 段缺少 `sources` 失败；临时候选改为短包名
-  `cmp` 并移除内联 `mcpp` 段后，`mcpp xpkg parse` 明确报告 Form A、三平台
-  `0.1.0`、`parse OK`。
-- 早期源码归档不含 `TcpListener`，确认 `--allow-dirty` 未装入 Phase 6C。归档对应的
-  Phase 6B 基线使用已安装的 mcpp `2026.8.28.1` 在隔离目录完成 Dev/Release 构建，
-  两套均为 10 个测试二进制、140/140 通过；归档内 example 的运行和 Release 构建也通过。
-- PR #11 的 Linux x86_64、macOS arm64、Windows x86_64 检查均通过；首轮耗时分别为
-  2m43s、2m45s、2m49s。
-
-## 已知问题 / 风险
-
-- 本机 `~/.mcpp` 仍保存旧 vendored xlings `2026.8.11.2` 且缺少 `subos_info`；
-  mcpp 提示 runtime facts inconclusive，但当前 LLVM 双配置构建、测试、示例和 benchmark
-  均成功。本阶段未修改用户全局环境。
-- 单 I/O driver 是 v1 简化；当前 readiness 只提供本机回归数据，不是 SLA。
-- 本机 IPv4 临时端口范围只有 301 个端口；不要连续重复完整 TCP 套件 50 轮。当前 5 轮竞态与
-  20 轮定向生命周期门禁是避免耗尽宿主资源的有界压力门。
-- bind 取消只在 driver setup 开始前协作生效；短同步 open/bind/listen 开始后由完成获胜。
-- 内置 RunLoop/ThreadPool 返回 Scheduler 会离开 I/O driver；自定义 Scheduler 必须自行提供
-  所需的线程转移语义，inline 实现不得在 driver continuation 中析构 `IoContext`。
-- listener 不提供 DNS、TLS、backlog 配置、socket option、多 pending accept 或自动 handler
-  所有权；这些都需要具体 consumer 与独立设计。
-- 运行中的 blocking callable 仍不可抢占；ThreadPool 仍为无界共享 FIFO。
-- 离开项目本地 `.xlings` 环境后，普通 xlings shim 目前只激活 mcpp `2026.8.11.2`，
-  无法解析 `.xlings.json` 要求的 `2026.8.28.1`；已安装的 `2026.8.28.1` 实际二进制可用，
-  归档隔离构建用它完成。不要把本地 `.xlings` 目录塞进发布包；正式 CI/index consumer
-  应按固定版本安装工具。
-- mcpp `2026.8.28.1` 自动描述符的内联 `mcpp` 段缺少 `sources`，且生成名是兼容全名；
-  index 候选必须规范化为短名 `cmp` 的 Form A。没有真实 CN 镜像时保留普通字符串 URL。
 
 ## 剩余工作
 
-1. 用户审查三平台全绿的 PR #11，并明确授权后再合并。
-2. 合并后同步 `main`、新建 Phase 7 分支，按 Design/Plan 加强三平台 Dev/Release CI，
-   并从干净候选重跑 161 项、example、readiness、归档及 Form A 验收。
-3. tag、GitHub Release、mcpp-index PR 与 index 合并均是独立远程操作，逐项等待授权。
+1. 合并前复核 PR #12 最新 HEAD 的三平台检查；获得单独授权后合并，同步干净 `main`，
+   无 `--allow-dirty` 重新生成并验证正式归档。
+2. tag、GitHub Release、mcpp-index PR、index 合并和最终独立 consumer 验证按 Design 的
+   独立授权门禁逐项执行。
 
 ## 推荐下一步
 
-由用户审查 PR #11 并确认是否授权合并；在此之前不执行 merge。
+停在 PR #12 合并门禁，等待用户明确授权。
diff --git a/.github/workflows/ci-linux.yml b/.github/workflows/ci-linux.yml
index bb0b3e5..7738282 100644
--- a/.github/workflows/ci-linux.yml
+++ b/.github/workflows/ci-linux.yml
@@ -51,13 +51,15 @@ jobs:
           xlings install -y
           mcpp --version
 
-      - name: Build library
-        run: mcpp build
-
-      - name: Run tests
-        run: mcpp test
+      - name: Qualify library
+        run: |
+          mcpp build --profile dev --strict --cache=off
+          mcpp test --profile dev --strict --cache=off
+          mcpp build --profile release --strict --cache=off
+          mcpp test --profile release --strict --cache=off
 
-      - name: Run example
+      - name: Qualify example
         run: |
           cd examples/basic
           mcpp run
+          mcpp build --profile release --strict --cache=off
diff --git a/.github/workflows/ci-macos.yml b/.github/workflows/ci-macos.yml
index b512808..ccde254 100644
--- a/.github/workflows/ci-macos.yml
+++ b/.github/workflows/ci-macos.yml
@@ -44,13 +44,15 @@ jobs:
           xlings install -y
           mcpp --version
 
-      - name: Build library
-        run: mcpp build
-
-      - name: Run tests
-        run: mcpp test
+      - name: Qualify library
+        run: |
+          mcpp build --profile dev --strict --cache=off
+          mcpp test --profile dev --strict --cache=off
+          mcpp build --profile release --strict --cache=off
+          mcpp test --profile release --strict --cache=off
 
-      - name: Run example
+      - name: Qualify example
         run: |
           cd examples/basic
           mcpp run
+          mcpp build --profile release --strict --cache=off
diff --git a/.github/workflows/ci-windows.yml b/.github/workflows/ci-windows.yml
index 7b76dfa..eee543b 100644
--- a/.github/workflows/ci-windows.yml
+++ b/.github/workflows/ci-windows.yml
@@ -51,16 +51,34 @@ jobs:
           xlings install -y
           mcpp --version
 
-      - name: Build library
+      - name: Validate strict-mode degradation
         shell: bash
-        run: mcpp build
+        run: |
+          # mcpp 2026.8.28.1 的 Windows LLVM 后端暂不生成 depfile；只放行这一条已知降级。
+          trap 'rm -f strict-mode.log' EXIT
+          set +e
+          mcpp build --profile dev --strict --cache=off --no-color > strict-mode.log 2>&1
+          strict_status=$?
+          set -e
+          cat strict-mode.log
+          if [ "$strict_status" -ne 0 ]; then
+            test "$strict_status" -eq 1
+            test "$(grep -Fc 'warning: this toolchain and platform combination emits no GNU depfile' strict-mode.log)" -eq 1
+            test "$(grep -Fc 'error: 1 degradation(s) reported and --strict is set' strict-mode.log)" -eq 1
+            test "$(grep -Ec '^(warning|error):' strict-mode.log)" -eq 2
+          fi
 
-      - name: Run tests
+      - name: Qualify library
         shell: bash
-        run: mcpp test
+        run: |
+          mcpp build --profile dev --cache=off
+          mcpp test --profile dev --cache=off
+          mcpp build --profile release --cache=off
+          mcpp test --profile release --cache=off
 
-      - name: Run example
+      - name: Qualify example
         shell: bash
         run: |
           cd examples/basic
           mcpp run
+          mcpp build --profile release --cache=off
diff --git a/README.md b/README.md
index 57b8b20..4f1098c 100644
--- a/README.md
+++ b/README.md
@@ -414,17 +414,25 @@ after CMP has a stable runtime API worth demonstrating.
 
 ## Development
 
-The local verification path is:
+On Linux and macOS, the local verification path is:
 
 ```bash
 mcpp build --profile dev --strict --cache=off
 mcpp test --profile dev --strict --cache=off
 mcpp build --profile release --strict --cache=off
 mcpp test --profile release --strict --cache=off
-cd examples/basic && mcpp run
+cd examples/basic
+mcpp run
+mcpp build --profile release --strict --cache=off
 ```
 
-CI runs the equivalent build, test, and standalone example flow on Linux, macOS, and Windows.
+On Windows, use the workflow's strict probe and remove `--strict` from the actual cache-off gates,
+as described below.
+
+CI runs cache-off Dev/Release library gates, the standalone example, and its Release build on all
+three platforms. Linux and macOS remain strict. With mcpp `2026.8.28.1`, Windows first uses a
+strict probe that allows only the known LLVM depfile degradation, then performs the cache-off gates
+without `--strict`; full rebuilds prevent that incremental limitation from reusing stale objects.
 The mcpp version is pinned by `.xlings.json`; contributors should not rely on an unrelated
 global mcpp installation.
 
diff --git a/README.zh.hant.md b/README.zh.hant.md
index e65babf..3db59a5 100644
--- a/README.zh.hant.md
+++ b/README.zh.hant.md
@@ -392,18 +392,24 @@ RunLoop 不是背景執行緒，也不會把阻塞程式碼自動變成非同步
 
 ## 開發與驗證
 
-本機驗證路徑：
+Linux 和 macOS 的本機驗證路徑：
 
 ```bash
 mcpp build --profile dev --strict --cache=off
 mcpp test --profile dev --strict --cache=off
 mcpp build --profile release --strict --cache=off
 mcpp test --profile release --strict --cache=off
-cd examples/basic && mcpp run
+cd examples/basic
+mcpp run
+mcpp build --profile release --strict --cache=off
 ```
 
-CI 在 Linux、macOS 和 Windows 上執行等價的建構、測試與獨立範例流程。mcpp 版本由
-`.xlings.json` 固定，不應依賴無關的全域 mcpp 安裝。
+Windows 應先使用 workflow 中的嚴格探針，並在實際無快取門禁中移除 `--strict`，原因見下文。
+
+CI 在三個平台都執行無快取 Dev/Release 函式庫門禁、獨立範例及其 Release 建構；Linux 和
+macOS 保持嚴格模式。mcpp `2026.8.28.1` 在 Windows 上先執行嚴格探針，只放行已知的 LLVM
+depfile 降級，再移除 `--strict` 執行無快取門禁；完整重建使這項增量限制無法沿用舊目的檔。
+mcpp 版本由 `.xlings.json` 固定，不應依賴無關的全域安裝。
 
 CMP 目前不追蹤 `mcpp.lock`，`.gitignore` 明確執行這項儲存庫約定。執行期相依放在
 `[dependencies]`，gtest 明確宣告在 `[dev-dependencies.compat]` 中。
diff --git a/README.zh.md b/README.zh.md
index 972f8a4..f4a8988 100644
--- a/README.zh.md
+++ b/README.zh.md
@@ -392,18 +392,24 @@ RunLoop 不是后台线程，也不会把阻塞代码自动变成异步代码。
 
 ## 开发与验证
 
-本地验证路径：
+Linux 和 macOS 的本地验证路径：
 
 ```bash
 mcpp build --profile dev --strict --cache=off
 mcpp test --profile dev --strict --cache=off
 mcpp build --profile release --strict --cache=off
 mcpp test --profile release --strict --cache=off
-cd examples/basic && mcpp run
+cd examples/basic
+mcpp run
+mcpp build --profile release --strict --cache=off
 ```
 
-CI 在 Linux、macOS 和 Windows 上执行等价的构建、测试和独立示例流程。mcpp 版本由
-`.xlings.json` 固定，不应依赖无关的全局 mcpp 安装。
+Windows 应先使用 workflow 中的严格探针，并在实际无缓存门禁中去掉 `--strict`，原因见下文。
+
+CI 在三个平台都执行无缓存 Dev/Release 库门禁、独立示例及其 Release 构建；Linux 和 macOS
+保持严格模式。mcpp `2026.8.28.1` 在 Windows 上先运行严格探针，只放行已知的 LLVM depfile
+降级，再去掉 `--strict` 执行无缓存门禁；完整重建使这一增量限制无法复用旧目标文件。mcpp
+版本由 `.xlings.json` 固定，不应依赖无关的全局安装。
 
 CMP 当前不跟踪 `mcpp.lock`，`.gitignore` 明确执行这一仓库约定。运行时依赖放在
 `[dependencies]`，gtest 明确声明在 `[dev-dependencies.compat]` 中。
diff --git a/docs/architecture.md b/docs/architecture.md
index 2604bde..f652778 100644
--- a/docs/architecture.md
+++ b/docs/architecture.md
@@ -497,7 +497,7 @@ Further partitions or implementation units are added only when another implement
 
 ## Verification
 
-Run from the repository root:
+Run from the repository root on Linux or macOS:
 
 ```text
 mcpp build --profile dev --strict --cache=off
@@ -506,10 +506,12 @@ mcpp build --profile release --strict --cache=off
 mcpp test --profile release --strict --cache=off
 cd examples/basic
 mcpp run
+mcpp build --profile release --strict --cache=off
 ```
 
-The expected result is a successful library build, 161 passing tests across ten binaries, and an
-example that prints `Coroutine result: 42`, `Concurrent result: 42`, `Worker pool result: 42`,
+The expected result is successful Dev/Release library and example builds, 161 passing tests across
+ten binaries, and an example that prints `Coroutine result: 42`, `Concurrent result: 42`,
+`Worker pool result: 42`,
 `Blocking result: 42`, `Task group result: 42`, `Recursive group result: 3`, `Event signalled`,
 `Reusable event cycles: 2`, `Mutex result: 42`, then
 `Coroutine cancelled` and exits with status 0. Tests retain the existing high-volume stack checks
@@ -522,6 +524,8 @@ temporary-file, and loopback-network counts and
 throughput are recorded in the
 [v1 readiness benchmark](benchmarks/2026-08-29-cmp-v1-readiness.md). ThreadPool counts, concurrency,
 and five-round Release measurements are recorded in the
-[ThreadPool benchmark](benchmarks/2026-08-29-cmp-thread-pool.md). The current Windows LLVM toolchain
-does not emit GNU depfiles. If a file included by a module interface changes, an incremental build
-can reuse an older BMI or object; `--cache=off` is used for a full local verification.
+[ThreadPool benchmark](benchmarks/2026-08-29-cmp-thread-pool.md). On Windows, use the workflow's
+strict probe and remove `--strict` from the actual cache-off gates. The current Windows LLVM toolchain
+does not emit GNU depfiles. Its CI strict probe allows exactly that known degradation and rejects
+any other warning or error; the actual Windows gates use `--cache=off` without `--strict`, so a full
+rebuild cannot reuse an older BMI or object.
diff --git a/docs/architecture.zh.hant.md b/docs/architecture.zh.hant.md
index 313fcdf..5a410a0 100644
--- a/docs/architecture.zh.hant.md
+++ b/docs/architecture.zh.hant.md
@@ -466,7 +466,7 @@ OneShotEvent、AsyncManualResetEvent 與 AsyncMutex 已形成真實的公開邊
 
 ## 驗證
 
-在儲存庫根目錄執行：
+在 Linux 或 macOS 的儲存庫根目錄執行：
 
 ```text
 mcpp build --profile dev --strict --cache=off
@@ -475,9 +475,11 @@ mcpp build --profile release --strict --cache=off
 mcpp test --profile release --strict --cache=off
 cd examples/basic
 mcpp run
+mcpp build --profile release --strict --cache=off
 ```
 
-預期結果是函式庫建置成功、十個二進位檔中的 161 項測試全部通過，而且範例依序輸出
+預期結果是函式庫和範例的 Dev/Release 建置成功、十個二進位檔中的 161 項測試全部通過，
+而且範例依序輸出
 `Coroutine result: 42`、`Concurrent result: 42`、`Worker pool result: 42`、
 `Blocking result: 42`、`Task group result: 42`、`Recursive group result: 3`、`Event signalled`、
 `Reusable event cycles: 2`、
@@ -488,6 +490,7 @@ mcpp run
 500 次 stop/accept 競態門檻；重點競態套件已連續執行多輪 Release 測試。計算、臨時檔案和回環
 網路的成功/失敗計數及吞吐記錄在
 [v1 可開發性壓測](benchmarks/2026-08-29-cmp-v1-readiness.md)。ThreadPool 的計數、並行和五輪
-Release 資料記錄在[執行緒池壓測](benchmarks/2026-08-29-cmp-thread-pool.md)。目前 Windows LLVM 工具鏈
-不會產生 GNU depfile；如果模組介面包含的檔案發生變更，增量建置可能沿用舊的 BMI 或
-目的檔。完整複驗時使用 `--cache=off`。
+Release 資料記錄在[執行緒池壓測](benchmarks/2026-08-29-cmp-thread-pool.md)。Windows 應使用
+workflow 中的嚴格探針，並在實際無快取門禁中移除 `--strict`。目前 Windows LLVM 工具鏈
+不會產生 GNU depfile；其 CI 嚴格探針只放行這一項已知降級，並拒絕其他 warning 或 error。
+實際 Windows 門禁使用 `--cache=off`，完整重建不會沿用舊的 BMI 或目的檔。
diff --git a/docs/architecture.zh.md b/docs/architecture.zh.md
index b0e2610..5764b26 100644
--- a/docs/architecture.zh.md
+++ b/docs/architecture.zh.md
@@ -463,7 +463,7 @@ OneShotEvent、AsyncManualResetEvent 与 AsyncMutex 已经形成真实的公共
 
 ## 验证
 
-在仓库根目录执行：
+在 Linux 或 macOS 的仓库根目录执行：
 
 ```text
 mcpp build --profile dev --strict --cache=off
@@ -472,9 +472,11 @@ mcpp build --profile release --strict --cache=off
 mcpp test --profile release --strict --cache=off
 cd examples/basic
 mcpp run
+mcpp build --profile release --strict --cache=off
 ```
 
-预期结果是库构建成功、十个二进制中的 161 项测试全部通过，并且示例依次输出
+预期结果是库和示例的 Dev/Release 构建成功、十个二进制中的 161 项测试全部通过，并且
+示例依次输出
 `Coroutine result: 42`、`Concurrent result: 42`、`Worker pool result: 42`、
 `Blocking result: 42`、`Task group result: 42`、`Recursive group result: 3`、`Event signalled`、
 `Reusable event cycles: 2`、
@@ -485,6 +487,7 @@ mcpp run
 stop/accept 竞态门禁；重点竞态套件已连续执行多轮 Release 测试。
 计算、临时文件和回环网络的成功/失败计数及吞吐记录在
 [v1 可开发性压测](benchmarks/2026-08-29-cmp-v1-readiness.md)。ThreadPool 的计数、并发和五轮
-Release 数据记录在[线程池压测](benchmarks/2026-08-29-cmp-thread-pool.md)。当前 Windows LLVM 工具链不会
-生成 GNU depfile；如果模块接口包含的文件发生变化，增量构建可能复用旧的 BMI 或目标文件。
-完整复验时使用 `--cache=off`。
+Release 数据记录在[线程池压测](benchmarks/2026-08-29-cmp-thread-pool.md)。Windows 应使用
+workflow 中的严格探针，并在实际无缓存门禁中去掉 `--strict`。当前 Windows LLVM 工具链
+不会生成 GNU depfile；其 CI 严格探针只放行这一条已知降级，并拒绝其他 warning 或 error。
+实际 Windows 门禁使用 `--cache=off`，完整重建不会复用旧的 BMI 或目标文件。
diff --git a/docs/superpowers/plans/2026-08-30-cmp-phase7-v1-release-qualification.md b/docs/superpowers/plans/2026-08-30-cmp-phase7-v1-release-qualification.md
index 1821e34..8ab7a52 100644
--- a/docs/superpowers/plans/2026-08-30-cmp-phase7-v1-release-qualification.md
+++ b/docs/superpowers/plans/2026-08-30-cmp-phase7-v1-release-qualification.md
@@ -2,7 +2,7 @@
 
 **Date:** 2026-08-30
 **Design:** `docs/superpowers/specs/2026-08-30-cmp-phase7-v1-release-qualification-design.md`
-**Status:** Pending Phase 6C merge and authorized branch transition
+**Status:** PR #12 open; three-platform qualification passed, merge and publication pending
 **Target package:** `mcpplibs.cmp` `0.1.0`
 
 ## Execution Rule
@@ -44,9 +44,15 @@ Correct only factual mismatches.
 Edit only `.github/workflows/ci-linux.yml`, `ci-macos.yml`, and `ci-windows.yml` unless a proven
 platform defect requires a targeted source/test fix.
 
-Replace the default-profile build/test steps with strict cache-off Dev and Release commands. Keep
-the existing pinned bootstrap and platform runner choices. Run the example once and add its strict
-cache-off Release build.
+Replace the default-profile build/test steps with strict cache-off Dev and Release commands on
+Linux and macOS. Keep the existing pinned bootstrap and platform runner choices. Run the example
+once and add its cache-off Release build.
+
+On Windows, mcpp `2026.8.28.1` with LLVM reports the known missing-depfile degradation before tests
+and `--strict` makes it fatal. Run one strict cache-off Dev probe and accept failure only when its
+warning/error set exactly matches that single degradation. Then run the same cache-off Dev/Release
+builds, all tests, example, and Release consumer build without `--strict`. Do not suppress output or
+allow any other diagnostic. If a future mcpp fixes the degradation, the strict probe passes directly.
 
 Do not add a new workflow, matrix abstraction, benchmark job, cache service, sanitizer job, or
 floating dependency.
@@ -56,7 +62,7 @@ review indentation and let GitHub Actions be the authoritative workflow parser.
 
 ## 4. Run local release qualification
 
-From the repository root:
+From the Linux qualification host's repository root:
 
 ```text
 mcpp build --profile dev --strict --cache=off
@@ -127,7 +133,7 @@ Review every changed line for:
 - Windows shell/path assumptions;
 - claims ahead of evidence;
 - generated or local files entering release inputs;
-- weakened test, strictness, or cache behavior;
+- weakened tests, an unbounded strict-mode exception, or cache behavior;
 - remote steps that blur authorization boundaries.
 
 Run whitespace/encoding checks without changing unrelated CRLF files. Update `.agent/HANDOFF.md`
@@ -172,4 +178,38 @@ a new independently reviewed Design.
 
 ## Completion Record
 
-Pending. Fill this section only with commands and remote checks actually completed.
+Local implementation record, 2026-08-30:
+
+- entry gate satisfied: Phase 6C PR #11 passed Linux, macOS, and Windows, merged with
+  authorization, and local `main` was synchronized to `02cb6b9` before creating
+  `release/phase7-v1-release-qualification`;
+- release-surface audit found no public Asio leak, version drift, missing export, or unsupported
+  capability claim;
+- the Linux and macOS workflows run strict cache-off Dev/Release root gates, the example, and its
+  Release build; Windows runs an exact allowlisted strict probe followed by matching cache-off
+  profiles without `--strict`; all three files parse as YAML;
+- local Dev and Release root suites each passed 10 binaries and 161/161 tests; the example run and
+  Release build passed;
+- five readiness rounds preserved compute `49,000/1,000/0`, file I/O `1,000/100/0`, and network
+  `20,000/100/0` success/expected-failure/unexpected-failure counts in every round;
+- the committed Phase 6C package preflight produced a provisional 206,341-byte archive with
+  SHA-256 `fe36976fe76225f6121cce81ca22a1b2392a833795795fb9a00ed52d98dba4de`;
+- its normalized Form A descriptor parsed, and the extracted archive passed strict cache-off Dev
+  and Release root builds and 161 tests, the example run, and its Release build;
+- final local self-review confirmed that only the three existing workflows and release-state
+  documentation changed; YAML structure, command parity, CRLF, and `git diff --check` passed.
+
+Project PR #12 was created with authorization. Its first Windows run proved that a universal
+`--strict` gate stops on mcpp's documented missing-depfile degradation before CMP tests execute;
+that failed run is not release evidence. The extracted replacement probe passes syntax checking,
+runs successfully when strict mode succeeds, accepts a synthetic exact known degradation, and
+rejects a synthetic additional warning.
+
+Correction commit `b2c0911` then passed PR #12 on all three platforms: Linux x86_64 in 2m53s,
+macOS arm64 in 3m08s, and Windows x86_64 in 5m22s. Log-level verification found two suites per
+platform, each with 10 binaries and 161/161 tests, plus the expected ten-line example and Release
+consumer build. Windows also completed the exact strict probe before its cache-off gates.
+
+PR #12 merge, the clean merged archive, tag, GitHub Release, mcpp-index PR, index CI/merge, and the
+isolated indexed consumer check remain pending. Each remote mutation retains the authorization
+boundary defined by the Design.
diff --git a/docs/superpowers/specs/2026-08-30-cmp-phase7-v1-release-qualification-design.md b/docs/superpowers/specs/2026-08-30-cmp-phase7-v1-release-qualification-design.md
index 140efdf..56f134c 100644
--- a/docs/superpowers/specs/2026-08-30-cmp-phase7-v1-release-qualification-design.md
+++ b/docs/superpowers/specs/2026-08-30-cmp-phase7-v1-release-qualification-design.md
@@ -1,9 +1,9 @@
 # CMP Phase 7 v1 Release Qualification Design
 
 **Date:** 2026-08-30
-**Status:** Reviewed design; implementation waits for Phase 6C merge and a new branch
+**Status:** PR #12 open; three-platform qualification passed, merge and publication pending
 **Package version:** `0.1.0`
-**Baseline:** Phase 6C local tree, 161/161 Dev and Release tests
+**Baseline:** Phase 6C PR #11 merged as `02cb6b9`, 161/161 Dev and Release tests
 
 ## Purpose
 
@@ -19,7 +19,8 @@ not yet made the compatibility promise implied by semantic version `1.0.0`.
 Phase 7 includes only:
 
 - freezing the public API implemented through Phase 6C;
-- strict cache-off Dev and Release CI on Linux x86_64, macOS arm64, and Windows x86_64;
+- strict cache-off Dev and Release CI on Linux x86_64 and macOS arm64, plus an allowlisted strict
+  probe and full cache-off Dev/Release CI on Windows x86_64;
 - standalone path-consumer verification in Dev and Release configurations;
 - local readiness and release-package validation;
 - checking package metadata and the generated mcpp-index descriptor;
@@ -41,6 +42,10 @@ Phase 7 implementation starts only after:
 The Design and Plan may exist before that gate. Runtime, CI, version, and release-file changes may
 not be mixed into the Phase 6C implementation branch.
 
+The gate was satisfied on 2026-08-30: PR #11 passed Linux x86_64, macOS arm64, and Windows x86_64,
+was merged with authorization, local `main` was fast-forwarded to `02cb6b9`, and the dedicated
+Phase 7 branch was created.
+
 ## Release Candidate Contract
 
 The release candidate exports the existing root module `mcpplibs.cmp` and keeps all current
@@ -79,8 +84,8 @@ become package inputs.
 
 ## CI Qualification
 
-Keep the three existing platform workflows and their current pinned xlings/mcpp environment. Each
-workflow must execute the same release-relevant commands:
+Keep the three existing platform workflows and their current pinned xlings/mcpp environment.
+Linux and macOS execute these release-relevant commands:
 
 ```text
 mcpp build --profile dev --strict --cache=off
@@ -92,6 +97,20 @@ mcpp run
 mcpp build --profile release --strict --cache=off
 ```
 
+The mcpp `2026.8.28.1` Windows LLVM backend intentionally reports `build/depfile` because it cannot
+emit GNU depfiles. `--strict` converts that known incremental-build degradation into a fatal error
+before the test suite runs. The Windows workflow therefore first executes a strict cache-off Dev
+build as a probe. A failed probe is accepted only when its complete warning/error set is exactly
+the single known depfile degradation and the corresponding strict-mode summary. It then executes
+the same Dev/Release profiles and example flow without `--strict`, retaining `--cache=off` for every
+explicit build/test gate. Any different strict diagnostic, compile failure, test failure, or example
+failure remains fatal. A future mcpp version that fixes Windows depfiles makes the probe pass
+directly without changing this workflow.
+
+This exception does not hide stale objects: cache-off qualification rebuilds from source, while the
+shared manifest remains under strict validation on Linux and macOS. Requiring an impossible Windows
+strict compile would test the build tool's documented limitation instead of CMP.
+
 The example run remains the external import and runtime check. Its Release build adds a consumer
 configuration check without depending on an unstable generated executable path.
 
@@ -144,6 +163,48 @@ The package manifest already owns that information.
 Do not add a second packager, release script, checked-in generated descriptor, or checked-in
 archive. mcpp already owns those transformations.
 
+## Local Qualification Evidence
+
+The Phase 7 branch completed the local gates on 2026-08-30:
+
+- strict cache-off Dev and Release root builds passed; both suites ran 10 binaries and all
+  161 tests passed;
+- the standalone example ran with its expected ten lines and its strict cache-off Release build
+  passed;
+- five Release readiness rounds preserved the exact hard counts in every round: compute
+  `49,000/1,000/0`, file I/O `1,000/100/0`, and network `20,000/100/0` for
+  success/expected failure/unexpected failure;
+- the five-round medians were `38.475 ms / 1,299,546.0 ops/s` for compute,
+  `217.363 ms / 5,060.7 ops/s` for file I/O, and
+  `2,622.017 ms / 7,665.9 ops/s` for network;
+- all three edited workflows parsed as YAML and contain matching profiles and cache-off gates;
+  Windows additionally contains the exact allowlisted strict probe described above;
+- an early `mcpp publish --dry-run --allow-dirty` from committed merged Phase 6C produced
+  `cmp-0.1.0.tar.gz` (206,341 bytes, provisional SHA-256
+  `fe36976fe76225f6121cce81ca22a1b2392a833795795fb9a00ed52d98dba4de`);
+- the normalized temporary descriptor parsed as native-package Form A, and the extracted archive
+  passed strict cache-off Dev and Release root builds and all 161 tests, the example run, and its
+  Release build.
+
+That archive intentionally excludes the dirty Phase 7 workflow and documentation edits. It proves
+the merged Phase 6C package contents and transformation only. The authoritative archive and hash
+must be regenerated from clean merged Phase 7. No runtime source or test file changed in this local
+qualification, so no new analyzer run was required.
+
+## Project PR Evidence
+
+PR #12 evaluated correction commit `b2c0911` on 2026-08-30 and passed all three workflows:
+
+- Linux x86_64 completed in 2m53s;
+- macOS arm64 completed in 3m08s;
+- Windows x86_64 completed in 5m22s, including the exact allowlisted strict probe.
+
+The complete logs show two suites per platform—Dev and Release—and every suite ran 10 test
+binaries with 161/161 tests passing. Each platform also printed the expected ten-line example and
+completed its Release consumer build. The Windows strict probe contained exactly the known missing
+GNU depfile warning and its strict-mode summary; the following cache-off Dev/Release gates passed.
+PR #12 remains open and merge authorization has not been granted.
+
 ## Remote Release Boundary
 
 These are distinct, separately authorized remote operations:
@@ -180,7 +241,8 @@ needed.
 Phase 7 is complete only when all of the following have authoritative evidence:
 
 1. Phase 6C and Phase 7 PRs are merged without unresolved checks;
-2. strict cache-off Dev and Release suites pass all 161 tests on Linux, macOS, and Windows;
+2. strict cache-off Dev and Release suites pass all 161 tests on Linux and macOS; the allowlisted
+   Windows strict probe and cache-off Dev/Release suites pass all 161 tests;
 3. the standalone example runs and its Release consumer build passes on all three platforms;
 4. five final local readiness rounds have zero unexpected failures and exact hard counts;
 5. the clean-tree publish dry-run succeeds;
@@ -195,7 +257,7 @@ index-verified.
 
 ## Self-review Corrections
 
-The reviews corrected seven likely scope errors:
+The reviews corrected eight likely scope errors:
 
 1. Keep `0.1.0`; a first usable capability set is not evidence for a `1.0.0` compatibility promise.
 2. Reuse `mcpp publish --dry-run`; a custom release script or redundant checked-in descriptor adds
@@ -209,7 +271,9 @@ The reviews corrected seven likely scope errors:
 6. Treat mcpp's generated descriptor as a starting point, not index-ready truth: the pinned tool
    emits an invalid inline source segment and the index requires the atomic name plus Form A.
 7. Do not mistake `--allow-dirty` for packaging dirty content. Its archive covers the committed
-   baseline, so final Phase 6C package evidence must wait for a clean merged candidate.
+   baseline, so final Phase 7 package evidence must wait for a clean merged candidate.
+8. Do not require mcpp's known Windows LLVM depfile degradation to disappear. Allowlist it with an
+   exact strict probe, then use cache-off full builds; Linux and macOS retain strict manifest gates.
 
 ## Final Roadmap Boundary