ci: gate merges on all platforms and structure issue intake

Pull requests: `lint (fmt, clippy)` and `pr-size` join the required checks
together with all five `Tests (<os>)` jobs and both `Release workspace` jobs,
so a PR can no longer merge with a red platform. Draft PRs run the three fast
platforms on every push; marking ready for review adds Intel Mac and Windows
ARM64. `pr-size` labels size/S-XL and fails above 1500 changed source lines
unless a maintainer adds `size-exempt`.

Issues: the bug form now asks for OS version, shell, program in the pane,
connection type, non-default settings and mandatory logs/screenshots, with
log paths per platform. `issue-labels` turns the dropdowns into os:/area:
labels, flags regressions, and parks reports without usable steps, evidence
or a version as needs-info with a checklist comment; an edit that fills the
gaps clears the label. config.yml pointed at the old nebula repository.

.github/ISSUE_TEMPLATE/bug_report.yml

@@ -1,114 +1,215 @@
-name: Bug Report
-description: Report a reproducible problem in Nebula
+name: Bug Report / 问题反馈
+description: Report a reproducible problem in Pebrel / 反馈一个可复现的问题
 title: "[Bug] "
 labels: ["bug", "triage"]
 body:
   - type: markdown
     attributes:
       value: |
-        Thanks for taking the time to report a bug. Please search existing issues first and include enough detail for someone else to reproduce the problem.
+        Thanks for taking the time to report a bug. A report we can reproduce gets fixed in the next release; one we cannot reproduce is asked for details and waits.
+        感谢反馈。**能复现的问题会进下个版本；说不清的问题只能先挂起等补充。** 下拉选项会自动转成标签。
 
   - type: dropdown
     id: platform
     attributes:
-      label: Operating system
+      label: Operating system / 操作系统
       options:
         - Windows
+        - Windows (WSL)
         - macOS
         - Linux
-        - Other
+        - Other / 其他
+    validations:
+      required: true
+
+  - type: input
+    id: os_version
+    attributes:
+      label: OS version and display / 系统版本与显示
+      description: Windows 11 23H2 / macOS 26.0 / Ubuntu 24.04; DPI scaling and monitor count if display-related. / 系统版本；渲染问题请附缩放比例与显示器数量。
+      placeholder: "Windows 11 23H2, 150% scaling, 2 monitors"
     validations:
       required: true
 
   - type: input
     id: version
     attributes:
-      label: Nebula version
-      description: Find it in Settings > Home, or on the latest release page.
-      placeholder: "e.g. 1.0.0"
+      label: Pebrel version / 版本号
+      description: Settings > Home, or the release page. / 设置 > 主页，或发布页。
+      placeholder: "e.g. 1.9.0"
     validations:
       required: true
 
   - type: dropdown
     id: install_source
     attributes:
-      label: How did you install Nebula?
+      label: How did you install Pebrel? / 安装方式
       options:
-        - GitHub Release (.msi / .exe / portable archive)
-        - Built from source (cargo build / cargo run)
-        - Other
+        - GitHub Release (.exe / .zip / .dmg / .AppImage / .deb)
+        - Built from source / 源码构建 (cargo build)
+        - Other / 其他
+    validations:
+      required: true
+
+  - type: dropdown
+    id: shell
+    attributes:
+      label: Shell in the affected pane / 出问题的面板用的 Shell
+      multiple: true
+      options:
+        - PowerShell 7 (pwsh)
+        - Windows PowerShell 5.1
+        - cmd
+        - WSL (bash / zsh inside WSL)
+        - Git Bash / MSYS2
+        - zsh
+        - bash
+        - fish
+        - nushell
+        - Not sure / 不确定
+    validations:
+      required: true
+
+  - type: dropdown
+    id: program
+    attributes:
+      label: Program running in the pane / 面板里正在跑的程序
+      description: What was running when it happened? / 出问题时面板里跑着什么？
+      multiple: true
+      options:
+        - Nothing, plain shell prompt / 什么都没跑
+        - Claude Code
+        - Codex CLI
+        - Pi
+        - Gemini CLI / other AI agent — 其他 AI agent
+        - Terminal multiplexer (screen etc.) — 终端复用器
+        - vim / nvim / other full-screen TUI — 全屏 TUI
+        - ssh client / remote shell — 远程 shell
+        - A long-running command (build, ping, adb…) — 长时间运行的命令
+        - Other / 其他
+    validations:
+      required: true
+
+  - type: dropdown
+    id: connection
+    attributes:
+      label: Connection type / 连接方式
+      options:
+        - Local terminal / 本地终端
+        - SSH via Pebrel host list / Pebrel 内 SSH 主机
+        - SSH via command line (ssh user@host) / 手动 ssh 命令
+        - WSL
+        - Other / 其他
     validations:
       required: true
 
   - type: dropdown
     id: area
     attributes:
-      label: Affected area
+      label: Affected area / 影响范围
+      description: Pick every area that applies. / 可多选。
       multiple: true
       options:
-        - Local terminal / shell
-        - SSH connection / authentication
-        - SFTP / file tree
-        - Tabs / splits / sessions
-        - AI assistant / providers
-        - Settings / backup / sync
-        - UI / layout / rendering
-        - Startup / crash
-        - Other
+        - Terminal rendering / scrolling / selection / garbled text — 终端渲染、滚动、选区、乱码
+        - Shell startup / PATH / start directory / context menu — Shell 启动、环境变量、起始目录、右键菜单
+        - SSH / proxy / jump host — SSH、代理、跳板
+        - File tree / drag-and-drop upload — 文件树、拖拽上传
+        - Tabs / splits / session restore / window size — 标签页、分屏、会话恢复、窗口尺寸
+        - AI agent (Claude Code / Codex / Pi / OMP) — AI Agent 集成与状态
+        - Markdown / LaTeX / image reader — 阅读页与内联渲染
+        - Settings / theme / layout / blur — 设置页、主题、布局、模糊背景
+        - Keyboard / IME / mouse / shortcuts / clipboard — 键盘、输入法、鼠标、快捷键、剪贴板
+        - Git / SVN
+        - Startup crash / hang / cannot exit — 启动崩溃、卡死、无法退出
+        - Performance / lag — 性能、卡顿
+        - Installer / update / packaging — 安装包、更新、打包
+        - Other / 其他
+    validations:
+      required: true
+
+  - type: dropdown
+    id: regression
+    attributes:
+      label: Did it work in a previous version? / 旧版本是否正常？
+      options:
+        - Not sure / 不确定
+        - Yes, it worked before / 旧版本正常，新版本坏了
+        - No, it never worked / 一直如此
     validations:
       required: true
 
   - type: dropdown
     id: reproducibility
     attributes:
-      label: Can you reproduce it?
+      label: Can you reproduce it? / 复现概率
       options:
-        - Always (100%)
-        - Often (>50%)
-        - Sometimes
-        - Once / not sure
+        - Always (100%) / 必现
+        - Often (>50%) / 经常
+        - Sometimes / 偶尔
+        - Once / not sure / 仅一次
     validations:
       required: true
 
   - type: textarea
     id: steps
     attributes:
-      label: Steps to reproduce
-      description: Use numbered steps and include the shell or connection type when relevant.
+      label: Steps to reproduce / 复现步骤
+      description: |
+        Numbered, from a fresh launch, ending at the exact moment it goes wrong. "It's laggy" or "it doesn't work" cannot be acted on.
+        从启动开始按序编号，写到出错的那一步为止。"很卡"、"不好用"这种一句话我们没法处理。
       placeholder: |
-        1. Open Nebula
-        2. Create or connect to ...
-        3. ...
+        1. Launch Pebrel, open a new PowerShell 7 tab
+        2. Run `codex` and ask it to ...
+        3. Split the pane with Ctrl+Shift+D
+        4. Output overlaps the input box at the bottom (see screenshot)
     validations:
       required: true
 
   - type: textarea
     id: expected
     attributes:
-      label: Expected behavior
+      label: Expected behavior / 期望结果
     validations:
       required: true
 
   - type: textarea
     id: actual
     attributes:
-      label: Actual behavior
+      label: Actual behavior / 实际结果
+      description: What you see instead, including any error text verbatim. / 实际看到什么，错误文字原样贴。
     validations:
       required: true
 
   - type: textarea
     id: logs
     attributes:
-      label: Logs / screenshots
-      description: Include relevant lines or screenshots. Remove passwords, private keys, tokens, host addresses, and other secrets.
+      label: Logs / screenshots / 日志或截图
+      description: |
+        Required. Screenshot or short recording for anything visual; log lines for crashes, hangs and connection problems. Remove passwords, keys, tokens and host addresses first.
+        必填。显示类问题贴截图或短录屏；崩溃、卡死、连接类问题贴日志。请先删掉密码、私钥、token、主机地址。
+        Logs / 日志位置：
+        - Windows: `%APPDATA%\Pebrel\pebrel_debug.log` and `%TEMP%\Pebrel-<pid>.log`
+        - macOS: `~/Library/Application Support/Pebrel/pebrel_debug.log` and `$TMPDIR/Pebrel-<pid>.log`
+        - Linux: `~/.config/pebrel/pebrel_debug.log` and `/tmp/Pebrel-<pid>.log`
       placeholder: Paste redacted logs or drag screenshots here.
+    validations:
+      required: true
+
+  - type: textarea
+    id: settings
+    attributes:
+      label: Non-default settings / 改过的设置
+      description: Theme, font, transparency/blur, GPU, shell profile, hooks — anything changed from defaults. Attach settings.txt if unsure. / 改过的主题、字体、透明/模糊、GPU、Shell 配置、hooks；不确定就贴 settings.txt。
+      placeholder: "Acrylic blur on, opacity 85%, font JetBrains Mono 13"
 
   - type: checkboxes
     id: checklist
     attributes:
-      label: Before submitting
+      label: Before submitting / 提交前确认
       options:
-        - label: I searched existing issues and did not find a duplicate
+        - label: I searched existing issues and did not find a duplicate / 已搜索，无重复
           required: true
-        - label: I removed passwords, private keys, tokens, and other secrets from this report
+        - label: I tested with the latest release / 已用最新版本试过
+          required: true
+        - label: I removed passwords, private keys, tokens and other secrets / 已删除敏感信息
           required: true

.github/ISSUE_TEMPLATE/config.yml

@@ -1,8 +1,11 @@
 blank_issues_enabled: false
 contact_links:
-  - name: Existing issues
-    url: https://github.com/Kuddev/nebula/issues
-    about: Search existing reports before opening a new issue.
-  - name: Latest release
-    url: https://github.com/Kuddev/nebula/releases/latest
-    about: Check whether the problem is already fixed in the latest version.
+  - name: Existing issues / 已有 issue
+    url: https://github.com/Kuddev/pebrel/issues
+    about: Search existing reports before opening a new issue. / 先搜一下是否已有人反馈。
+  - name: Latest release / 最新版本
+    url: https://github.com/Kuddev/pebrel/releases/latest
+    about: Check whether the problem is already fixed in the latest version. / 看看最新版是否已修复。
+  - name: Discord
+    url: https://discord.gg/VFn4rcxmhn
+    about: Quick questions and discussion. / 快速提问与交流。

.github/ISSUE_TEMPLATE/feature_request.yml

@@ -1,5 +1,5 @@
-name: Feature Request
-description: Suggest an improvement or new capability for Nebula
+name: Feature Request / 功能建议
+description: Suggest an improvement or new capability for Pebrel / 提出改进或新功能
 title: "[Feature] "
 labels: ["enhancement", "triage"]
 body:
@@ -7,12 +7,13 @@ body:
     attributes:
       value: |
         Describe the workflow problem first, then the change you want. Concrete examples make the request easier to evaluate.
+        先说现在哪里不顺手，再说希望怎么改。具体的使用场景最有说服力。
 
   - type: textarea
     id: problem
     attributes:
-      label: Problem / pain point
-      description: What is hard, missing, or repetitive today?
+      label: Problem / pain point — 现在的痛点
+      description: What is hard, missing, or repetitive today? / 什么事难做、缺失、或者每天重复？
       placeholder: When I work with multiple terminals or hosts, I cannot ...
     validations:
       required: true
@@ -20,50 +21,66 @@ body:
   - type: textarea
     id: solution
     attributes:
-      label: Proposed solution
-      description: What would you like Nebula to do?
+      label: Proposed solution — 希望的做法
+      description: What would you like Pebrel to do? / 希望 Pebrel 怎么做？
     validations:
       required: true
 
   - type: textarea
     id: alternatives
     attributes:
-      label: Alternatives considered
-      description: List workarounds, other tools, or designs you considered.
-    validations:
-      required: true
+      label: Alternatives considered — 现在的替代办法
+      description: Workarounds, other tools, or designs you considered. / 目前怎么绕过去的，别的终端怎么做的。
+
+  - type: dropdown
+    id: platform
+    attributes:
+      label: Platform / 平台
+      description: Leave empty if it applies everywhere. / 全平台通用可不选。
+      multiple: true
+      options:
+        - Windows
+        - Windows (WSL)
+        - macOS
+        - Linux
 
   - type: dropdown
     id: area
     attributes:
-      label: Related area
+      label: Related area / 相关范围
       multiple: true
       options:
-        - Local terminal / shell
-        - SSH / SFTP
-        - Tabs / splits / sessions
-        - AI assistant / providers
-        - Settings / backup / sync
-        - UI / UX
-        - Other
+        - Terminal rendering / scrolling / selection — 终端渲染、滚动、选区
+        - Shell startup / PATH / start directory / context menu — Shell 启动、环境变量、起始目录、右键菜单
+        - SSH / proxy / jump host — SSH、代理、跳板
+        - File tree / drag-and-drop upload — 文件树、拖拽上传
+        - Tabs / splits / session restore / window size — 标签页、分屏、会话恢复、窗口尺寸
+        - AI agent (Claude Code / Codex / Pi / OMP) — AI Agent 集成与状态
+        - Markdown / LaTeX / image reader — 阅读页与内联渲染
+        - Settings / theme / layout / blur — 设置页、主题、布局、模糊背景
+        - Keyboard / IME / mouse / shortcuts / clipboard — 键盘、输入法、鼠标、快捷键、剪贴板
+        - Git / SVN
+        - Performance / lag — 性能、卡顿
+        - Installer / update / packaging — 安装包、更新、打包
+        - Other / 其他
     validations:
       required: true
 
   - type: dropdown
     id: priority
     attributes:
-      label: How important is this to you?
+      label: How important is this to you? / 对你有多重要
       options:
-        - Nice to have
-        - Would improve my daily workflow
-        - Blocking / critical for my use case
+        - Nice to have / 有更好
+        - Would improve my daily workflow / 每天都会用到
+        - Blocking / critical for my use case / 没有就没法用
     validations:
       required: true
 
   - type: checkboxes
     id: checklist
     attributes:
-      label: Before submitting
+      label: Before submitting / 提交前确认
       options:
-        - label: I searched existing issues for similar requests
+        - label: I searched existing issues for similar requests / 已搜索，无重复
           required: true

.github/workflows/issue-labels.yml

@@ -0,0 +1,154 @@
+name: Issue labels
+
+# Turns the dropdown answers of the issue forms (.github/ISSUE_TEMPLATE/*.yml)
+# into labels and flags bug reports that are too thin to act on. Labels are
+# only added; maintainers' manual labels are never removed.
+
+on:
+  issues:
+    types: [opened, edited]
+
+permissions:
+  issues: write
+
+concurrency:
+  group: issue-labels-${{ github.event.issue.number }}
+  cancel-in-progress: true
+
+jobs:
+  label-from-form:
+    name: label-from-form
+    runs-on: ubuntu-24.04
+    timeout-minutes: 5
+    steps:
+      - name: Map form answers to labels
+        uses: actions/github-script@v7
+        with:
+          script: |
+            const issue = context.payload.issue;
+            const body = issue.body || "";
+
+            // Issue forms render every field as "### <label>\n\n<value>";
+            // multi-select dropdowns join their values with ", ".
+            const fields = {};
+            const re = /^### (.+?)\s*\n+([\s\S]*?)(?=^### |\s*$(?![\s\S]))/gm;
+            for (const m of body.matchAll(re)) {
+              const value = m[2].trim();
+              if (value && value !== "_No response_") fields[m[1].trim()] = value;
+            }
+            const field = (prefix) => {
+              const key = Object.keys(fields).find((k) => k.startsWith(prefix));
+              return key ? fields[key] : "";
+            };
+            const answers = (prefix) => {
+              const value = field(prefix);
+              return value ? value.split(", ").map((s) => s.trim()) : [];
+            };
+
+            // Match on the English lead of each option so wording tweaks in
+            // the Chinese half never silently break the mapping.
+            const OS = [
+              ["Windows (WSL)", ["os: windows", "os: wsl"]],
+              ["Windows", ["os: windows"]],
+              ["macOS", ["os: macos"]],
+              ["Linux", ["os: linux"]],
+            ];
+            const AREA = [
+              ["Terminal rendering", ["area: terminal"]],
+              ["Shell startup", ["area: shell"]],
+              ["SSH", ["area: ssh"]],
+              ["File tree", ["area: files"]],
+              ["Tabs / splits", ["area: session"]],
+              ["AI agent", ["area: agents"]],
+              ["Markdown", ["area: reader"]],
+              ["Settings", ["area: ui"]],
+              ["Keyboard", ["area: input"]],
+              ["Git", ["area: git"]],
+              ["Startup crash", ["crash"]],
+              ["Performance", ["performance"]],
+              ["Installer", ["area: ci-release"]],
+            ];
+            const PROGRAM = [
+              ["Claude Code", ["area: agents"]],
+              ["Codex", ["area: agents"]],
+              ["Pi", ["area: agents"]],
+              ["Gemini", ["area: agents"]],
+            ];
+            const CONNECTION = [
+              ["SSH", ["area: ssh"]],
+              ["WSL", ["os: wsl"]],
+            ];
+            const SHELL = [["WSL", ["os: wsl"]]];
+            const pick = (values, table) => {
+              const out = [];
+              for (const v of values) {
+                const hit = table.find(([lead]) => v.startsWith(lead));
+                if (hit) out.push(...hit[1]);
+              }
+              return out;
+            };
+
+            const labels = new Set([
+              ...pick([...answers("Operating system"), ...answers("Platform")], OS),
+              ...pick([...answers("Affected area"), ...answers("Related area")], AREA),
+              ...pick(answers("Program running in the pane"), PROGRAM),
+              ...pick(answers("Connection type"), CONNECTION),
+              ...pick(answers("Shell in the affected pane"), SHELL),
+            ]);
+            if (answers("Did it work in a previous version").some((v) => v.startsWith("Yes"))) {
+              labels.add("regression");
+            }
+
+            // Thin bug reports: no usable steps or no evidence. Ask once, on
+            // open; an edit that fills the gaps clears the label.
+            const existing = new Set(issue.labels.map((l) => l.name));
+            const isBug = existing.has("bug");
+            const steps = field("Steps to reproduce");
+            const logs = field("Logs / screenshots");
+            const missing = [];
+            // CJK carries more per character; weigh it double. Numbered
+            // multi-line steps count as an attempt regardless of length.
+            const compact = steps.replace(/\s+/g, "");
+            const weight = compact.length + (compact.match(/[　-鿿豈-﫿]/g) || []).length;
+            const numbered = (steps.match(/^\s*\d+[.、)]/gm) || []).length >= 2;
+            if (isBug && weight < 60 && !numbered) {
+              missing.push("**Steps to reproduce / 复现步骤**: numbered steps from launch to the failing moment / 从启动开始按序写到出错那一步");
+            }
+            if (isBug && !logs) {
+              missing.push("**Logs / screenshots / 日志或截图**: a screenshot for visual issues, `pebrel_debug.log` for crashes, hangs and connection problems / 显示类问题贴截图，崩溃卡死连接类贴日志");
+            }
+            if (isBug && !/\d+\.\d+/.test(field("Pebrel version"))) {
+              missing.push("**Pebrel version / 版本号**: e.g. `1.9.0` (Settings > Home)");
+            }
+
+            const toAdd = [...labels].filter((l) => !existing.has(l));
+            if (missing.length) {
+              if (!existing.has("needs-info")) toAdd.push("needs-info");
+              if (context.payload.action === "opened") {
+                await github.rest.issues.createComment({
+                  ...context.repo,
+                  issue_number: issue.number,
+                  body: [
+                    "Thanks for the report. It is missing details we need before anyone can reproduce it, so it is parked as `needs-info`. Please edit the issue and fill in:",
+                    "感谢反馈。目前缺少复现所需的信息，已标记 `needs-info`。请**编辑 issue** 补齐以下内容，补齐后标签会自动移除：",
+                    "",
+                    ...missing.map((m) => `- ${m}`),
+                    "",
+                    "Reports without these are closed after 14 days of silence. / 14 天无补充将关闭。",
+                  ].join("\n"),
+                });
+              }
+            } else if (isBug && existing.has("needs-info") && context.payload.action === "edited") {
+              await github.rest.issues.removeLabel({
+                ...context.repo, issue_number: issue.number, name: "needs-info",
+              }).catch(() => {});
+            }
+
+            core.info(`form labels: ${[...labels].join(", ") || "(none)"}; adding: ${toAdd.join(", ") || "(none)"}; missing: ${missing.length}`);
+            if (toAdd.length) {
+              await github.rest.issues.addLabels({
+                ...context.repo,
+                issue_number: issue.number,
+                labels: toAdd,
+              });
+            }

.github/workflows/linux-lua.yml

@@ -11,7 +11,9 @@ on:
   push:
     branches: [main]
   # Include docs, catalogs, assets and workflow edits: every PR gets test results.
+  # ready_for_review starts the scarce-runner jobs that draft pushes skip.
   pull_request:
+    types: [opened, synchronize, reopened, ready_for_review]
   merge_group:
 
 concurrency:
@@ -29,13 +31,61 @@ jobs:
     with:
       runtime_run: ${{ inputs.windows_diagnostics_run }}
 
+  lint:
+    if: ${{ inputs.windows_diagnostics_run == '' }}
+    name: lint (fmt, clippy)
+    runs-on: ubuntu-24.04
+    timeout-minutes: 30
+    env:
+      CARGO_TERM_COLOR: always
+    steps:
+      - uses: actions/checkout@v4
+      - name: Install native dependencies
+        run: |
+          sudo apt-get update
+          sudo apt-get install -y --no-install-recommends             build-essential cmake pkg-config libasound2-dev libdbus-1-dev libudev-dev             libfontconfig1-dev libfreetype6-dev libegl1-mesa-dev libgl1-mesa-dev             libwayland-dev libxkbcommon-dev libxkbcommon-x11-dev             libx11-xcb-dev libxcb1-dev libxcb-render0-dev             libxcb-shape0-dev libxcb-xfixes0-dev libxcb-xinput-dev libxcb-xkb-dev libxi-dev
+      - uses: dtolnay/rust-toolchain@stable
+        with:
+          toolchain: "1.97.1"
+          components: rustfmt, clippy
+      - name: rustfmt
+        run: cargo fmt --all -- --check
+      - uses: ./.github/actions/rust-cache
+        id: rust-cache
+        with:
+          key: clippy
+          target: target/ci
+      - name: clippy
+        # Advisory until the existing warning backlog is cleared; then drop
+        # continue-on-error and add -D warnings.
+        continue-on-error: true
+        run: cargo clippy --locked --config .github/ci-profile.toml --profile ci --workspace --all-targets
+      - name: Save compiled workload
+        if: ${{ always() && steps.rust-cache.outcome == 'success' && steps.rust-cache.outputs.cache-hit != 'true' }}
+        uses: actions/cache/save@v4
+        with:
+          path: target/ci
+          key: ${{ steps.rust-cache.outputs.cache-key }}
+
+  # Draft pull requests run the three fast platforms (about 20 minutes) on
+  # every push. The scarce runners (Intel Mac, Windows ARM64) join once the
+  # pull request is marked ready for review and on main, so every merge is
+  # verified on all five platforms without paying their queue time while a
+  # change is still being iterated. Skipped jobs report success, which keeps
+  # all five names usable as required checks; a ready PR cannot merge red.
   native-tests:
     if: ${{ inputs.windows_diagnostics_run == '' }}
     name: Tests (${{ matrix.os }})
     strategy:
       fail-fast: false
       matrix:
-        os: [ubuntu-24.04, windows-2022, windows-11-arm, macos-26, macos-26-intel]
+        os: [ubuntu-24.04, windows-2022, macos-26]
+        tier: [core]
+        include:
+          - os: windows-11-arm
+            tier: full
+          - os: macos-26-intel
+            tier: full
     runs-on: ${{ matrix.os }}
     timeout-minutes: 120
     env:
@@ -45,13 +95,15 @@ jobs:
       RUST_TOOLCHAIN: "1.97.1"
     steps:
       - uses: actions/checkout@v4
+        if: ${{ (matrix.tier != 'full' || github.event_name != 'pull_request' || !github.event.pull_request.draft) }}
         with:
           fetch-depth: 0
       - uses: actions/setup-python@v5
+        if: ${{ (matrix.tier != 'full' || github.event_name != 'pull_request' || !github.event.pull_request.draft) }}
         with:
           python-version: "3.13"
       - name: Install native dependencies
-        if: runner.os == 'Linux'
+        if: ${{ (matrix.tier != 'full' || github.event_name != 'pull_request' || !github.event.pull_request.draft) && (runner.os == 'Linux') }}
         run: |
           sudo apt-get update
           sudo apt-get install -y --no-install-recommends \
@@ -62,15 +114,16 @@ jobs:
             libx11-xcb-dev libxcb1-dev libxcb-render0-dev \
             libxcb-shape0-dev libxcb-xfixes0-dev libxcb-xinput-dev libxcb-xkb-dev libxi-dev
       - name: Require the AppKit SDK used by this PR
-        if: runner.os == 'macOS'
+        if: ${{ (matrix.tier != 'full' || github.event_name != 'pull_request' || !github.event.pull_request.draft) && (runner.os == 'macOS') }}
         run: |
           sdk_version="$(xcrun --sdk macosx --show-sdk-version)"
           test "${sdk_version%%.*}" -ge 26
       - uses: dtolnay/rust-toolchain@stable
+        if: ${{ (matrix.tier != 'full' || github.event_name != 'pull_request' || !github.event.pull_request.draft) }}
         with:
           toolchain: "1.97.1"
       - name: Require a native Windows ARM64 compiler
-        if: matrix.os == 'windows-11-arm'
+        if: ${{ (matrix.tier != 'full' || github.event_name != 'pull_request' || !github.event.pull_request.draft) && (matrix.os == 'windows-11-arm') }}
         shell: pwsh
         run: |
           if ([System.Runtime.InteropServices.RuntimeInformation]::OSArchitecture -ne 'Arm64') {
@@ -81,6 +134,7 @@ jobs:
             throw 'The Windows ARM job must compile and run ARM64 tests.'
           }
       - uses: ./.github/actions/rust-cache
+        if: ${{ (matrix.tier != 'full' || github.event_name != 'pull_request' || !github.event.pull_request.draft) }}
         id: rust-cache
         with:
           key: native
@@ -90,25 +144,26 @@ jobs:
             ${{ runner.os != 'macOS' && 'target/release' || '' }}
           legacy-key: ${{ runner.os == 'Windows' && 'windows' || runner.os == 'Linux' && 'linux' || 'macos' }}
       - name: Prepare pinned Windows console runtime
-        if: runner.os == 'Windows'
+        if: ${{ (matrix.tier != 'full' || github.event_name != 'pull_request' || !github.event.pull_request.draft) && (runner.os == 'Windows') }}
         shell: pwsh
         run: |
           $architecture = if ('${{ runner.arch }}' -eq 'ARM64') { 'arm64' } else { 'x64' }
           ./scripts/prepare-windows-runtime.ps1 -Destination assets/windows/conhost -Architecture $architecture
           ./scripts/tests/windows-runtime.tests.ps1 -Architecture $architecture
       - name: Test complete workspace and native harness
+        if: ${{ (matrix.tier != 'full' || github.event_name != 'pull_request' || !github.event.pull_request.draft) }}
         run: python scripts/ci_native_tests.py
       - name: Check source architecture
-        if: ${{ !cancelled() && runner.os == 'Linux' }}
+        if: ${{ (matrix.tier != 'full' || github.event_name != 'pull_request' || !github.event.pull_request.draft) && (!cancelled() && runner.os == 'Linux') }}
         run: python scripts/check_architecture.py
       - name: Test translation and allocation contracts
-        if: ${{ !cancelled() && runner.os == 'Linux' }}
+        if: ${{ (matrix.tier != 'full' || github.event_name != 'pull_request' || !github.event.pull_request.draft) && (!cancelled() && runner.os == 'Linux') }}
         run: cargo test --locked --manifest-path tools/i18n-contract/Cargo.toml
       - name: Check the release workspace
-        if: ${{ !cancelled() && runner.os != 'macOS' }}
+        if: ${{ (matrix.tier != 'full' || github.event_name != 'pull_request' || !github.event.pull_request.draft) && (!cancelled() && runner.os != 'macOS') }}
         run: cargo check --locked --workspace --release
       - name: Save compiled workload
-        if: ${{ always() && steps.rust-cache.outcome == 'success' && steps.rust-cache.outputs.cache-hit != 'true' }}
+        if: ${{ (matrix.tier != 'full' || github.event_name != 'pull_request' || !github.event.pull_request.draft) && (always() && steps.rust-cache.outcome == 'success' && steps.rust-cache.outputs.cache-hit != 'true') }}
         uses: actions/cache/save@v4
         with:
           path: |
@@ -116,7 +171,7 @@ jobs:
             ${{ runner.os != 'macOS' && 'target/release' || '' }}
           key: ${{ steps.rust-cache.outputs.cache-key }}
       - name: Upload native compilation timings
-        if: always()
+        if: ${{ (matrix.tier != 'full' || github.event_name != 'pull_request' || !github.event.pull_request.draft) && (always()) }}
         uses: actions/upload-artifact@v4
         with:
           name: native-timings-${{ matrix.os }}
@@ -145,32 +200,38 @@ jobs:
       RUST_TOOLCHAIN: "1.97.1"
     steps:
       - uses: actions/checkout@v4
+        if: ${{ (matrix.os != 'macos-26-intel' || github.event_name != 'pull_request' || !github.event.pull_request.draft) }}
       - uses: actions/setup-python@v5
+        if: ${{ (matrix.os != 'macos-26-intel' || github.event_name != 'pull_request' || !github.event.pull_request.draft) }}
         with:
           python-version: "3.13"
       - name: Require the AppKit SDK used by this PR
+        if: ${{ (matrix.os != 'macos-26-intel' || github.event_name != 'pull_request' || !github.event.pull_request.draft) }}
         run: |
           sdk_version="$(xcrun --sdk macosx --show-sdk-version)"
           test "${sdk_version%%.*}" -ge 26
       - uses: dtolnay/rust-toolchain@stable
+        if: ${{ (matrix.os != 'macos-26-intel' || github.event_name != 'pull_request' || !github.event.pull_request.draft) }}
         with:
           toolchain: "1.97.1"
       - uses: ./.github/actions/rust-cache
+        if: ${{ (matrix.os != 'macos-26-intel' || github.event_name != 'pull_request' || !github.event.pull_request.draft) }}
         id: rust-cache
         with:
           key: native-release-check
           target: target/release
           legacy-key: macos
       - name: Check the release workspace
+        if: ${{ (matrix.os != 'macos-26-intel' || github.event_name != 'pull_request' || !github.event.pull_request.draft) }}
         run: cargo check --locked --workspace --release --timings
       - name: Save compiled workload
-        if: ${{ always() && steps.rust-cache.outcome == 'success' && steps.rust-cache.outputs.cache-hit != 'true' }}
+        if: ${{ (matrix.os != 'macos-26-intel' || github.event_name != 'pull_request' || !github.event.pull_request.draft) && always() && steps.rust-cache.outcome == 'success' && steps.rust-cache.outputs.cache-hit != 'true' }}
         uses: actions/cache/save@v4
         with:
           path: target/release
           key: ${{ steps.rust-cache.outputs.cache-key }}
       - name: Upload release check compilation timings
-        if: always()
+        if: ${{ (matrix.os != 'macos-26-intel' || github.event_name != 'pull_request' || !github.event.pull_request.draft) && always() }}
         uses: actions/upload-artifact@v4
         with:
           name: release-check-timings-${{ matrix.os }}

.github/workflows/pr-size.yml

@@ -0,0 +1,81 @@
+name: PR size
+
+# Labels every pull request by the number of changed source lines and fails
+# oversized ones so they get split before review. Generated and documentation
+# files do not count. Maintainers bypass the limit with the `size-exempt`
+# label (a release integration branch is the usual case).
+
+on:
+  pull_request:
+    types: [opened, synchronize, reopened, labeled, unlabeled]
+  # Required checks must also report inside the merge queue; the size was
+  # already enforced on the pull request, so this is a pass-through.
+  merge_group:
+
+permissions:
+  pull-requests: write
+
+concurrency:
+  group: pr-size-${{ github.event.pull_request.number || github.ref }}
+  cancel-in-progress: true
+
+jobs:
+  pr-size:
+    name: pr-size
+    runs-on: ubuntu-24.04
+    timeout-minutes: 5
+    steps:
+      - uses: actions/github-script@v7
+        if: github.event_name == 'pull_request'
+        with:
+          script: |
+            const pr = context.payload.pull_request;
+            const LIMIT = 1500;
+            const IGNORED = [
+              /^Cargo\.lock$/, /(^|\/)package-lock\.json$/, /(^|\/)yarn\.lock$/,
+              /^docs\//, /\.md$/, /\.snap$/, /\.svg$/, /\.png$/, /\.ico$/,
+              /^assets\//, /^locales\//, /\.lock$/,
+            ];
+
+            const files = await github.paginate(github.rest.pulls.listFiles, {
+              ...context.repo, pull_number: pr.number, per_page: 100,
+            });
+            let counted = 0, ignored = 0;
+            for (const f of files) {
+              const lines = f.additions + f.deletions;
+              if (IGNORED.some((re) => re.test(f.filename))) ignored += lines; else counted += lines;
+            }
+
+            const size = counted <= 100 ? "S" : counted <= 500 ? "M" : counted <= LIMIT ? "L" : "XL";
+            const labels = new Set(pr.labels.map((l) => l.name));
+            const wanted = `size/${size}`;
+            for (const l of labels) {
+              if (l.startsWith("size/") && l !== wanted) {
+                await github.rest.issues.removeLabel({ ...context.repo, issue_number: pr.number, name: l }).catch(() => {});
+              }
+            }
+            if (!labels.has(wanted)) {
+              await github.rest.issues.addLabels({ ...context.repo, issue_number: pr.number, labels: [wanted] });
+            }
+
+            core.info(`counted ${counted} lines (${ignored} ignored) in ${files.length} files -> ${wanted}`);
+            if (size !== "XL" || labels.has("size-exempt")) return;
+
+            const marker = "<!-- pr-size-guard -->";
+            const comments = await github.paginate(github.rest.issues.listComments, {
+              ...context.repo, issue_number: pr.number, per_page: 100,
+            });
+            if (!comments.some((c) => c.body && c.body.includes(marker))) {
+              await github.rest.issues.createComment({
+                ...context.repo, issue_number: pr.number,
+                body: [
+                  marker,
+                  `This pull request changes **${counted}** source lines (limit ${LIMIT}; docs, lockfiles and assets excluded). Changes this large cannot be reviewed reliably, so please split it into independent pull requests — one behaviour change each, with its own tests.`,
+                  "",
+                  `本 PR 改动 **${counted}** 行源码（上限 ${LIMIT}，不计文档、lockfile、资源）。这么大的改动无法可靠审查，请拆成多个相互独立的 PR，每个只做一件事并附测试。`,
+                  "",
+                  "Maintainers can add the `size-exempt` label for integration branches. / 维护者可对集成分支加 `size-exempt` 放行。",
+                ].join("\n"),
+              });
+            }
+            core.setFailed(`${counted} changed source lines exceed the ${LIMIT}-line limit; split the pull request or add size-exempt.`);

CONTRIBUTING.md

@@ -74,6 +74,8 @@ raise any existing tracked private artifact for an explicit maintainer decision.
    `architecture/notes/`; ordinary fixes do not create a note.
 2. Keep one conceptual change per PR. A necessary extraction and its behavior
    tests may accompany the feature; unrelated rewrites and formatting may not.
+   The `pr-size` check fails above 1500 changed source lines (docs, lockfiles
+   and assets excluded); split instead of asking for an exemption.
 3. Put shared rules in their existing authority. UI modules adapt those rules;
    they must not fork persistence, state transitions or domain behavior.
 4. Add regression tests that fail for the defect, and test error/cancellation
@@ -122,9 +124,14 @@ compile check into a claim that UI tests or a packaged application were run.
 
 ## Review and enforcement
 
-`architecture-contracts` is the stable PR job name. Maintainers must enable it as
-a required check and require Code Owner approval in the target branch ruleset;
-see the [activation checklist](docs/project-constraints.md#server-side-activation).
+`architecture-contracts`, `lint (fmt, clippy)`, `pr-size`, the five
+`Tests (<os>)` jobs and both `Release workspace (<os>)` jobs are required checks
+on `main`, together with Code Owner approval; see the
+[activation checklist](docs/project-constraints.md#server-side-activation).
+Open pull requests as drafts while iterating: drafts run Linux, Windows x64 and
+Apple Silicon on every push (about 20 minutes). Marking the PR ready for review
+adds Intel Mac and Windows ARM64, so every merge is verified on all five
+platforms without paying for the scarce runners on each push.
 Local hooks are convenient, but bypassable; they are not the enforcement boundary.
 Submitting a workflow or `CODEOWNERS` file does not configure server-side rules.
 
@@ -137,6 +144,8 @@ feature work; there is no routine `--skip-architecture` option.
 ## 中文摘要
 
 - 先读架构图、工程合同和决策记录；按职责拆分，不按行号切片。
+- 一个 PR 只做一件事；改动超过 1500 行源码（不计文档、lockfile、资源）`pr-size` 会失败，请拆分。
+- 迭代期间用 Draft PR：每次 push 跑 Linux / Windows x64 / Apple Silicon（约 20 分钟）；标记 Ready for review 后补跑 Intel Mac 与 Windows ARM64，十项必需检查全绿才能合并。
 - 2000 行是现有仓库的防灾上限，800 行只提示审查，不是“大厂标准”。
 - 普通功能 PR 不得增加存量债务；有问题的规则可以修订，但要有反例、测试和维护者审批。
 - 新增核心抽象、依赖方向、持久化或线程模型改变要先说明设计，不强迫每个小修复写 ADR。