fix(onboarding): clear session before skip_existing completion (MUL-2166)

Emacs review: handleWelcomeSkip ran after startOnboardingSession had
already fired during the shell's mount effect, so the skip_existing
onboarding_completed event was carrying the session id we documented
it would omit. Clear the session before completing so the event
matches the doc and HogQL IS NOT NULL filter isolates real funnel
completions cleanly.

Co-authored-by: multica-agent <github@multica.ai>

.env.example

@@ -101,6 +101,13 @@ ALLOWED_ORIGINS=
 # `Authorization: Bearer <token>`.
 # REALTIME_METRICS_TOKEN=
 
+# GitHub App integration (Settings → Integrations "Connect GitHub")
+# Both must be set for the Connect button to enable and for webhooks to be
+# accepted; leave empty to disable the integration. See docs/github-integration.
+# GITHUB_APP_SLUG is the tail of https://github.com/apps/<slug>.
+GITHUB_APP_SLUG=
+GITHUB_WEBHOOK_SECRET=
+
 # Frontend
 FRONTEND_PORT=3000
 FRONTEND_ORIGIN=http://localhost:3000

README.md

@@ -20,7 +20,7 @@ Turn coding agents into real teammates — assign tasks, track progress, compoun
 [![CI](https://github.com/multica-ai/multica/actions/workflows/ci.yml/badge.svg)](https://github.com/multica-ai/multica/actions/workflows/ci.yml)
 [![GitHub stars](https://img.shields.io/github/stars/multica-ai/multica?style=flat)](https://github.com/multica-ai/multica/stargazers)
 
-[Website](https://multica.ai) · [Cloud](https://multica.ai/app) · [X](https://x.com/MulticaAI) · [Self-Hosting](SELF_HOSTING.md) · [Contributing](CONTRIBUTING.md)
+[Website](https://multica.ai) · [Cloud](https://multica.ai) · [X](https://x.com/MulticaAI) · [Self-Hosting](SELF_HOSTING.md) · [Contributing](CONTRIBUTING.md)
 
 **English | [简体中文](README.zh-CN.md)**
 

README.zh-CN.md

@@ -20,7 +20,7 @@
 [![CI](https://github.com/multica-ai/multica/actions/workflows/ci.yml/badge.svg)](https://github.com/multica-ai/multica/actions/workflows/ci.yml)
 [![GitHub stars](https://img.shields.io/github/stars/multica-ai/multica?style=flat)](https://github.com/multica-ai/multica/stargazers)
 
-[官网](https://multica.ai) · [云服务](https://multica.ai/app) · [X](https://x.com/MulticaAI) · [自部署指南](SELF_HOSTING.md) · [参与贡献](CONTRIBUTING.md)
+[官网](https://multica.ai) · [云服务](https://multica.ai) · [X](https://x.com/MulticaAI) · [自部署指南](SELF_HOSTING.md) · [参与贡献](CONTRIBUTING.md)
 
 **[English](README.md) | 简体中文**
 

apps/desktop/electron-builder.yml

@@ -46,20 +46,31 @@ linux:
   # Yaru). Forcing `multica` makes every Linux identity slot agree and
   # matches `StartupWMClass=Multica` (productName-derived).
   executableName: multica
-  # Pin StartupWMClass explicitly to the WM_CLASS that Electron emits on
-  # X11. Electron derives WM_CLASS from `app.getName()`, which in packaged
-  # builds resolves to `productName` (`Multica`). Without an explicit
-  # `StartupWMClass`, electron-builder writes `productName` as the default
-  # — making this declaration redundant with current settings — but
-  # pinning the value here turns a silent future drift (e.g. if anyone
-  # renames productName or sets app.setName at boot) into a visible diff
-  # against this file. The WM_CLASS ↔ StartupWMClass match is what lets
-  # GNOME associate the running window with the `.desktop` entry and
-  # therefore render the right icon. The post-build verification step in
-  # PR #2437 is `xprop WM_CLASS` on a real Ubuntu install.
+  # Pin StartupWMClass to the WM_CLASS Electron emits on X11. Electron
+  # derives WM_CLASS from `app.getName()`, which reads the *packaged*
+  # ASAR's `package.json` — `productName` if present, otherwise `name`.
+  # PR #2437 assumed electron-builder.yml's productName fed app.getName()
+  # directly; it does not. With our source package.json carrying only
+  # `name: "@multica/desktop"`, packaged Electron emitted
+  # `WM_CLASS=@multica/desktop`, which broke association with this entry
+  # and reproduced #2515 on Ubuntu 0.2.31. The fix lives in two places
+  # outside this file — `productName: "Multica"` on the source
+  # package.json (so the ASAR carries it) and `app.setName("Multica")`
+  # in the production branch of `src/main/index.ts` (belt-and-braces).
+  # Keep `StartupWMClass: Multica` pinned here so any future drift in
+  # those two anchors shows up as a diff against this declaration.
+  # Verification on a real Ubuntu install: `xprop WM_CLASS` on a running
+  # window prints `Multica` for both fields.
   desktop:
     entry:
       StartupWMClass: Multica
+  # Point at pre-rendered hicolor sizes. electron-builder *can* generate
+  # 16/24/32/48/64/128/256/512 from a single build/icon.png, but the
+  # auto-generation silently shipped only the 1024×1024 source in our
+  # v0.2.31 .deb (#2515 reproduces this) — leaving GNOME's hicolor lookup
+  # with no usable size and falling back to the theme default. Shipping
+  # the sizes from source removes the toolchain dependency entirely.
+  icon: build/icons
   target:
     - AppImage
     - deb

apps/desktop/eslint.config.mjs

@@ -10,10 +10,11 @@ export default [
       globals: { ...globals.node },
     },
   },
-  // Security: every renderer-controlled URL that reaches the OS shell must
-  // flow through openExternalSafely in src/main/external-url.ts (scheme
-  // allowlist). Enforce it statically so a direct shell.openExternal call
-  // cannot silently regress the protection.
+  // Security: every renderer-controlled URL that reaches the OS shell or the
+  // native download system must flow through the safe wrappers in
+  // src/main/external-url.ts (scheme allowlist). Enforce it statically so
+  // direct shell.openExternal / webContents.downloadURL calls cannot silently
+  // regress the protection.
   {
     files: ["src/main/**/*.ts"],
     rules: {
@@ -25,6 +26,12 @@ export default [
           message:
             "Do not call shell.openExternal directly. Use openExternalSafely from './external-url' so the http/https allowlist stays enforced.",
         },
+        {
+          selector:
+            "CallExpression[callee.object.property.name='webContents'][callee.property.name='downloadURL']",
+          message:
+            "Do not call webContents.downloadURL directly. Use downloadURLSafely from './external-url' so the http/https allowlist stays enforced.",
+        },
       ],
     },
   },

apps/desktop/package.json

@@ -1,5 +1,6 @@
 {
   "name": "@multica/desktop",
+  "productName": "Multica",
   "version": "0.1.0",
   "private": true,
   "description": "Multica Desktop — native desktop client for the Multica platform.",

apps/desktop/src/main/external-url.ts

@@ -1,4 +1,4 @@
-import { shell } from "electron";
+import { shell, type BrowserWindow } from "electron";
 
 // True when the URL parses and uses http/https — the only schemes we let
 // reach `shell.openExternal`. Scheme comparison is safe because the WHATWG
@@ -19,6 +19,19 @@ export function openExternalSafely(url: string): Promise<void> | void {
   return shell.openExternal(url);
 }
 
+// Canonical wrapper around webContents.downloadURL. All renderer-controlled
+// URLs that trigger a native download MUST flow through here; direct calls
+// to `webContents.downloadURL` elsewhere in the main process are banned by
+// the no-restricted-syntax rule in apps/desktop/eslint.config.mjs.
+// Reuses the same http/https allowlist as openExternalSafely.
+export function downloadURLSafely(win: BrowserWindow, url: string): void {
+  if (getHttpProtocol(url) === null) {
+    console.warn(`[security] blocked downloadURL: ${describeScheme(url)}`);
+    return;
+  }
+  win.webContents.downloadURL(url);
+}
+
 function getHttpProtocol(url: string): "http:" | "https:" | null {
   try {
     const { protocol } = new URL(url);

apps/desktop/src/main/index.ts

@@ -5,7 +5,7 @@ import { electronApp, optimizer, is } from "@electron-toolkit/utils";
 import fixPath from "fix-path";
 import { setupAutoUpdater } from "./updater";
 import { setupDaemonManager } from "./daemon-manager";
-import { openExternalSafely } from "./external-url";
+import { openExternalSafely, downloadURLSafely } from "./external-url";
 import { installContextMenu } from "./context-menu";
 import { getAppVersion } from "./app-version";
 import { loadRuntimeConfig } from "./runtime-config-loader";
@@ -133,6 +133,27 @@ function createWindow(): void {
       preload: join(__dirname, "../preload/index.js"),
       sandbox: false,
       webSecurity: false,
+      // Required for the Chromium PDF viewer (PDFium) to activate inside
+      // iframes — used by the attachment preview modal for application/pdf
+      // files. Default is false in Electron; without it <iframe src=*.pdf>
+      // renders blank.
+      //
+      // Security trade-off, accepted intentionally:
+      //   1. This window already runs with `webSecurity: false` + `sandbox: false`,
+      //      so `plugins: true` does NOT meaningfully widen the renderer's
+      //      attack surface beyond what is already accepted.
+      //   2. The only PDFs that reach an iframe here are signed CloudFront URLs
+      //      we ourselves issued (see useDownloadAttachment); user-supplied URLs
+      //      are routed through `setWindowOpenHandler` → `openExternalSafely` and
+      //      cannot land in this renderer.
+      //   3. Chromium's PDFium plugin is itself sandboxed inside its own process
+      //      and only handles the `application/pdf` MIME — it does not expose
+      //      Flash, Java, or other historical plugin surfaces.
+      //
+      // If we ever tighten `webSecurity` / `sandbox`, revisit this by hosting
+      // the PDF viewer in a dedicated BrowserView with `plugins: true` scoped
+      // to that view, keeping the main renderer plugin-free.
+      plugins: true,
       additionalArguments: [`--multica-locale=${systemLocale}`],
     },
   });
@@ -212,6 +233,14 @@ const DEV_APP_NAME = process.env.DESKTOP_APP_SUFFIX
 if (is.dev) {
   app.setName(DEV_APP_NAME);
   app.setPath("userData", join(app.getPath("appData"), DEV_APP_NAME));
+} else {
+  // Pin the production app name in code. Electron's Linux WM_CLASS is set
+  // from app.getName() when the first BrowserWindow is realized; the
+  // packaged ASAR's package.json `productName` already steers app.getName()
+  // to "Multica", but anchoring it here makes WM_CLASS ↔ StartupWMClass
+  // (declared in electron-builder.yml) survive a regression in
+  // productName / the build pipeline. Must run before requestSingleInstanceLock().
+  app.setName("Multica");
 }
 
 // --- Protocol registration -----------------------------------------------
@@ -288,6 +317,14 @@ if (!gotTheLock) {
       return openExternalSafely(url);
     });
 
+    ipcMain.handle("file:download-url", (_event, url: string) => {
+      if (!mainWindow) {
+        console.warn("[download] ignored file:download-url — mainWindow torn down");
+        return;
+      }
+      downloadURLSafely(mainWindow, url);
+    });
+
     // Sync IPC: app version + normalized OS for preload. Sync (not invoke) so
     // preload can attach the values to `desktopAPI.appInfo` before any renderer
     // code reads them, ensuring the very first HTTP request from the renderer

apps/desktop/src/preload/index.d.ts

@@ -19,6 +19,9 @@ interface DesktopAPI {
   onInviteOpen: (callback: (invitationId: string) => void) => () => void;
   /** Open a URL in the default browser. */
   openExternal: (url: string) => Promise<void>;
+  /** Download a file by URL through Electron's native download system.
+   *  Shows a native save dialog. On non-desktop platforms this is undefined. */
+  downloadURL: (url: string) => Promise<void>;
   /** Hide macOS traffic lights for full-screen modals; restore when false. */
   setImmersiveMode: (immersive: boolean) => Promise<void>;
   /** Show a native OS notification for a new inbox item. */

apps/desktop/src/preload/index.ts

@@ -89,6 +89,11 @@ const desktopAPI = {
   },
   /** Open a URL in the default browser */
   openExternal: (url: string) => ipcRenderer.invoke("shell:openExternal", url),
+  /** Download a file by URL through Electron's native download system.
+   *  Shows a save dialog and saves to disk. Unlike openExternal, this
+   *  avoids browser rendering of HTML files on Linux.
+   *  On non-desktop platforms this property is undefined. */
+  downloadURL: (url: string) => ipcRenderer.invoke("file:download-url", url),
   /** Toggle immersive mode — hide macOS traffic lights for full-screen modals */
   setImmersiveMode: (immersive: boolean) =>
     ipcRenderer.invoke("window:setImmersive", immersive),

apps/desktop/src/renderer/src/routes.tsx

@@ -14,11 +14,13 @@ import { AgentDetailPage } from "./pages/agent-detail-page";
 import { RuntimeDetailPage } from "./pages/runtime-detail-page";
 import { IssuesPage } from "@multica/views/issues/components";
 import { ProjectsPage } from "@multica/views/projects/components";
+import { DashboardPage } from "@multica/views/dashboard";
 import { AutopilotsPage } from "@multica/views/autopilots/components";
 import { MyIssuesPage } from "@multica/views/my-issues";
 import { SkillsPage } from "@multica/views/skills";
 import { DesktopRuntimesPage } from "./components/desktop-runtimes-page";
 import { AgentsPage } from "@multica/views/agents";
+import { SquadsPage, SquadDetailPage as SquadDetailPageView } from "@multica/views/squads/components";
 import { InboxPage } from "@multica/views/inbox";
 import { SettingsPage } from "@multica/views/settings";
 import { ErrorBoundary } from "@multica/ui/components/common/error-boundary";
@@ -145,7 +147,18 @@ export const appRoutes: RouteObject[] = [
             element: <AgentDetailPage />,
             handle: { title: "Agent" },
           },
+          { path: "squads", element: <SquadsPage />, handle: { title: "Squads" } },
+          {
+            path: "squads/:id",
+            element: <SquadDetailPageView />,
+            handle: { title: "Squad" },
+          },
           { path: "inbox", element: <InboxPage />, handle: { title: "Inbox" } },
+          {
+            path: "usage",
+            element: <DashboardPage />,
+            handle: { title: "Usage" },
+          },
           {
             path: "settings",
             element: (

apps/desktop/src/renderer/src/stores/tab-store.test.ts

@@ -180,6 +180,61 @@ describe("useTabStore actions", () => {
     expect(s.byWorkspace.acme.tabs[0].id).not.toBe(onlyTabId); // fresh tab
   });
 
+  it("defers disposing the closed tab router until after the store update", () => {
+    vi.useFakeTimers();
+    try {
+      const store = useTabStore.getState();
+      store.switchWorkspace("acme");
+      const closedTabId = store.addTab("/acme/settings", "Settings", "Settings");
+      const closingTab = useTabStore
+        .getState()
+        .byWorkspace.acme.tabs.find((t) => t.id === closedTabId);
+      const dispose = vi.mocked(closingTab!.router.dispose);
+
+      store.closeTab(closedTabId);
+
+      expect(dispose).not.toHaveBeenCalled();
+      expect(
+        useTabStore.getState().byWorkspace.acme.tabs.some((t) => t.id === closedTabId),
+      ).toBe(false);
+
+      vi.runAllTimers();
+
+      expect(dispose).toHaveBeenCalledOnce();
+    } finally {
+      vi.useRealTimers();
+    }
+  });
+
+  it("ignores router-sync updates from a tab after it has been closed", () => {
+    const store = useTabStore.getState();
+    store.switchWorkspace("acme");
+    const closedTabId = store.addTab("/acme/settings", "Settings", "Settings");
+
+    store.closeTab(closedTabId);
+    const before = useTabStore.getState().byWorkspace.acme;
+
+    store.updateTab(closedTabId, { path: "/acme/runtimes", icon: "Monitor" });
+    store.updateTabHistory(closedTabId, 1, 2);
+
+    expect(useTabStore.getState().byWorkspace.acme).toBe(before);
+    expect(
+      useTabStore.getState().byWorkspace.acme.tabs.some((t) => t.id === closedTabId),
+    ).toBe(false);
+  });
+
+  it("does not replace the tab group for no-op router-sync updates", () => {
+    const store = useTabStore.getState();
+    store.switchWorkspace("acme");
+    const tab = useTabStore.getState().byWorkspace.acme.tabs[0];
+    const before = useTabStore.getState().byWorkspace.acme;
+
+    store.updateTab(tab.id, { path: tab.path, icon: tab.icon, title: tab.title });
+    store.updateTabHistory(tab.id, tab.historyIndex, tab.historyLength);
+
+    expect(useTabStore.getState().byWorkspace.acme).toBe(before);
+  });
+
   it("validateWorkspaceSlugs drops groups for slugs not in the valid set and repoints active", () => {
     const store = useTabStore.getState();
     store.switchWorkspace("acme");

apps/desktop/src/renderer/src/stores/tab-store.ts

@@ -350,7 +350,10 @@ export const useTabStore = create<TabStore>()(
         const { slug, group, index } = hit;
 
         const closing = group.tabs[index];
-        closing.router.dispose();
+        const disposeClosingRouter = () => {
+          // Let React unmount the tab's RouterProvider before disposing it.
+          window.setTimeout(() => closing.router.dispose(), 0);
+        };
 
         if (group.tabs.length === 1) {
           // Last tab in this workspace — reseed a default so the workspace
@@ -363,6 +366,7 @@ export const useTabStore = create<TabStore>()(
               [slug]: { tabs: [fresh], activeTabId: fresh.id },
             },
           });
+          disposeClosingRouter();
           return;
         }
 
@@ -378,6 +382,7 @@ export const useTabStore = create<TabStore>()(
             [slug]: { tabs: nextTabs, activeTabId: nextActiveTabId },
           },
         });
+        disposeClosingRouter();
       },
 
       setActiveTab(tabId) {
@@ -402,6 +407,13 @@ export const useTabStore = create<TabStore>()(
         const { slug, group, index } = hit;
         const current = group.tabs[index];
         const next: Tab = { ...current, ...patch };
+        if (
+          next.path === current.path &&
+          next.title === current.title &&
+          next.icon === current.icon
+        ) {
+          return;
+        }
         const nextTabs = [...group.tabs];
         nextTabs[index] = next;
         set({
@@ -418,6 +430,12 @@ export const useTabStore = create<TabStore>()(
         if (!hit) return;
         const { slug, group, index } = hit;
         const current = group.tabs[index];
+        if (
+          current.historyIndex === historyIndex &&
+          current.historyLength === historyLength
+        ) {
+          return;
+        }
         const next: Tab = { ...current, historyIndex, historyLength };
         const nextTabs = [...group.tabs];
         nextTabs[index] = next;

apps/docs/content/docs/developers/conventions.mdx

@@ -160,6 +160,7 @@ Chinese term reference:
 | Confirm / Continue / Back | 确认 / 继续 / 返回 |
 | Edit / New / Create / Add | 编辑 / 新建 / 创建 / 添加 |
 | Remove / Send / Open / Close | 移除 / 发送 / 打开 / 关闭 |
+| Preview / Download / Upload | 预览 / 下载 / 上传 |
 | Done / Loading... | 完成 / 加载中... |
 | Profile / Account / Appearance | 个人资料 / 账号 / 外观 |
 | Theme / Language | 主题 / 语言 |

apps/docs/content/docs/developers/conventions.zh.mdx

@@ -160,6 +160,7 @@ Multica 的产品名词分两类：
 | Confirm / Continue / Back | 确认 / 继续 / 返回 |
 | Edit / New / Create / Add | 编辑 / 新建 / 创建 / 添加 |
 | Remove / Send / Open / Close | 移除 / 发送 / 打开 / 关闭 |
+| Preview / Download / Upload | 预览 / 下载 / 上传 |
 | Done / Loading... | 完成 / 加载中... |
 | Profile / Account / Appearance | 个人资料 / 账号 / 外观 |
 | Theme / Language | 主题 / 语言 |

apps/docs/content/docs/environment-variables.mdx

@@ -141,6 +141,22 @@ For a full explanation of how each parameter affects daemon behavior, see [Daemo
 **Leaving `FRONTEND_ORIGIN` unset creates two silent failures**: (1) invite email links point at `https://app.multica.ai` (the hosted domain), and clicking them doesn't bring users back to your self-hosted instance; (2) WebSocket Origin checks fall back to `localhost:3000 / 5173 / 5174`, so every WebSocket connection in a production deployment is rejected and the frontend appears to "lose real-time updates."
 </Callout>
 
+## GitHub integration
+
+The [GitHub PR ↔ issue integration](/github-integration) needs two variables. Set both to enable Connect GitHub in Settings and accept incoming webhooks.
+
+| Variable | Default | Description |
+|---|---|---|
+| `GITHUB_APP_SLUG` | empty | The slug of your GitHub App (the tail of `https://github.com/apps/<slug>`). Drives the Settings → Integrations install button URL |
+| `GITHUB_WEBHOOK_SECRET` | empty | The Webhook secret you set on the GitHub App. Used for HMAC-SHA256 verification of every `pull_request` / `installation` delivery, and as the HMAC key for the setup-callback state token |
+
+**Behavior when either is unset:**
+
+- `Connect GitHub` in Settings → Integrations is **disabled** and shows a "not configured" hint to admins.
+- The `/api/webhooks/github` endpoint returns **`503 github webhooks not configured`** — Multica refuses to process events with no secret rather than treating every signature as valid.
+
+**Note:** `GITHUB_WEBHOOK_SECRET` is reused as the signing key for the install-flow state token, so operators only need to manage one secret. It is **not** the GitHub App's *Client* secret — Client secrets are OAuth-related and not used by this integration. See [GitHub integration → Self-host setup](/github-integration#self-host-setup) for the full walkthrough.
+
 ## Usage analytics
 
 By default, the server reports to Multica's official PostHog instance. To opt out, set `ANALYTICS_DISABLED=true`.
@@ -154,5 +170,6 @@ By default, the server reports to Multica's official PostHog instance. To opt ou
 ## Next
 
 - [Sign-in and signup configuration](/auth-setup) — how to actually configure the auth-related variables above and where the traps are
+- [GitHub integration](/github-integration) — how to set up the GitHub App that backs `GITHUB_APP_SLUG` / `GITHUB_WEBHOOK_SECRET`
 - [Troubleshooting](/troubleshooting) — symptoms and fixes for common misconfigurations
 - [Daemon and runtimes](/daemon-runtimes) — what the `MULTICA_DAEMON_*` parameters actually do

apps/docs/content/docs/environment-variables.zh.mdx

@@ -141,6 +141,22 @@ Multica 存储用户上传的附件（评论里的图片、文件等）。**优
 **`FRONTEND_ORIGIN` 不设就有两个静默失败**：（1）邀请邮件里的链接指向 `https://app.multica.ai`（托管版的域名），用户点了跳不回你的 self-host 实例；（2）WebSocket 连接的 Origin 校验回落到 `localhost:3000 / 5173 / 5174`，生产部署的 WebSocket 全部被拒，前端看起来「实时更新不工作」。
 </Callout>
 
+## GitHub 集成
+
+[GitHub PR ↔ issue 集成](/github-integration) 依赖两个环境变量。两个都配上才会启用 Settings 里的 Connect GitHub 并接受 webhook。
+
+| 环境变量 | 默认值 | 说明 |
+|---|---|---|
+| `GITHUB_APP_SLUG` | 空 | 你的 GitHub App slug（`https://github.com/apps/<slug>` 的尾部）。Settings → Integrations 里安装按钮的跳转 URL 用它拼 |
+| `GITHUB_WEBHOOK_SECRET` | 空 | 你在 GitHub App 上设置的 Webhook secret。每条 `pull_request` / `installation` delivery 都用它做 HMAC-SHA256 校验；同一个值也用作 setup 回调里 state token 的签名密钥 |
+
+**任一变量未设时：**
+
+- Settings → Integrations 里 `Connect GitHub` 按钮 **disable**，对 admin 显示「not configured」提示
+- `/api/webhooks/github` 直接返回 **`503 github webhooks not configured`**——secret 没配置时 Multica 拒绝处理任何 webhook 事件，而不是把所有签名当 valid
+
+**注意：** `GITHUB_WEBHOOK_SECRET` 同时被复用为 install 流程里 state token 的签名密钥，所以运维只需要维护一个 secret。它**不是** GitHub App 的 *Client* secret——Client secret 是 OAuth 用的，和本集成无关。完整配置流程见 [GitHub 集成 → Self-Host 配置](/github-integration#self-host-配置)。
+
 ## 用量统计
 
 默认上报到 Multica 官方 PostHog 实例。不想上报就把 `ANALYTICS_DISABLED=true`。
@@ -154,5 +170,6 @@ Multica 存储用户上传的附件（评论里的图片、文件等）。**优
 ## 下一步
 
 - [登录与注册配置](/auth-setup) —— 上面 auth 相关的那几个环境变量怎么真的配、陷阱在哪
+- [GitHub 集成](/github-integration) —— `GITHUB_APP_SLUG` / `GITHUB_WEBHOOK_SECRET` 背后的 GitHub App 怎么建
 - [故障排查](/troubleshooting) —— 配错了常见的症状和修复
 - [守护进程与运行时](/daemon-runtimes) —— `MULTICA_DAEMON_*` 参数的行为含义

apps/docs/content/docs/github-integration.mdx

@@ -0,0 +1,183 @@
+---
+title: GitHub integration
+description: Connect a GitHub App once, then PRs whose branch, title, or body reference an issue identifier auto-attach to that issue — and merging the PR moves the issue to Done.
+---
+
+import { Callout } from "fumadocs-ui/components/callout";
+
+Connect a GitHub account or organization once in **Settings → Integrations**. After that, any pull request whose branch name, title, or body contains an issue identifier (for example `MUL-123`) is **auto-linked** to that [issue](/issues), appears under **Pull requests** in the issue sidebar, and — when the PR is merged — moves the issue to **Done**.
+
+There is no per-issue setup. The whole flow is identifier-driven.
+
+## What the integration does
+
+| Surface | Behavior |
+|---|---|
+| **Settings → Integrations** | Workspace admins see a GitHub card with a **Connect GitHub** button. Clicking it opens GitHub's App install page; after install you bounce back to Settings. |
+| **Issue sidebar → Pull requests** | Every PR auto-linked to this issue, with title, repo, state (`Open` / `Draft` / `Merged` / `Closed`), and author. Click a row to jump to the PR on GitHub. |
+| **Webhook (background)** | On every `pull_request` event, Multica upserts the PR row, scans the PR for issue identifiers, and (re)builds the link rows. Idempotent — replaying a delivery is a no-op. |
+| **Auto-status on merge** | When a PR transitions to `merged`, every linked issue not already `Done` or `Cancelled` is moved to `Done`. The status change is timeline-logged with source `github_pr_merged`. |
+
+Only the PR itself is mirrored. Commits, branch refs without an open PR, and CI check states are **not** modeled. The integration is intentionally narrow.
+
+## How identifiers are matched
+
+The webhook extracts identifiers from three fields, in this order: **PR head branch**, **PR title**, **PR body**. The matcher is:
+
+- Case-insensitive — `mul-123`, `MUL-123`, `Mul-123` all match.
+- Bounded — a `\b` on the left and a digit anchor on the right keep it from grabbing version numbers like `v1.2-3` or email-style strings.
+- Workspace-scoped — only matches the workspace's own [issue prefix](/workspaces). `FOO-1` in a workspace whose prefix is `MUL` is ignored, even if the integer matches another issue.
+- Deduplicated — listing `MUL-1, MUL-1` in the body links the issue once.
+
+You can reference **multiple issues** in one PR. `Closes MUL-1, MUL-2` links the PR to both, and merging it advances both to `Done`.
+
+## The auto-merge-to-Done rule
+
+When a PR's `merged` field flips to `true`, every linked issue is evaluated:
+
+| Issue current status | Result |
+|---|---|
+| `done` | No change (already terminal). |
+| `cancelled` | **No change** — cancelled means the user explicitly abandoned the work; the integration does not override that signal. |
+| Anything else (`todo`, `in_progress`, `in_review`, `blocked`, `backlog`) | Moved to `done`. |
+
+Closing a PR **without** merging it only updates the PR card's state to `Closed`. The linked issues stay where they were — the user is the one who decides what closing-without-merge means.
+
+<Callout type="info">
+The action is attributed to the `system` actor on the timeline. Subscribers of the issue receive an inbox notification for the status change, the same way they would if a human had moved it.
+</Callout>
+
+## What's not auto-linked
+
+- **Identifiers in commit messages** — only branch / title / body are scanned. A commit titled `MUL-123: fix login` does not auto-link unless the same string also appears in the PR title or body.
+- **Identifiers in PR comments** — only the PR's own metadata is scanned; later GitHub comments are ignored.
+- **PRs in repos the App isn't installed on** — without the App, Multica never receives the webhook.
+- **Manually linking a PR to an issue** — there is no UI for this yet. If your team's convention puts identifiers in a place Multica isn't reading, add them to the PR title or body.
+
+## Disconnecting
+
+In **Settings → Integrations** there is no installation list — you manage existing installations from GitHub directly:
+
+- **From GitHub** — uninstall the Multica GitHub App at `https://github.com/settings/installations` (personal) or `https://github.com/organizations/<org>/settings/installations` (org). Multica receives the `installation.deleted` webhook and drops the row in real time; any open Settings tab updates without a refresh.
+- **Disconnect from inside Multica is admin-only** — the Settings card is hidden for non-admins.
+
+After disconnect, mirrored PR rows stay in the database so historical issue sidebars still show what was linked, but no new webhook events from that installation will be accepted.
+
+## Permissions and visibility
+
+- **Connect / disconnect** require workspace **owner or admin**. Members see the card description but no Connect button.
+- The **Pull requests** sidebar on an issue is visible to anyone who can read the issue — same permissions as the rest of issue detail.
+- The GitHub App requests **read-only** access to pull requests and metadata. Multica never pushes commits, comments, or status checks back to GitHub.
+
+## Self-host setup
+
+If you're running Multica on Multica Cloud, the integration is already configured — skip this section.
+
+For self-host, you create one GitHub App, point it at your server, and set two environment variables. The whole flow is below.
+
+### 1. Create a GitHub App
+
+Go to one of:
+
+- Personal account → `https://github.com/settings/apps/new`
+- Organization → `https://github.com/organizations/<org>/settings/apps/new`
+
+Fill in:
+
+| Field | Value |
+|---|---|
+| **GitHub App name** | Anything recognizable, e.g. `Multica` or `Multica (staging)`. |
+| **Homepage URL** | Your Multica frontend, e.g. `https://multica.example.com`. |
+| **Callback URL** | Leave blank — Multica doesn't use OAuth user identity. |
+| **Setup URL** | `https://<api-host>/api/github/setup`. **Check "Redirect on update"**. |
+| **Webhook → Active** | Enabled. |
+| **Webhook URL** | `https://<api-host>/api/webhooks/github`. |
+| **Webhook secret** | Generate a long random string (e.g. `openssl rand -hex 32`). You'll paste the same value into Multica's env in step 2. |
+| **Permissions → Repository → Pull requests** | **Read-only**. |
+| **Permissions → Repository → Metadata** | Read-only (mandatory). |
+| **Subscribe to events** | Tick **Pull request**. |
+| **Where can this GitHub App be installed?** | Your choice. `Only on this account` is fine for single-org setups. |
+
+After **Create GitHub App**, note two things from the App's detail page:
+
+- The **public link** at the top — its tail is the slug. `https://github.com/apps/multica-acme` → slug = `multica-acme`.
+- The **webhook secret** you just generated (you can't read it back from GitHub later — save it now).
+
+<Callout type="warning">
+**Webhook secret ≠ Client secret.** The App settings page has both fields stacked together. The **Webhook secret** is what signs `pull_request` payloads — that's the one Multica needs. The **Client secret** is for OAuth and is not used by this integration. Mixing them up produces a confusing `401 invalid signature` on every webhook delivery.
+</Callout>
+
+### 2. Set environment variables
+
+On the API server:
+
+```dotenv
+GITHUB_APP_SLUG=multica-acme
+GITHUB_WEBHOOK_SECRET=<the webhook secret you generated>
+```
+
+Both variables are required. If either is missing:
+
+- `Connect GitHub` in Settings is **disabled** and shows a "not configured" hint.
+- The `/api/webhooks/github` endpoint returns **`503 github webhooks not configured`** — Multica refuses to process events with no secret, rather than silently treating every signature as valid.
+
+`FRONTEND_ORIGIN` must also be set (it already is for any production self-host); the setup callback bounces the user back to `<FRONTEND_ORIGIN>/settings` after install.
+
+Restart the API after setting the env vars.
+
+### 3. Run migrations
+
+The integration ships its tables in migration `079_github_integration`. If you're upgrading an older deployment:
+
+```bash
+make migrate-up
+```
+
+Three tables get created: `github_installation`, `github_pull_request`, `issue_pull_request`. They cascade-delete with their workspace, so removing a workspace cleans them up automatically.
+
+### 4. Connect from the UI
+
+In Multica:
+
+1. Open **Settings → Integrations** as an owner or admin.
+2. Click **Connect GitHub**. GitHub opens in a new tab.
+3. Pick the repositories to grant access to and **Install**.
+4. GitHub redirects back to `<api-host>/api/github/setup`, which records the installation and bounces you to `<FRONTEND_ORIGIN>/settings?github_connected=1`.
+
+After that, open any PR whose branch / title / body contains an issue identifier — within a few seconds the Pull requests block appears on that issue's detail page.
+
+### 5. Verify with a curl probe
+
+If GitHub's **Recent Deliveries** page reports `401 invalid signature` after install, the two sides have different secrets. The fastest way to find out which side is wrong is to bypass GitHub:
+
+```bash
+SECRET="<the value you put in GITHUB_WEBHOOK_SECRET>"
+BODY='{"zen":"test"}'
+SIG=$(printf '%s' "$BODY" | openssl dgst -sha256 -hmac "$SECRET" -hex | awk '{print $NF}')
+curl -i -X POST https://<api-host>/api/webhooks/github \
+  -H "X-Hub-Signature-256: sha256=$SIG" \
+  -H "X-GitHub-Event: ping" \
+  -H "Content-Type: application/json" \
+  -d "$BODY"
+```
+
+| HTTP status | Meaning | Fix |
+|---|---|---|
+| `200` `{"ok":"pong"}` | Server's loaded secret matches your `$SECRET`. The mismatch is on GitHub. | Edit the App → Webhook secret → **paste the same value** → **Save changes** (clicking out of the field without Save keeps the old secret). Redeliver. |
+| `401 invalid signature` | Server's loaded secret is **not** what you think it is. | Confirm the env var landed in the running process (e.g. `kubectl exec` → `echo -n "$GITHUB_WEBHOOK_SECRET" | wc -c`). Re-deploy. |
+| `503 github webhooks not configured` | `GITHUB_WEBHOOK_SECRET` is empty in the process. | Set the env var, restart the API. |
+
+## Limitations
+
+A few rough edges to be aware of today:
+
+- **No manual link UI yet** — the only way to link a PR is to have the identifier in its branch, title, or body.
+- **No CI / check state** — only the PR itself is mirrored. Build status, review comments, and reviewers are not surfaced in Multica.
+- **No workspace-level config** for the merge → Done rule — it's a fixed default (`merged → done`, unless `cancelled`). Workspace-customizable mappings are a future addition.
+- **Multi-PR-to-one-issue is conservative on merge** — if two PRs both reference `MUL-123` and the first one merges, the issue is moved to `Done` immediately. A follow-up change to wait for all linked PRs to resolve before advancing is in progress.
+
+## Next
+
+- [Issues](/issues) — the issue identifiers (`MUL-123`) referenced from PRs
+- [Workspaces](/workspaces) — where the workspace-specific issue prefix is set
+- [Environment variables](/environment-variables) — full env reference, including the GitHub variables above

apps/docs/content/docs/github-integration.zh.mdx

@@ -0,0 +1,183 @@
+---
+title: GitHub 集成
+description: 一次性连接 GitHub App，之后 PR 的分支名、标题或正文里写了 issue 编号（例如 MUL-123），就会自动挂到那个 issue 上——PR 合并时 issue 自动转 Done。
+---
+
+import { Callout } from "fumadocs-ui/components/callout";
+
+在 **Settings → Integrations** 里一次性连一个 GitHub 账号或组织。之后任何 PR 只要分支名、标题或正文里出现 issue 编号（例如 `MUL-123`），就会**自动关联**到那个 [issue](/issues)，出现在 issue 详情页右侧的 **Pull requests** 区块里——PR 合并时，issue 自动转 **Done**。
+
+没有 per-issue 的配置，整个流程是「编号驱动」的。
+
+## 集成做了什么
+
+| 出现位置 | 行为 |
+|---|---|
+| **Settings → Integrations** | 工作区 owner / admin 看到一个 GitHub 卡片，里面有 **Connect GitHub** 按钮。点击会打开 GitHub 的 App 安装页；装好后跳回 Settings。 |
+| **Issue 详情侧栏 → Pull requests** | 列出所有自动关联到该 issue 的 PR，含标题、仓库、状态（`Open` / `Draft` / `Merged` / `Closed`）和作者。点一行跳到 GitHub。 |
+| **Webhook（后台）** | 每次 `pull_request` 事件触发：upsert PR 行 → 扫描里面的 issue 编号 →（重新）建立 link。幂等——重投 delivery 不会产生重复记录。 |
+| **Merge 自动改 status** | PR 转 `merged` 时，所有已关联且状态不是 `Done` / `Cancelled` 的 issue 会被推到 `Done`。时间线里以 source 为 `github_pr_merged` 记录。 |
+
+只镜像 PR 本身。Commit、没开 PR 的分支、CI 检查状态都**不**入库——集成有意保持窄边界。
+
+## 编号是怎么匹配的
+
+Webhook 从三个字段抽取编号，顺序是：**PR head 分支** → **PR 标题** → **PR 正文**。匹配规则：
+
+- 大小写不敏感——`mul-123`、`MUL-123`、`Mul-123` 都能匹配
+- 有边界——左侧 `\b`、右侧只接数字，避免误抓 `v1.2-3`、email 地址等
+- 限定到本工作区——只匹配本工作区的 [issue prefix](/workspaces)。前缀是 `MUL` 的工作区里，PR 出现 `FOO-1` 不会匹配，即使数字撞另一个 issue 也不会
+- 自动去重——`Closes MUL-1, MUL-1` 只关联一次
+
+一个 PR 里**可以同时引用多个 issue**。比如 `Closes MUL-1, MUL-2`：PR 同时关联两个 issue，合并时两个 issue 都会转 `Done`。
+
+## Merge 自动转 Done 的规则
+
+PR 的 `merged` 字段翻成 `true` 时，逐个评估关联的 issue：
+
+| Issue 当前状态 | 结果 |
+|---|---|
+| `done` | 不变（已经是终态）|
+| `cancelled` | **不变**——cancelled 是用户明确放弃工作的信号，集成不覆盖 |
+| 其他（`todo` / `in_progress` / `in_review` / `blocked` / `backlog`）| 转成 `done` |
+
+PR **关闭但没合并**——只更新 PR 卡片的状态为 `Closed`，issue 状态不变。"关闭但不合并"语义因团队而异，Multica 不替用户做决定。
+
+<Callout type="info">
+状态变更的 actor 是 `system`。订阅了该 issue 的成员会收到 inbox 通知，和成员手动改状态时一致。
+</Callout>
+
+## 哪些情况不会自动关联
+
+- **Commit message 里的编号**——只扫 PR 的分支 / 标题 / 正文。一个 commit message 写 `MUL-123: fix login` 不会触发关联，除非同样的字符串也出现在 PR 标题或正文里
+- **PR 评论里的编号**——只扫 PR 自己的元数据，后续的 GitHub comment 不读
+- **App 没安装的仓库里的 PR**——没 App，Multica 收不到 webhook
+- **手动把 PR 关联到 issue**——暂时没有这个 UI。如果你们的约定把编号放到 Multica 不扫的地方，请改放到 PR 标题或正文里
+
+## 断开连接
+
+**Settings → Integrations** 里没有 installation 列表——现有 installation 直接到 GitHub 上管理：
+
+- **从 GitHub 卸载** —— 个人在 `https://github.com/settings/installations`、组织在 `https://github.com/organizations/<org>/settings/installations` 卸载 Multica App。Multica 收到 `installation.deleted` webhook 后立刻删行；任何已打开的 Settings tab 实时更新，不用刷新
+- **Multica 这边的断开是 admin only** —— 卡片对非 admin 不显示连接操作
+
+断开之后，已经镜像的 PR 行保留在数据库里——历史 issue 侧栏仍能显示当时关联的 PR，但来自这个 installation 的新 webhook 事件不再被接受。
+
+## 权限和可见性
+
+- **Connect / Disconnect** 需要工作区 **owner 或 admin**。普通成员能看到卡片描述但看不到 Connect 按钮
+- **Pull requests** 侧栏对所有能看到该 issue 的成员可见——和 issue 详情页其他部分权限一致
+- GitHub App 申请的是 PR 和 Metadata 的 **只读** 权限。Multica 从不向 GitHub 推 commit、评论或 status check
+
+## Self-Host 配置
+
+如果你在 Multica Cloud 上，集成已经配好——跳过本节。
+
+Self-Host 需要：建一个 GitHub App、指向你的 server、设两个环境变量。完整流程如下。
+
+### 1. 创建一个 GitHub App
+
+到下面其中一个页面：
+
+- 个人账号 → `https://github.com/settings/apps/new`
+- 组织 → `https://github.com/organizations/<org>/settings/apps/new`
+
+按下表填写：
+
+| 字段 | 值 |
+|---|---|
+| **GitHub App name** | 任何能辨识的名字，例如 `Multica` 或 `Multica (staging)` |
+| **Homepage URL** | 你的 Multica 前端，例如 `https://multica.example.com` |
+| **Callback URL** | 留空——本集成不使用 OAuth 用户身份 |
+| **Setup URL** | `https://<api-host>/api/github/setup`。**勾选 "Redirect on update"** |
+| **Webhook → Active** | 启用 |
+| **Webhook URL** | `https://<api-host>/api/webhooks/github` |
+| **Webhook secret** | 生成一个长随机字符串（例如 `openssl rand -hex 32`）。这个值会同样填到 step 2 的 env 里 |
+| **Permissions → Repository → Pull requests** | **Read-only** |
+| **Permissions → Repository → Metadata** | Read-only（必填）|
+| **Subscribe to events** | 勾选 **Pull request** |
+| **Where can this GitHub App be installed?** | 自选。单组织部署建议选 `Only on this account` |
+
+点 **Create GitHub App** 之后，从详情页记下两件事：
+
+- 顶部 **public link** 的尾部即 slug。`https://github.com/apps/multica-acme` → slug = `multica-acme`
+- 你刚生成的 **webhook secret**（GitHub 之后不会再让你读取这个值——现在就保存好）
+
+<Callout type="warning">
+**Webhook secret ≠ Client secret。** App 设置页里两个字段紧挨着。**Webhook secret** 用于签 `pull_request` payload，这才是 Multica 需要的那个；**Client secret** 是 OAuth 用的，和本集成无关。混淆这两个会得到「每条 webhook 都 `401 invalid signature`」的诡异症状。
+</Callout>
+
+### 2. 配置环境变量
+
+API server 上：
+
+```dotenv
+GITHUB_APP_SLUG=multica-acme
+GITHUB_WEBHOOK_SECRET=<你刚生成的 webhook secret>
+```
+
+两个都必填。任何一个缺失：
+
+- Settings 里 `Connect GitHub` 按钮会被 **disable**，并显示「not configured」提示
+- `/api/webhooks/github` 直接返回 **`503 github webhooks not configured`**——Multica 在 secret 没配置时拒绝处理事件，不会出现「没 secret 也接受 webhook」的安全坑
+
+`FRONTEND_ORIGIN` 也必须设置（任何生产 self-host 都已经设了）——setup 回调结束后用它把用户跳回 `<FRONTEND_ORIGIN>/settings`。
+
+设完 env 重启 API。
+
+### 3. 执行 migration
+
+集成的表在 migration `079_github_integration` 里。如果是升级既有部署：
+
+```bash
+make migrate-up
+```
+
+会创建三张表：`github_installation`、`github_pull_request`、`issue_pull_request`。三张表都 cascade 跟随 workspace——删工作区会自动清理。
+
+### 4. 在 UI 里连接
+
+到 Multica：
+
+1. 以 owner 或 admin 身份打开 **Settings → Integrations**
+2. 点 **Connect GitHub**，GitHub 在新 tab 打开
+3. 选择要授权的仓库，点 **Install**
+4. GitHub 跳回 `<api-host>/api/github/setup`，落库后再跳到 `<FRONTEND_ORIGIN>/settings?github_connected=1`
+
+之后在任意一个仓库开一个分支 / 标题 / 正文带本工作区 issue 编号的 PR——几秒内对应 issue 的详情页上就能看到 Pull requests 区块。
+
+### 5. 用 curl 自检
+
+如果 GitHub 的 **Recent Deliveries** 里第一次 PR 事件就报 `401 invalid signature`，说明两边的 secret 不一致。绕过 GitHub 直接测 server 是最快的定位方法：
+
+```bash
+SECRET="<你填给 GITHUB_WEBHOOK_SECRET 的值>"
+BODY='{"zen":"test"}'
+SIG=$(printf '%s' "$BODY" | openssl dgst -sha256 -hmac "$SECRET" -hex | awk '{print $NF}')
+curl -i -X POST https://<api-host>/api/webhooks/github \
+  -H "X-Hub-Signature-256: sha256=$SIG" \
+  -H "X-GitHub-Event: ping" \
+  -H "Content-Type: application/json" \
+  -d "$BODY"
+```
+
+| HTTP 状态 | 含义 | 修法 |
+|---|---|---|
+| `200` `{"ok":"pong"}` | server 加载的 secret 和你 `$SECRET` 一致——GitHub 那边的 secret 才是错的 | 编辑 App → Webhook secret 字段**粘相同的值** → **必须点 Save changes**（不点 Save 等于没改）→ Redeliver |
+| `401 invalid signature` | server 加载的 secret **不是**你以为的那个 | 进容器确认 env 实际生效（例如 `kubectl exec` → `echo -n "$GITHUB_WEBHOOK_SECRET" \| wc -c`），重新部署 |
+| `503 github webhooks not configured` | `GITHUB_WEBHOOK_SECRET` 在进程里是空的 | 配上 env，重启 API |
+
+## 已知限制
+
+目前还没做的几个边界：
+
+- **手动 link UI 暂未提供**——关联 PR 的唯一方法是把 issue 编号写到 PR 分支 / 标题 / 正文
+- **不读 CI / check 状态**——只镜像 PR 本身，构建状态、reviewer 评论、reviewer 列表都没接进 Multica
+- **没有工作区级别的 merge → status 映射配置**——默认固定是 `merged → done`（cancelled 除外）。可配置映射是后续迭代
+- **同 issue 多 PR 时，merge 行为偏激进**——两个 PR 都引用 `MUL-123` 时，第一个 merge 就把 issue 转 Done。"等所有关联 PR 都解决再推进 issue 状态"的优化已经在做了
+
+## 下一步
+
+- [Issues](/issues) —— PR 引用的 issue 编号（`MUL-123`）的来源
+- [工作区](/workspaces) —— 工作区 issue prefix 的设置位置
+- [环境变量](/environment-variables) —— 完整 env 清单，包含上面提到的 GitHub 变量

apps/docs/content/docs/meta.json

@@ -27,6 +27,8 @@
     "autopilots",
     "---Inbox---",
     "inbox",
+    "---Integrations---",
+    "github-integration",
     "---Self-hosting & ops---",
     "environment-variables",
     "auth-setup",

apps/docs/content/docs/meta.zh.json

@@ -26,6 +26,8 @@
     "autopilots",
     "---收件箱---",
     "inbox",
+    "---集成---",
+    "github-integration",
     "---自部署运维---",
     "environment-variables",
     "auth-setup",

apps/docs/content/docs/self-host-quickstart.mdx

@@ -115,5 +115,6 @@ Same flow as Cloud — see [Cloud quickstart → Steps 5-6](/cloud-quickstart#5-
 
 - [Environment variables](/environment-variables) — full env reference
 - [Auth setup](/auth-setup) — Resend / OAuth / signup allowlist in detail
+- [GitHub integration](/github-integration) — connect a GitHub App so PRs auto-link to issues and merging closes them
 - [Troubleshooting](/troubleshooting) — start here when things go wrong
 - [Desktop app](/desktop-app) — optional Desktop setup via `~/.multica/desktop.json`; the web frontend + CLI remains the quickest self-host path

apps/docs/content/docs/self-host-quickstart.zh.mdx

@@ -114,5 +114,6 @@ multica setup self-host
 
 - [环境变量](/environment-variables) —— 完整 env 清单
 - [登录与注册配置](/auth-setup) —— Resend / OAuth / 注册白名单详细配置
+- [GitHub 集成](/github-integration) —— 连一个 GitHub App，让 PR 自动关联 issue、merge 时自动转 Done
 - [故障排查](/troubleshooting) —— 遇到问题先来这里
 - [桌面应用](/desktop-app) —— 可以通过 `~/.multica/desktop.json` 连接 Desktop；Web 前端 + CLI 仍然是最快的自部署路径

apps/web/app/[workspaceSlug]/(dashboard)/squads/[id]/page.tsx

@@ -0,0 +1 @@
+export { SquadDetailPage as default } from "@multica/views/squads";

apps/web/app/[workspaceSlug]/(dashboard)/squads/page.tsx

@@ -0,0 +1 @@
+export { SquadsPage as default } from "@multica/views/squads";

apps/web/app/[workspaceSlug]/(dashboard)/usage/page.tsx

@@ -0,0 +1 @@
+export { DashboardPage as default } from "@multica/views/dashboard";

apps/web/features/landing/i18n/en.ts

@@ -284,6 +284,56 @@ export function createEnDict(allowSignup: boolean): LandingDict {
       fixes: "Bug Fixes",
     },
     entries: [
+      {
+        version: "0.2.32",
+        date: "2026-05-13",
+        title: "Usage Insights, Chat Renaming & Smoother Desktop Flows",
+        changes: [],
+        features: [
+          "Usage now shows workspace and project token activity, runtime trends, and per-agent rankings in one place",
+          "Chat sessions can be renamed directly from the chat header",
+          "Feedback reports can include screenshots or files so teams have the context they need",
+        ],
+        improvements: [
+          "The Usage page has clearer naming and a more dynamic agent leaderboard",
+          "New chats and completed chat responses update more smoothly with fewer loading flashes",
+          "Self-hosted GitHub setup is easier to configure and the setup docs point to the right cloud URL",
+          "User-installed Codex skills are available automatically when new tasks run",
+        ],
+        fixes: [
+          "Empty successful agent responses are marked completed instead of blocked",
+          "Pasted mentions in instruction editors keep their mention links",
+          "Desktop attachment downloads use the native Linux flow and tab closing no longer loops",
+          "Gemini and Windows runtime startup checks are more reliable in unattended runs",
+          "Long GitHub repository lists stay usable when adding project resources",
+        ],
+      },
+      {
+        version: "0.2.31",
+        date: "2026-05-12",
+        title: "GitHub Integration, Chat Attachments & Safer Issue Navigation",
+        changes: [],
+        features: [
+          "Connect GitHub so linked pull requests appear on Multica issues, sync their status, and close the Multica issue automatically when the PR closes",
+          "Chat messages can include file attachments and image previews",
+          "Agents and runtimes can now be kept public or private for clearer team access",
+          "Stopping a single agent task now asks for confirmation before it is terminated",
+          "New GitHub integration docs cover both hosted and self-hosted setup",
+        ],
+        improvements: [
+          "Issue links land more reliably on the exact comment or activity you opened",
+          "Long issue timelines scroll more smoothly",
+          "The feedback dialog now points contributors toward GitHub discussions and issues",
+          "Self-hosted Caddy guidance now calls out real-time connection requirements",
+          "Linux desktop packages show the Multica app icon again",
+        ],
+        fixes: [
+          "Downloaded attachments keep their original filenames",
+          "Local attachments are served more reliably, and upload controls stay disabled until files are ready",
+          "Issue creation dialogs keep their text fields at the correct height",
+          "Runtime documentation links point to the correct page",
+        ],
+      },
       {
         version: "0.2.30",
         date: "2026-05-11",

apps/web/features/landing/i18n/zh.ts

@@ -284,6 +284,56 @@ export function createZhDict(allowSignup: boolean): LandingDict {
       fixes: "问题修复",
     },
     entries: [
+      {
+        version: "0.2.32",
+        date: "2026-05-13",
+        title: "用量洞察、聊天重命名与桌面体验优化",
+        changes: [],
+        features: [
+          "Usage 页面集中展示 workspace 和 project 的 token 使用、runtime 趋势和 agent 排名",
+          "聊天会话可以直接在聊天页顶部重命名",
+          "反馈时可以附带截图或文件，方便团队快速理解问题",
+        ],
+        improvements: [
+          "Dashboard 更名为 Usage，并加入更清晰的 agent 排行展示",
+          "新聊天和消息完成状态切换更顺，不再频繁闪加载状态",
+          "自托管 GitHub 配置更完整，文档里的云端链接也已修正",
+          "用户安装的 Codex Skills 会自动带入新的 agent 任务",
+        ],
+        fixes: [
+          "没有输出内容但成功完成的 agent 任务会显示为 completed，不再误判为 blocked",
+          "在指令编辑器中粘贴的 mention 会保留可点击链接",
+          "Linux 桌面端下载附件时走系统原生流程，关闭标签页也不再触发循环跳转",
+          "Gemini 和 Windows runtime 的启动检查更稳定，适合无人值守执行",
+          "添加项目资源时，较长的 GitHub 仓库列表可以正常滚动",
+        ],
+      },
+      {
+        version: "0.2.31",
+        date: "2026-05-12",
+        title: "GitHub 集成、聊天附件与 Issue 定位优化",
+        changes: [],
+        features: [
+          "接入 GitHub 后，关联的 Pull Request 会显示在 Multica Issue 中，状态会同步到 Multica，关闭 PR 后会自动关闭对应 Issue",
+          "聊天消息支持添加文件附件和图片预览",
+          "Agent 和 runtime 可以设置公开或私有，方便控制团队可见范围",
+          "停止单个 agent 任务前会先弹出确认，避免误操作",
+          "新增 GitHub 集成文档，覆盖托管版和自托管配置",
+        ],
+        improvements: [
+          "打开 Issue 链接时，会更稳定地定位到指定评论或动态",
+          "很长的 Issue 时间线滚动更顺畅",
+          "反馈入口更明确地引导用户到 GitHub 参与讨论和反馈",
+          "自托管 Caddy 配置文档补充实时连接要求",
+          "Linux 桌面端安装包恢复显示 Multica 应用图标",
+        ],
+        fixes: [
+          "下载附件时保留原始文件名",
+          "本地附件访问更稳定，上传按钮会等文件准备好后再可用",
+          "创建 Issue 弹窗里的文本框高度显示正确",
+          "Runtime 文档入口跳转到正确页面",
+        ],
+      },
       {
         version: "0.2.30",
         date: "2026-05-11",

docker-compose.selfhost.yml

@@ -58,9 +58,11 @@ services:
       APP_ENV: ${APP_ENV:-production}
       MULTICA_DEV_VERIFICATION_CODE: ${MULTICA_DEV_VERIFICATION_CODE:-}
       MULTICA_APP_URL: ${MULTICA_APP_URL:-http://localhost:3000}
-      ALLOW_SIGNUP: ${ALLOW_SIGNUP:-true}                                     
-      ALLOWED_EMAILS: ${ALLOWED_EMAILS:-}                                     
-      ALLOWED_EMAIL_DOMAINS: ${ALLOWED_EMAIL_DOMAINS:-} 
+      ALLOW_SIGNUP: ${ALLOW_SIGNUP:-true}
+      ALLOWED_EMAILS: ${ALLOWED_EMAILS:-}
+      ALLOWED_EMAIL_DOMAINS: ${ALLOWED_EMAIL_DOMAINS:-}
+      GITHUB_APP_SLUG: ${GITHUB_APP_SLUG:-}
+      GITHUB_WEBHOOK_SECRET: ${GITHUB_WEBHOOK_SECRET:-}
     restart: unless-stopped
 
   frontend:

docs/agent-quick-create-plan.md

@@ -0,0 +1,555 @@
+# Agent 快速创建 — 三阶段实施计划
+
+> Status: Draft (设计阶段,未动工)
+> Owner: TBD
+> Last updated: 2026-05-13
+
+## TL;DR
+
+- **目标**:降低用户创建 Agent 的门槛,从「手工填表 + 一个个挑 skill」演进到「一键模板」「AI 推荐 skill」「AI 直接创建 agent」三档
+- **三阶段**:Template(必做、独立)→ Skill Finder(AI 推荐 skill)→ AI Create Agent(AI 直接创建)
+- **架构关键**:Phase 2/3 复用现有 Quick-create Issue 基础设施(派任务给 agent + tool calling + inbox 通知),不引入新 LLM 调用路径
+- **不需要新基础设施**:无 SSE、无 server-side LLM、无新 WS channel
+- **soft blocker**:两处 routine 重构(`createSkillWithFiles` TX 拆分、skill 同名 find-or-create)
+- **不做**:接入 Anthropic 官方 marketplace(plugin 体系跟单体 skill 形态不匹配)、接入 ClawHub(战略对位错误 + 实际使用率低,见 §5)
+
+---
+
+## 1. 背景与目标
+
+### 1.1 当前现状
+
+当前用户创建一个 Agent 需要走的步骤:
+
+1. 进 `/agents` 页面 → 点 "Create Agent"
+2. 手工填 name / description / runtime / model
+3. 手工写 instructions(空白文本框,用户自己思考措辞)
+4. 创建完后进 Agent 详情页 → 点 "Add Skill" → 一个一个挑 skill 关联
+5. 如果 workspace 还没有需要的 skill,得先去别处建/导入 skill(`POST /api/skills/import` 支持 skills.sh / GitHub / ClawHub 三种 URL)
+
+**痛点**:
+- 用户得**预先知道**自己需要哪些 skill,这要求他对 skill 生态熟悉
+- 写 instructions 是空白文本编辑,大多数用户不知道写什么
+- 跨多页操作,体感上"创建一个能用的 Agent"是个项目,不是个动作
+
+### 1.2 三阶段方案
+
+| Phase | 提供给用户的能力 | 是否需要 AI | 独立可发布 |
+|---|---|---|---|
+| **1. Template** | 选模板 → 自动 import 模板带的 skill + 预填 instructions | 否 | ✅ |
+| **2. Skill Finder** | 描述需求 → AI 推荐 skill 列表 → 一键导入到 workspace | ✅ | ✅(独立功能,任何场景都能用) |
+| **3. AI Create Agent** | 描述需求 → AI 自己 find skill + 写 instructions + 创建 agent | ✅ | 依赖 Phase 2 |
+
+每个 phase **本身有用户价值**,不需要等下一个 phase 才能用:
+- Phase 1 用户能用模板创建 agent,即使后两阶段没做
+- Phase 2 用户能在任何地方"用 AI 找 skill"(创建 agent 时、给现有 agent 加 skill 时、单纯逛 skill 时)
+- Phase 3 是 1+2 的组合
+
+### 1.3 不在范围内
+
+明确不做的事(及理由,见 §5):
+- 接入 Anthropic 官方 plugin marketplace(`anthropics/claude-plugins-official`)
+- 接入 ClawHub 的"发现/搜索"层(import 路径已经存在,但是死代码,建议下线)
+- 让 AI 直接装 skill 到用户本地 `~/.claude/skills/`(npx skills CLI 行为)
+- Server-side LLM 调用(后端目前没有 LLM SDK,这条路引入新基础设施,而 Quick-create 模式可以避开)
+
+---
+
+## 2. 关键概念回顾
+
+> 这一节给没参与前期讨论的同事看。已经熟悉 skill 系统的可跳到 §3。
+
+### 2.1 Skill 是什么
+
+Skill 是一个**按需加载的能力包**,本质是 SKILL.md 文件 + 可选附件。Anthropic 2025-12 把它发布为开放标准(agentskills.io),Cursor / OpenAI / GitHub Copilot 等都已采纳——同一份 SKILL.md 跨多个 agent 工具都能用。
+
+每个 runtime(Claude Code / Cursor / Codex 等)启动时**自动扫**自己约定的目录(`~/.claude/skills/`、`.cursor/skills/` 等),读 SKILL.md 的 frontmatter 形成"我手上有这些 skill"的清单注入 system prompt。具体 skill 正文只在被触发时才进 context。
+
+### 2.2 Multica 的 Skill 数据模型
+
+3 张表(migration `008_structured_skills.up.sql`):
+
+| 表 | 关键字段 |
+|---|---|
+| `skill` | `id, workspace_id, name, description, content (=SKILL.md 正文), config (含 origin 元数据)` |
+| `skill_file` | `skill_id, path, content`(SKILL.md 的附件,如 examples/*.md、scripts/*.py) |
+| `agent_skill` | `agent_id, skill_id`(M:N 关联) |
+
+**关键约束**:`UNIQUE(workspace_id, name)` — 同 workspace 内 skill 名字必须唯一。
+
+### 2.3 Skill 流转链路(数据库 → runtime)
+
+任务运行时,skill 从 PG 到 runtime 的完整路径:
+
+```
+1. 数据库:skill + skill_file + agent_skill 三张表的行
+
+2. Daemon claim 任务:
+   POST /api/runtimes/{runtimeId}/tasks/claim
+   handler/daemon.go:1018-1098 (ClaimTaskByRuntime)
+   → service/task.go:1447-1463 (LoadAgentSkills)
+   → 把 agent 关联的所有 skill 全文塞进 HTTP 响应
+
+3. Daemon 算工作目录:
+   server/internal/daemon/execenv/execenv.go:114, 124
+   workDir = {WorkspacesRoot}/{wsID}/{shortTaskID}/workdir
+
+4. Daemon 按 runtime 算 skill 目录:
+   server/internal/daemon/execenv/context.go:121-158 (resolveSkillsDir)
+   claude   → {workDir}/.claude/skills
+   cursor   → {workDir}/.cursor/skills
+   codex    → 特殊:{codexHome}/skills
+
+5. Daemon 把字符串写成磁盘文件:
+   context.go:175-204 (writeSkillFiles)
+   核心就两行 os.WriteFile
+
+6. Daemon 启动 runtime,cwd = workDir
+   runtime 自己扫 .claude/skills/(等)→ 加载 frontmatter
+
+7. 任务结束:os.RemoveAll(workDir)
+   PG 是真相源,workDir 是每次任务临时复印件
+```
+
+**核心 invariant**:Multica 不教 runtime 怎么用 skill,只把文件摆到 runtime 已经会扫的位置。
+
+### 2.4 Template = Instructions + Skill 引用
+
+Template 是个**静态 JSON 定义**,包含:
+- 预写好的 instructions
+- 一组 skill 引用(用 URL 指向 skills.sh / GitHub)
+
+用户选模板时,后端:
+1. 对每个 skill 引用,**复用现有 `/api/skills/import` 的 fetcher**(`fetchFromSkillsSh` / `fetchFromGitHub`)拉内容
+2. 物化到 workspace(同名复用 / 新建)
+3. CreateAgent + setAgentSkills
+4. 整个流程一个事务
+
+skill 引用为什么用 URL 而不是内联 SKILL.md 内容:
+- 复用现有 import 基础设施,零新代码
+- skill 内容跟 GitHub 同步,不需要 vendoring 进 multica 仓库
+- 模板 JSON 体积小,git review 友好
+
+### 2.5 Quick-create Issue 模式(Phase 2/3 复用的基础设施)
+
+当前 `POST /api/issues/quick-create`(handler/issue.go:877-982)的流程:
+
+```
+1. 后端 enqueue 任务:
+   - agent_task_queue 加一行,issue_id = NULL,context JSONB = {type: "quick-create", prompt: ...}
+   - 立即返回 202 Accepted + task_id
+
+2. Daemon claim 任务时识别 quick-create:
+   - 检查 task.Context != nil AND !task.IssueID.Valid
+   - 解析为 QuickCreateContext (service/task.go:1810-1811)
+
+3. Daemon 构造 prompt:
+   - daemon/prompt.go:45-106 (buildQuickCreatePrompt)
+   - 把用户的自然语言 prompt 作为语义核心
+   - 加上"调用 multica issue create CLI 命令"的指令
+
+4. Agent 跑 LLM + tool calling:
+   - LLM 输出形如 `multica issue create --title="..." --description="..."` 的命令
+   - daemon 执行 CLI 命令,CLI 调 POST /api/issues 创建 issue
+   - CLI 自动在请求里带上 MULTICA_QUICK_CREATE_TASK_ID env(daemon/daemon.go:2081)
+     → 让创建出来的 issue 带 origin_type='quick_create' + origin_id=<task_id>
+
+5. 后端 link + 通知:
+   - 完成检测:GetIssueByOrigin(workspace_id, "quick_create", task_id)
+   - LinkTaskToIssue(task_id, issue_id) 把任务行的 issue_id 补上
+   - 写 inbox_item 通知用户(notifyQuickCreateCompleted, service/task.go:1908-1920)
+```
+
+**关键洞察**:这个模式**完全通用化**了。复用它只需要:
+1. 新的 context JSONB type(比如 `"skill-find"`、`"agent-create"`)
+2. 新的 prompt builder
+3. 新的"完成检测 + inbox 通知"
+
+不需要任何 daemon / 任务队列层面的改动。
+
+---
+
+## 3. 三阶段详细设计
+
+### Phase 1:Agent Template
+
+**目标**:用户选模板 → 一键得到一个可用的 agent(自带 skill + instructions),不需要 AI 参与。
+
+#### 设计
+
+- **Template 定义存放**:静态 JSON,commit 在 `server/internal/agenttmpl/templates/*.json`
+- **Template JSON 形态**:
+  ```json
+  {
+    "slug": "code-reviewer",
+    "name": "Code Reviewer",
+    "description": "审代码用的 agent",
+    "instructions": "你审代码,关注 N+1 查询、错误处理、类型安全...",
+    "skills": [
+      { "source_url": "https://skills.sh/obra/superpowers/tdd" },
+      { "source_url": "https://github.com/foo/bar/tree/main/skills/code-style" }
+    ]
+  }
+  ```
+- **新 endpoint**:`POST /api/agents/from-template`
+  - 请求:`{template_slug, name, runtime_id, ...overrides}`
+  - 后端流程(**全部在一个事务里**):
+    1. 加载 template JSON
+    2. 对每个 skill source_url:
+       - 调用 `detectImportSource(url)`(skill.go:586-617)分发到对应 fetcher
+       - 通过 GetSkillByWorkspaceAndName 检查 workspace 是否已有同名 skill
+         - 有 → 复用现有 skill_id
+         - 无 → 调 `createSkillWithFilesInTx`(待重构,见 §4)物化
+    3. `CreateAgent`(复用 agent.go:CreateAgent 的内部逻辑)
+    4. 批量 `AddAgentSkill` 关联
+  - 响应:`{agent: {...}, imported_skill_ids: [...], reused_skill_ids: [...]}`
+- **前端**:`CreateAgentDialog`(packages/views/agents/components/create-agent-dialog.tsx)加 "From template" 模式,跟现有 manual / duplicate 模式并列
+  - 模板选择器 → 预览(instructions + skill 列表)→ 提交调新 endpoint
+  - 响应里的 `reused_skill_ids` 用 toast 提示"以下 skill 已存在,沿用了 workspace 现有版本"
+
+#### 起步模板清单(初版,可调)
+
+- `code-reviewer` — 代码审查
+- `tdd-pair` — TDD 配对编程
+- `db-reviewer` — 数据库 / SQL 审查
+- `pr-summarizer` — PR 摘要
+- `docs-writer` — 文档撰写
+
+具体每个模板选哪些 skill URL,在 Phase 1 启动时单独决定(需要逛 skills.sh 选高质量 skill)。
+
+#### Phase 1 改动清单
+
+| 文件 / 位置 | 改动 |
+|---|---|
+| `server/internal/agenttmpl/`(新包) | 加载 JSON 模板的代码 |
+| `server/internal/agenttmpl/templates/*.json`(新文件) | 5 个起步模板 |
+| `server/internal/handler/agent.go` | 新 handler `CreateAgentFromTemplate` |
+| `server/internal/handler/skill_create.go` | **重构**:拆出 `createSkillWithFilesInTx` 变体(见 §4) |
+| `server/pkg/db/queries/skill.sql` | 加 `GetSkillByWorkspaceAndName`(见 §4) |
+| `server/cmd/server/router.go` | 注册新 endpoint |
+| `packages/views/agents/components/create-agent-dialog.tsx` | 加 template 模式 |
+| `packages/core/api/agent.ts` | 加 `createAgentFromTemplate` API 调用 |
+| `packages/views/agents/components/template-picker.tsx`(新文件) | 模板选择器组件 |
+
+### Phase 2:Skill Finder
+
+**目标**:用户用自然语言描述需求(如"我想审 SQL"),AI 推荐一组 skill,用户勾选一键导入到 workspace。
+
+#### 设计
+
+- **架构选型**:走 quick-create 模式,**不是后端直接调 LLM**
+- **新 endpoint**:`POST /api/skills/find`
+  - 请求:`{prompt, agent_id}`(agent_id 是用来跑这个 LLM 任务的 agent,跟 Quick-create Issue 一样要求预先有 agent)
+  - 后端流程:
+    1. enqueue 任务:`agent_task_queue` 加一行,context JSONB = `{type: "skill-find", prompt}`
+    2. 返回 202 + task_id
+- **Daemon prompt builder**:`daemon/prompt.go` 加 `buildSkillFindPrompt`(类比 buildQuickCreatePrompt)
+  - 喂给 agent 的 prompt 大致:
+    ```
+    用户需求:{user_prompt}
+    
+    你的任务:从以下 curated skill 清单里选 3-5 个最相关的推荐给用户。
+    
+    可选 skill 清单(JSON):
+    {curated_skill_index}
+    
+    输出:调用 `multica skill find --output-results '<JSON>'` 命令,
+    JSON 形态为 [{name, description, source_url, reason}, ...]
+    ```
+- **CLI 命令**(新):`multica skill find --output-results <JSON>`
+  - 不发起 HTTP 请求,只把 JSON 写到 daemon 通过 env 指定的临时文件
+  - daemon 读这个文件,把内容塞进 inbox notification 的 payload
+- **Curated skill 索引**:`server/internal/agenttmpl/skill_index.json`(新文件)
+  - 几十到上百条精选 skill,每条:`{name, description, source_url, tags, install_count}`
+  - 维护方式:工程师/产品手工维护,代码 review 卡内容质量
+  - MVP **不做**实时 GitHub Code Search 或 skills.sh 爬虫
+- **完成通知**:写 inbox_item,type = `skill_find_done`,payload 含推荐结果数组
+- **前端**:
+  - 独立"Find Skill"页面(`/skills/find` 或 `/skills?ai=true`)
+  - skill list page 上"用 AI 找 skill"按钮入口
+  - 用户输入 prompt → 提交 → 等通知 → inbox item 里展示 skill 卡片(name + description + source_url + reason)
+  - 用户勾选 → 一键批量调现有 `POST /api/skills/import`(每个 skill 一次,可考虑加 batch endpoint 但 MVP 不必要)
+
+#### Phase 2 改动清单
+
+| 文件 / 位置 | 改动 |
+|---|---|
+| `server/internal/handler/skill.go` | 新 handler `FindSkill`(enqueue task) |
+| `server/internal/service/task.go` | 加 `EnqueueSkillFindTask` + 完成检测 + inbox 通知 |
+| `server/internal/daemon/prompt.go` | 加 `buildSkillFindPrompt` |
+| `server/internal/daemon/daemon.go` | 加 `SkillFindContext` 识别 + env 注入 |
+| `server/cmd/multica/cmd_skill.go` | 加 `find --output-results` 子命令 |
+| `server/internal/agenttmpl/skill_index.json`(新文件) | curated 清单 |
+| `packages/views/skills/components/find-skills-dialog.tsx`(新文件) | UI |
+| `packages/core/api/skill.ts` | 加 `findSkills` API |
+| `packages/views/inbox/items/skill-find-result.tsx`(新文件) | inbox item 渲染 |
+
+### Phase 3:AI Create Agent
+
+**目标**:用户描述需求,AI 自己 find skill + 写 instructions + 创建 agent。
+
+#### 设计
+
+- **架构选型**:走 quick-create 模式,**组合 Phase 2 的 find 能力 + 新的 agent create CLI**
+- **新 endpoint**:`POST /api/agents/ai-draft`
+  - 请求:`{prompt, host_agent_id}`(host_agent_id 是跑这个元任务的 agent)
+  - 后端:enqueue 任务,context = `{type: "agent-create", prompt}`,返回 202 + task_id
+- **Daemon prompt builder**:`buildAgentCreatePrompt` 指挥 agent 三步走:
+  ```
+  1. 调用 `multica skill find --output-results ...` 选 skill
+     (或直接看 curated 清单选)
+  2. 基于选定 skill 写 instructions
+  3. 调用 `multica agent create --name ... --instructions ... --skill-ids ...`
+     创建 agent 并关联 skill
+  ```
+- **CLI 命令**(新):`multica agent create`
+  - 后端 handler 已存在(handler/agent.go:CreateAgent),只需要绑 CLI(~50 行)
+  - 创建时带 `MULTICA_AI_DRAFT_TASK_ID` env,服务端用它做 origin 标记 + LinkTaskToAgent
+- **完成通知**:inbox_item type = `agent_draft_done`,payload 含 agent_id + 摘要
+- **前端**:`CreateAgentDialog` 加 "AI" 模式
+  - 输入需求 → 提交 → 等通知 → inbox 通知里点击 → 跳新 agent 详情页(用户在那儿编辑/调整)
+
+#### Phase 3 改动清单
+
+| 文件 / 位置 | 改动 |
+|---|---|
+| `server/internal/handler/agent.go` | 新 handler `AIDraftAgent`(enqueue task) |
+| `server/internal/service/task.go` | 加 `EnqueueAgentDraftTask` + 完成检测 + inbox 通知 |
+| `server/internal/daemon/prompt.go` | 加 `buildAgentCreatePrompt` |
+| `server/cmd/multica/cmd_agent.go` | 加 `create` 子命令(handler 已有) |
+| `packages/views/agents/components/create-agent-dialog.tsx` | 加 "AI" 模式 |
+| `packages/core/api/agent.ts` | 加 `aiDraftAgent` API |
+| `packages/views/inbox/items/agent-draft-result.tsx`(新文件) | inbox item 渲染 |
+
+---
+
+## 4. Blocker 清单与修复方案
+
+### 4.1 [SOFT] `createSkillWithFiles` 不可组合事务
+
+**问题**:`server/internal/handler/skill_create.go:21-71` 这个函数自己 `Begin()` 一个事务,执行完 `Commit()`。Phase 1 需要在外层事务里**多次**调用它(import N 个 skill + createAgent + setAgentSkills 都在一个 TX),但现在没法这么用。
+
+**影响范围**:Phase 1
+
+**修复方案**:
+
+```go
+// 拆成两个函数(保持原 API 向后兼容):
+
+// 新增:接受外部 qtx,不管事务
+func createSkillWithFilesInTx(
+    ctx context.Context,
+    qtx *db.Queries,
+    input skillCreateInput,
+) (*SkillWithFilesResponse, error) {
+    // 不 Begin/Commit,只调 qtx.CreateSkill + qtx.UpsertSkillFile loop
+}
+
+// 改造:原函数变成包装层,内部调 InTx 版
+func (h *Handler) createSkillWithFiles(
+    ctx context.Context,
+    input skillCreateInput,
+) (*SkillWithFilesResponse, error) {
+    tx, _ := h.TxStarter.Begin(ctx)
+    defer tx.Rollback()
+    qtx := h.Queries.WithTx(tx)
+    result, err := createSkillWithFilesInTx(ctx, qtx, input)
+    if err != nil { return nil, err }
+    tx.Commit()
+    return result, nil
+}
+```
+
+旧调用方完全不变。Phase 1 新 endpoint 自己 Begin,然后多次调 `*InTx` 变体,最后统一 Commit。
+
+**工作量**:小(< 100 行重构)
+
+### 4.2 [SOFT] Skill 同名冲突
+
+**问题**:`skill` 表有 `UNIQUE(workspace_id, name)` 约束。Phase 1 模板导入时,如果模板里的 skill 跟 workspace 已有 skill 同名,INSERT 会报 PG 错误 23505,整个 from-template 流程挂掉。
+
+**影响范围**:Phase 1
+
+**修复方案**:加 find-or-create 模式:
+
+1. 新 query `GetSkillByWorkspaceAndName`(`server/pkg/db/queries/skill.sql`)
+2. Phase 1 流程改成:
+   - 对每个模板 skill,先查 workspace 是否已有同名
+   - 有 → 复用现有 skill_id,跳过 import
+   - 无 → 调 `createSkillWithFilesInTx` 物化
+3. 响应里返回 `reused_skill_ids: [...]`,前端 toast "以下 skill 已存在,沿用现有版本"
+
+**不选择"覆盖"或"加后缀"的原因**:用户可能已经改过本地版本,覆盖会丢用户修改;加后缀污染 skill 列表。
+
+**工作量**:小(< 50 行 + 1 条 sqlc query)
+
+### 4.3 [SOFT] 缺 `multica skill find` CLI
+
+**影响范围**:Phase 2
+
+**方案**:加一个 CLI 子命令,模仿 `multica skill import` 的实现(`server/cmd/multica/cmd_skill.go:55-60, 323-357`)。**注意**:这个命令不发 HTTP 请求,只是 LLM agent 用来"输出推荐结果"的 channel——它把 LLM 推荐的 JSON 写到 daemon 指定的临时文件,daemon 读完塞进 inbox notification。
+
+**工作量**:小(~80 行)
+
+### 4.4 [SOFT] 缺 `multica agent create` CLI
+
+**影响范围**:Phase 3
+
+**方案**:后端 handler 已有(`handler/agent.go:CreateAgent`),只需在 `server/cmd/multica/cmd_agent.go` 加 `create` 子命令。
+
+**工作量**:小(~50 行)
+
+### 4.5 [非 blocker] System Agent 问题
+
+**之前误判为 hard blocker,实际不是**:
+
+Quick-create Issue 当前的设计就要求用户**预先有一个 agent** 才能用——AI 路径不为"零 agent 起步"服务。Phase 2/3 沿用这个前提,所以**新 workspace 没 agent 时 AI 功能不可用**是符合现有产品模型的,不需要 bootstrap 一个 system agent。
+
+产品自然解锁路径:
+1. 新用户进 workspace
+2. 用 **Phase 1 Template**(无需 AI、无需现有 agent)创建第一个 agent
+3. 之后 Phase 2/3 即可用,host_agent 就用刚创建的那个
+
+---
+
+## 5. 关键设计决策(及理由)
+
+### 5.1 为什么不接 Anthropic 官方 marketplace?
+
+**结构错配**。Anthropic 官方 marketplace(`anthropics/claude-plugins-official`)是 **plugin 体系**:每个 plugin 是个 bundle,包含 `.claude-plugin/plugin.json` + `skills/` + `agents/` + `hooks/` + `.mcp.json`。
+
+Multica 只有**单体 skill**(SKILL.md + skill_file),没有 plugin / bundle 概念。要接入得新写 plugin parser + 拆分逻辑,工作量大,而 skills.sh 已经覆盖了同一批高质量内容(skills.sh 后端就是 GitHub raw,绝大多数 skill 作者就在 GitHub 上,Anthropic plugin 体系里的 skill 通常也在作者的 GitHub repo 里有单体副本)。
+
+### 5.2 为什么走 quick-create 模式而不是后端直接调 LLM?
+
+代码事实:`server/` 目前**完全没有任何 LLM SDK**(grep `anthropic-sdk-go` / `openai-go` / 任何 LLM provider 都是 0 命中)。所有 LLM 调用都通过 daemon → runtime → CLI 这条路。
+
+走 quick-create 模式的优势:
+- **不引入新基础设施**(SSE / LLM client / API key 管理)
+- **复用 agent 的 instructions / model / runtime 配置**(用户已经在某个 agent 里配置过的偏好自动生效)
+- **统一计费 / 用量监控**(LLM 调用都计在用户 agent 的 quota 里)
+
+代价:
+- 用户得**预先有一个 agent**(参见 §4.5,这跟 Quick-create Issue 现状一致)
+- LLM 调用通过 daemon 多一跳,延迟略增(但不阻塞 202 响应)
+
+### 5.3 为什么 Skill Finder 是 endpoint 不是 SKILL.md?
+
+**Skill Finder 名字里的 "Skill" 是它的产物(找的是 skill),不是它自己实现成 SKILL.md**。
+
+如果做成 SKILL.md 文件:
+- 它得装进某个 agent 里才能用 → 单点功能变得需要前置配置
+- skill 教 agent 调什么?调 `npx skills`(装到本地,目标错)?调 Multica API(那要写 tool channel,绕一大圈)
+- AI 创建 Agent(Phase 3)那条路要"启动 agent → agent 调 skill → skill 调 tool",链路复杂三倍
+
+做成 endpoint:
+- 用户独立可用(独立 UI 入口)
+- AI 创建 Agent 后端直接调 endpoint,两个功能共用一段逻辑
+- 简单
+
+### 5.4 Curated Skill 索引 vs 实时搜索
+
+**MVP 用 curated 清单**(几十条精选 URL + 摘要 commit 在 repo 里)。理由:
+- 质量可控
+- 不踩 GitHub Code Search rate limit
+- 不被 LLM 编 URL(LLM 知识 cutoff + hallucinate URL 是真问题)
+- 维护成本低
+
+进阶可加 `search_skills(query)` tool 实时打 GitHub Code Search,等用户反馈"清单太窄"再做。
+
+### 5.5 不做 ClawHub(顺手清理建议)
+
+**现状**:`POST /api/skills/import` 当前支持 3 个 source(`fetchFromClawHub` skill.go:642-744、`fetchFromSkillsSh` skill.go:757-879、`fetchFromGitHub` skill.go:1363-1463)。ClawHub 是个独立 HTTP 客户端,不复用 GitHub 基础设施。
+
+**判断**(详见之前讨论):
+- ClawHub 服务的是 OpenClaw 平台(Multica 同生态位竞品的内容生态)
+- UI 没有发现/搜索层,用户只能粘 URL,而 ClawHub 装机量远低于 skills.sh,用户主动逛的概率极低
+- 独立代码路径,API 演进时单独跟进
+
+**建议**(独立于本计划,可以一起做也可以延后):
+- 跑 `SELECT count(*) FROM skill WHERE config->'origin'->>'type' = 'clawhub'` 看实际使用量
+- 接近 0 → 渐进下线(先去 UI SourceCard,后续 release 删 fetcher)
+- 有量 → 留着,但仍不为它做新功能
+
+---
+
+## 6. 实施依赖与排期
+
+```
+[Phase 1] Template
+  └── 独立,无依赖
+  └── 包含 2 个 soft blocker 的修复(§4.1 §4.2)
+       ↓
+[Phase 2] Skill Finder
+  └── 依赖 Phase 1 中的 skill import 路径(已存在,沿用)
+  └── 含 1 个 soft blocker(§4.3)
+       ↓
+[Phase 3] AI Create Agent
+  └── 依赖 Phase 2(复用 find skill 能力)
+  └── 含 1 个 soft blocker(§4.4)
+```
+
+**真实排期建议**:
+- Phase 1 可单独发版,有独立价值
+- Phase 2 独立可发版(找 skill 是高频独立场景)
+- Phase 3 等 Phase 2 ready 后开始
+
+每个 phase 启动时单独开 PR 设计 doc,本文档只是路线图。
+
+---
+
+## 7. 风险与缓解
+
+| 风险 | 缓解 |
+|---|---|
+| GitHub rate limit(模板 import 多个 skill 时) | 已有 `GITHUB_TOKEN` env 支持(skill.go:1163-1166),5000/h 配额够用。生产环境确保配置 |
+| 模板里引用的 skill repo 被作者删除 | from-template handler 容错:某个 skill fetch 失败 → 整个事务回滚,前端展示具体哪个 URL 挂了。模板自己也定期 review |
+| LLM 推荐编造 URL(Phase 2) | 用 curated 清单作为 context,**不让 LLM 自由发挥 URL**,推荐范围限定在清单内 |
+| Phase 3 LLM 写出离谱 instructions | 用户在 inbox 通知里点击 → 跳新 agent 详情页**编辑模式**,不直接进入"已就绪"状态。用户必须确认 |
+| 模板格式后续要演进(加字段) | Template JSON 加 `version` 字段,后端按 version 兼容老格式 |
+| Curated skill 清单过时(作者改 repo / 删 skill) | 加 CI 任务定期跑一遍清单 URL,挂掉的报警通知维护者 |
+
+---
+
+## 8. 不在本文档范围(已识别的下一步话题)
+
+- 跨 workspace 模板共享 / marketplace 化(用户能把自己的 agent 存成模板分享)
+- 实时 GitHub Code Search tool(Phase 2 进阶)
+- Server-side LLM 调用基础设施(如果未来需要 streaming 等场景)
+- ClawHub 下线决策(独立讨论,见 §5.5)
+- Skill 版本管理(workspace skill 版本号 / 升级提示)
+
+---
+
+## 附录 A:代码索引
+
+> 给接手开发的同事的快速参考。每条 file:line 都在本计划里被引用过,记录在这里方便跳转。
+
+| 主题 | 位置 |
+|---|---|
+| Skill DB 模型 | `server/migrations/008_structured_skills.up.sql:4-32` |
+| Skill 创建 handler + 事务 | `server/internal/handler/skill.go:143-162` + `skill_create.go:21-71` |
+| Skill import 入口(支持 3 个 source) | `server/internal/handler/skill.go:1538` |
+| Skill import source 分发 | `server/internal/handler/skill.go:586-617` (`detectImportSource`) |
+| Skills.sh fetcher | `server/internal/handler/skill.go:757-879` (`fetchFromSkillsSh`) |
+| GitHub fetcher | `server/internal/handler/skill.go:1363-1463` (`fetchFromGitHub`) |
+| ClawHub fetcher | `server/internal/handler/skill.go:642-744` (`fetchFromClawHub`) |
+| Agent 创建 handler | `server/internal/handler/agent.go:380-399` (request) + `:422-564` (CreateAgent) |
+| Agent 创建 sqlc | `server/pkg/db/queries/agent.sql:19-25` |
+| Agent-Skill 关联 sqlc | `server/pkg/db/queries/agent.sql:86-103` |
+| 当前 Agent Duplication(前端模式) | `packages/views/agents/components/agents-page.tsx:286-301`(post-create skill copy) |
+| Agent 创建 dialog | `packages/views/agents/components/create-agent-dialog.tsx` |
+| Skill add dialog | `packages/views/agents/components/skill-add-dialog.tsx` |
+| Quick-create Issue handler | `server/internal/handler/issue.go:877-982` (`QuickCreateIssue`) |
+| Quick-create task enqueue | `server/internal/service/task.go:488+` (`EnqueueQuickCreateTask`) |
+| Daemon claim + load skills | `server/internal/handler/daemon.go:1018-1098` + `service/task.go:1447-1463` |
+| Daemon prompt build | `server/internal/daemon/prompt.go:17-36` (dispatch) + `:45-106` (`buildQuickCreatePrompt`) |
+| Daemon execenv prepare | `server/internal/daemon/execenv/execenv.go:103-176` |
+| Skill 目录约定(runtime mapping) | `server/internal/daemon/execenv/context.go:121-158` (`resolveSkillsDir`) |
+| Skill 文件落盘 | `server/internal/daemon/execenv/context.go:175-204` (`writeSkillFiles`) |
+| Quick-create 完成检测 + inbox | `server/internal/service/task.go:1810-1949` |
+| LinkTaskToIssue | `server/internal/handler/agent.go:97-105` |
+| Quick-create Issue 前端 modal | `packages/views/modals/quick-create-issue.tsx:48-570+` |
+| Multica CLI 入口 | `server/cmd/multica/main.go:62-79` |
+| Skill CLI 命令 | `server/cmd/multica/cmd_skill.go:17-96`(已有 import,无 find) |
+| Agent CLI 命令 | `server/cmd/multica/cmd_agent.go:101-112`(已有 list/get,无 create) |

docs/analytics.md

@@ -366,6 +366,28 @@ not have a workspace yet.
 |---|---|---|
 | `workspace_id` | string (UUID) | Present only when the user already has a workspace. |
 | `source` | string | Always `onboarding`. |
+| `onboarding_session_id` | string (UUID) | Issued on this event and persisted to client storage. Stamped on every onboarding_* event until the funnel terminates. Lets HogQL correlate a full funnel back to a single start, even when `distinct_id` is shared across multiple sessions or skip paths. |
+
+## `onboarding_session_id`
+
+All in-product onboarding events carry an `onboarding_session_id` so the funnel
+can be reconstructed without joining on `distinct_id` alone. The id is generated
+client-side at `onboarding_started`, persisted across reloads, attached to every
+subsequent onboarding event, and cleared on `onboarding_completed`.
+
+Paths that bypass the in-product funnel emit `onboarding_completed` with the
+property omitted:
+
+- `skip_existing` from Welcome — the Welcome step clears the session before
+  completing, since the user never entered any real onboarding step. Their
+  earlier `onboarding_started` *does* carry a session id, but their completion
+  does not.
+- `invite_accept` — server-side completion path that never receives a session
+  id from the client.
+
+Funnel queries should filter `onboarding_session_id IS NOT NULL` on
+`onboarding_completed` to isolate real funnel completions from these
+soft-completions.
 
 ### `onboarding_questionnaire_submitted`
 
@@ -382,6 +404,7 @@ re-emit — the funnel counts users, not edits.
 | `team_size_has_other` | bool | `true` when the user filled the Q1 free-text escape. |
 | `role_has_other` | bool | Ditto Q2. |
 | `use_case_has_other` | bool | Ditto Q3. |
+| `onboarding_session_id` | string (UUID) | Forwarded from the client; lets the questionnaire submission join back to its `onboarding_started`. |
 
 Person properties set with `$set` (not once — users can go back and
 change answers before submitting again):
@@ -424,6 +447,7 @@ which exit the user took.
 | `workspace_id` | string (UUID) | Present for workspace-linked onboarding completions. |
 | `completion_path` | string | One of `full` / `runtime_skipped` / `cloud_waitlist` / `skip_existing` / `invite_accept` / `unknown`. See below. |
 | `joined_cloud_waitlist` | bool | Derived from `user.cloud_waitlist_email`. Orthogonal to `completion_path` — a user may submit the waitlist form and still pick CLI. |
+| `onboarding_session_id` | string (UUID) | Present for `full` / `runtime_skipped` / `cloud_waitlist` paths (the in-product funnel). Omitted for `skip_existing` / `invite_accept` because those bypass the funnel and never received a session. |
 
 Person properties set with `$set_once`:
 

packages/core/agents/queries.ts

@@ -82,3 +82,30 @@ export function agentTasksOptions(wsId: string, agentId: string) {
     refetchOnWindowFocus: true,
   });
 }
+
+// Agent templates are workspace-independent: a static catalog served from
+// the server's embedded JSON. Cache effectively forever — the only way the
+// list / detail change is a server deploy, and a hard reload picks that up.
+export const agentTemplateKeys = {
+  all: () => ["agent-templates"] as const,
+  list: () => [...agentTemplateKeys.all(), "list"] as const,
+  detail: (slug: string) => [...agentTemplateKeys.all(), "detail", slug] as const,
+};
+
+export function agentTemplateListOptions() {
+  return queryOptions({
+    queryKey: agentTemplateKeys.list(),
+    queryFn: () => api.listAgentTemplates(),
+    staleTime: Infinity,
+    gcTime: 30 * 60 * 1000,
+  });
+}
+
+export function agentTemplateDetailOptions(slug: string) {
+  return queryOptions({
+    queryKey: agentTemplateKeys.detail(slug),
+    queryFn: () => api.getAgentTemplate(slug),
+    staleTime: Infinity,
+    gcTime: 30 * 60 * 1000,
+  });
+}

packages/core/analytics/index.ts

@@ -13,6 +13,7 @@
 // backend returns an empty key and this module stays inert.
 
 import posthog from "posthog-js";
+import { getOnboardingSessionId } from "../onboarding/session";
 
 export const EVENT_SCHEMA_VERSION = 2;
 
@@ -283,6 +284,14 @@ function withClientEventProperties(
   if (next.is_demo === undefined) {
     next.is_demo = false;
   }
+  // Attach the active onboarding session id when one is in progress.
+  // Stamped on every client event (not just onboarding_*) so a stray
+  // event fired mid-funnel can still be joined back; HogQL filters by
+  // event name when it cares.
+  if (next.onboarding_session_id === undefined) {
+    const sessionId = getOnboardingSessionId();
+    if (sessionId) next.onboarding_session_id = sessionId;
+  }
   return next;
 }
 

packages/core/api/client.test.ts

@@ -200,6 +200,60 @@ describe("ApiClient", () => {
     });
   });
 
+  describe("getAttachmentTextContent", () => {
+    it("returns body text and the original content type from the X-* header", async () => {
+      vi.stubGlobal(
+        "fetch",
+        vi.fn().mockResolvedValue(
+          new Response("# heading\n\nbody\n", {
+            status: 200,
+            headers: {
+              "Content-Type": "text/plain; charset=utf-8",
+              "X-Original-Content-Type": "text/markdown",
+            },
+          }),
+        ),
+      );
+
+      const client = new ApiClient("https://api.example.test");
+      const { text, originalContentType } =
+        await client.getAttachmentTextContent("att-1");
+
+      expect(text).toBe("# heading\n\nbody\n");
+      expect(originalContentType).toBe("text/markdown");
+    });
+
+    it("throws PreviewTooLargeError on 413", async () => {
+      const { PreviewTooLargeError } = await import("./client");
+      vi.stubGlobal(
+        "fetch",
+        vi.fn().mockResolvedValue(
+          new Response("", { status: 413, statusText: "Payload Too Large" }),
+        ),
+      );
+
+      const client = new ApiClient("https://api.example.test");
+      await expect(client.getAttachmentTextContent("att-1")).rejects.toBeInstanceOf(
+        PreviewTooLargeError,
+      );
+    });
+
+    it("throws PreviewUnsupportedError on 415", async () => {
+      const { PreviewUnsupportedError } = await import("./client");
+      vi.stubGlobal(
+        "fetch",
+        vi.fn().mockResolvedValue(
+          new Response("", { status: 415, statusText: "Unsupported Media Type" }),
+        ),
+      );
+
+      const client = new ApiClient("https://api.example.test");
+      await expect(client.getAttachmentTextContent("att-1")).rejects.toBeInstanceOf(
+        PreviewUnsupportedError,
+      );
+    });
+  });
+
   describe("chat attachment wiring", () => {
     it("uploadFile includes chat_session_id in the FormData body", async () => {
       const fetchMock = vi.fn().mockResolvedValue(

packages/core/api/client.ts

@@ -11,6 +11,10 @@ import type {
   ListIssuesParams,
   Agent,
   CreateAgentRequest,
+  AgentTemplate,
+  AgentTemplateSummary,
+  CreateAgentFromTemplateRequest,
+  CreateAgentFromTemplateResponse,
   UpdateAgentRequest,
   AgentTask,
   AgentActivityBucket,
@@ -38,6 +42,9 @@ import type {
   RuntimeHourlyActivity,
   RuntimeUsageByAgent,
   RuntimeUsageByHour,
+  DashboardUsageDaily,
+  DashboardUsageByAgent,
+  DashboardAgentRunTime,
   RuntimeUpdate,
   RuntimeModelListRequest,
   RuntimeLocalSkillListRequest,
@@ -84,6 +91,8 @@ import type {
   GitHubPullRequest,
   ListGitHubInstallationsResponse,
   GitHubConnectResponse,
+  Squad,
+  SquadMember,
 } from "../types";
 import type { OnboardingCompletionPath } from "../onboarding/types";
 import { type Logger, noopLogger } from "../logger";
@@ -91,10 +100,19 @@ import { createRequestId } from "../utils";
 import { getCurrentSlug } from "../platform/workspace-storage";
 import { parseWithFallback } from "./schema";
 import {
+  AgentTemplateSchema,
+  AgentTemplateSummaryListSchema,
   AttachmentResponseSchema,
   ChildIssuesResponseSchema,
   CommentsListSchema,
+  CreateAgentFromTemplateResponseSchema,
+  DashboardAgentRunTimeListSchema,
+  DashboardUsageByAgentListSchema,
+  DashboardUsageDailyListSchema,
+  EMPTY_AGENT_TEMPLATE_DETAIL,
+  EMPTY_AGENT_TEMPLATE_SUMMARY_LIST,
   EMPTY_ATTACHMENT,
+  EMPTY_CREATE_AGENT_FROM_TEMPLATE_RESPONSE,
   EMPTY_LIST_ISSUES_RESPONSE,
   EMPTY_TIMELINE_ENTRIES,
   ListIssuesResponseSchema,
@@ -190,6 +208,27 @@ export class ApiError extends Error {
   }
 }
 
+// Thrown by getAttachmentTextContent when the server refuses to inline a
+// file because it exceeds the 2 MB cap. UI maps to a "too large, please
+// download" affordance with the Download CTA still available.
+export class PreviewTooLargeError extends Error {
+  constructor() {
+    super("attachment too large for inline preview");
+    this.name = "PreviewTooLargeError";
+  }
+}
+
+// Thrown by getAttachmentTextContent when the server's text whitelist
+// rejects the content type. Normally the client's isPreviewable() guard
+// catches this earlier, but the two whitelists can drift — surfacing the
+// 415 as a typed error makes the drift visible.
+export class PreviewUnsupportedError extends Error {
+  constructor() {
+    super("attachment type not supported for inline preview");
+    this.name = "PreviewUnsupportedError";
+  }
+}
+
 export class ApiClient {
   private baseUrl: string;
   private token: string | null = null;
@@ -264,15 +303,23 @@ export class ApiClient {
     }
   }
 
-  private async fetch<T>(path: string, init?: RequestInit): Promise<T> {
+  // Sends the request with the standard headers (auth, CSRF, request id,
+  // client identity) and runs the shared error path (401 → handleUnauthorized,
+  // structured ApiError, status-aware log level). Returns the raw Response so
+  // callers can decide how to decode the body — JSON for the typed `fetch<T>`
+  // path, plain text for the attachment-preview proxy, etc.
+  private async fetchRaw(
+    path: string,
+    init?: RequestInit & { extraHeaders?: Record<string, string> },
+  ): Promise<Response> {
     const rid = createRequestId();
     const start = Date.now();
     const method = init?.method ?? "GET";
 
     const headers: Record<string, string> = {
-      "Content-Type": "application/json",
       "X-Request-ID": rid,
       ...this.authHeaders(),
+      ...(init?.extraHeaders ?? {}),
       ...((init?.headers as Record<string, string>) ?? {}),
     };
 
@@ -293,12 +340,18 @@ export class ApiClient {
     }
 
     this.logger.info(`← ${res.status} ${path}`, { rid, duration: `${Date.now() - start}ms` });
+    return res;
+  }
 
+  private async fetch<T>(path: string, init?: RequestInit): Promise<T> {
+    const res = await this.fetchRaw(path, {
+      ...init,
+      extraHeaders: { "Content-Type": "application/json" },
+    });
     // Handle 204 No Content
     if (res.status === 204) {
       return undefined as T;
     }
-
     return res.json() as Promise<T>;
   }
 
@@ -339,6 +392,7 @@ export class ApiClient {
   async markOnboardingComplete(payload?: {
     completion_path?: OnboardingCompletionPath;
     workspace_id?: string;
+    onboarding_session_id?: string;
   }): Promise<User> {
     return this.fetch("/api/me/onboarding/complete", {
       method: "POST",
@@ -358,6 +412,7 @@ export class ApiClient {
 
   async patchOnboarding(payload: {
     questionnaire?: Record<string, unknown>;
+    onboarding_session_id?: string;
   }): Promise<User> {
     return this.fetch("/api/me/onboarding", {
       method: "PATCH",
@@ -628,6 +683,51 @@ export class ApiClient {
     });
   }
 
+  async listAgentTemplates(): Promise<AgentTemplateSummary[]> {
+    const raw = await this.fetch<unknown>("/api/agent-templates");
+    return parseWithFallback(
+      raw,
+      AgentTemplateSummaryListSchema,
+      EMPTY_AGENT_TEMPLATE_SUMMARY_LIST,
+      { endpoint: "GET /api/agent-templates" },
+    );
+  }
+
+  async getAgentTemplate(slug: string): Promise<AgentTemplate> {
+    const raw = await this.fetch<unknown>(
+      `/api/agent-templates/${encodeURIComponent(slug)}`,
+    );
+    // Round-trip the requested slug into the fallback so a malformed
+    // detail response still produces a navigable record matching the URL
+    // the user clicked.
+    return parseWithFallback(
+      raw,
+      AgentTemplateSchema,
+      { ...EMPTY_AGENT_TEMPLATE_DETAIL, slug },
+      { endpoint: "GET /api/agent-templates/:slug" },
+    );
+  }
+
+  /** Creates an agent from a curated template. The server fetches every
+   *  referenced skill URL in parallel, materializes them into the workspace
+   *  (find-or-create by name), and writes the agent + skill bindings in a
+   *  single transaction. On any upstream fetch failure, the entire write is
+   *  rolled back and the API returns 422 with `failed_urls`. */
+  async createAgentFromTemplate(
+    data: CreateAgentFromTemplateRequest,
+  ): Promise<CreateAgentFromTemplateResponse> {
+    const raw = await this.fetch<unknown>("/api/agents/from-template", {
+      method: "POST",
+      body: JSON.stringify(data),
+    });
+    return parseWithFallback(
+      raw,
+      CreateAgentFromTemplateResponseSchema,
+      EMPTY_CREATE_AGENT_FROM_TEMPLATE_RESPONSE,
+      { endpoint: "POST /api/agents/from-template" },
+    );
+  }
+
   async updateAgent(id: string, data: UpdateAgentRequest): Promise<Agent> {
     return this.fetch(`/api/agents/${id}`, {
       method: "PUT",
@@ -700,6 +800,58 @@ export class ApiClient {
     return this.fetch(`/api/runtimes/${runtimeId}/usage/by-hour?${search}`);
   }
 
+  // ---------------------------------------------------------------------------
+  // Workspace dashboard — three independent rollups for `/{slug}/dashboard`.
+  // Each accepts an optional `project_id` to narrow the scope to one project.
+  // Cost is computed client-side from the model pricing table (same contract
+  // as the per-runtime endpoints above).
+  // ---------------------------------------------------------------------------
+
+  async getDashboardUsageDaily(
+    params: { days?: number; project_id?: string | null },
+  ): Promise<DashboardUsageDaily[]> {
+    const search = new URLSearchParams();
+    if (params.days) search.set("days", String(params.days));
+    if (params.project_id) search.set("project_id", params.project_id);
+    const raw = await this.fetch<unknown>(`/api/dashboard/usage/daily?${search}`);
+    return parseWithFallback<DashboardUsageDaily[]>(
+      raw,
+      DashboardUsageDailyListSchema,
+      [],
+      { endpoint: "GET /api/dashboard/usage/daily" },
+    );
+  }
+
+  async getDashboardUsageByAgent(
+    params: { days?: number; project_id?: string | null },
+  ): Promise<DashboardUsageByAgent[]> {
+    const search = new URLSearchParams();
+    if (params.days) search.set("days", String(params.days));
+    if (params.project_id) search.set("project_id", params.project_id);
+    const raw = await this.fetch<unknown>(`/api/dashboard/usage/by-agent?${search}`);
+    return parseWithFallback<DashboardUsageByAgent[]>(
+      raw,
+      DashboardUsageByAgentListSchema,
+      [],
+      { endpoint: "GET /api/dashboard/usage/by-agent" },
+    );
+  }
+
+  async getDashboardAgentRunTime(
+    params: { days?: number; project_id?: string | null },
+  ): Promise<DashboardAgentRunTime[]> {
+    const search = new URLSearchParams();
+    if (params.days) search.set("days", String(params.days));
+    if (params.project_id) search.set("project_id", params.project_id);
+    const raw = await this.fetch<unknown>(`/api/dashboard/agent-runtime?${search}`);
+    return parseWithFallback<DashboardAgentRunTime[]>(
+      raw,
+      DashboardAgentRunTimeListSchema,
+      [],
+      { endpoint: "GET /api/dashboard/agent-runtime" },
+    );
+  }
+
   async initiateUpdate(
     runtimeId: string,
     targetVersion: string,
@@ -1080,6 +1232,13 @@ export class ApiClient {
     await this.fetch(`/api/chat/sessions/${id}`, { method: "DELETE" });
   }
 
+  async updateChatSession(id: string, data: { title: string }): Promise<ChatSession> {
+    return this.fetch(`/api/chat/sessions/${id}`, {
+      method: "PATCH",
+      body: JSON.stringify(data),
+    });
+  }
+
   async listChatMessages(sessionId: string): Promise<ChatMessage[]> {
     return this.fetch(`/api/chat/sessions/${sessionId}/messages`);
   }
@@ -1134,6 +1293,38 @@ export class ApiClient {
     await this.fetch(`/api/attachments/${id}`, { method: "DELETE" });
   }
 
+  // Fetches the raw bytes of a text-previewable attachment.
+  //
+  // The endpoint sidesteps CloudFront CORS (not configured on the CDN) and
+  // bypasses Content-Disposition: attachment for the `text/*` family, both
+  // of which would otherwise prevent the renderer from getting the body.
+  // The server always replies with `text/plain; charset=utf-8` for safety;
+  // the original MIME ships back in the `X-Original-Content-Type` header so
+  // the preview dispatcher can choose between markdown / html / plain code.
+  //
+  // Routes through `fetchRaw` so it inherits the standard auth headers,
+  // 401 → handleUnauthorized recovery, request-id logging, and ApiError
+  // shape. 413 / 415 are translated to typed `Preview*Error` instances so
+  // the modal can render specific fallbacks instead of generic failure.
+  async getAttachmentTextContent(
+    id: string,
+  ): Promise<{ text: string; originalContentType: string }> {
+    let res: Response;
+    try {
+      res = await this.fetchRaw(`/api/attachments/${id}/content`);
+    } catch (err) {
+      if (err instanceof ApiError) {
+        if (err.status === 413) throw new PreviewTooLargeError();
+        if (err.status === 415) throw new PreviewUnsupportedError();
+      }
+      throw err;
+    }
+    return {
+      text: await res.text(),
+      originalContentType: res.headers.get("X-Original-Content-Type") ?? "",
+    };
+  }
+
   // Projects
   async listProjects(params?: { status?: string }): Promise<ListProjectsResponse> {
     const search = new URLSearchParams();
@@ -1256,6 +1447,43 @@ export class ApiClient {
     });
   }
 
+  // Squads
+  async listSquads(): Promise<Squad[]> {
+    return this.fetch(`/api/squads`);
+  }
+
+  async getSquad(id: string): Promise<Squad> {
+    return this.fetch(`/api/squads/${id}`);
+  }
+
+  async createSquad(data: { name: string; description?: string; leader_id: string }): Promise<Squad> {
+    return this.fetch("/api/squads", { method: "POST", body: JSON.stringify(data) });
+  }
+
+  async updateSquad(id: string, data: { name?: string; description?: string; instructions?: string; leader_id?: string; avatar_url?: string }): Promise<Squad> {
+    return this.fetch(`/api/squads/${id}`, { method: "PUT", body: JSON.stringify(data) });
+  }
+
+  async deleteSquad(id: string): Promise<void> {
+    await this.fetch(`/api/squads/${id}`, { method: "DELETE" });
+  }
+
+  async listSquadMembers(squadId: string): Promise<SquadMember[]> {
+    return this.fetch(`/api/squads/${squadId}/members`);
+  }
+
+  async addSquadMember(squadId: string, data: { member_type: string; member_id: string; role?: string }): Promise<SquadMember> {
+    return this.fetch(`/api/squads/${squadId}/members`, { method: "POST", body: JSON.stringify(data) });
+  }
+
+  async removeSquadMember(squadId: string, data: { member_type: string; member_id: string }): Promise<void> {
+    await this.fetch(`/api/squads/${squadId}/members`, { method: "DELETE", body: JSON.stringify(data) });
+  }
+
+  async updateSquadMemberRole(squadId: string, data: { member_type: string; member_id: string; role: string }): Promise<SquadMember> {
+    return this.fetch(`/api/squads/${squadId}/members/role`, { method: "PATCH", body: JSON.stringify(data) });
+  }
+
   // Autopilots
   async listAutopilots(params?: { status?: string }): Promise<ListAutopilotsResponse> {
     const search = new URLSearchParams();

packages/core/api/index.ts

@@ -1,4 +1,9 @@
-export { ApiClient, ApiError } from "./client";
+export {
+  ApiClient,
+  ApiError,
+  PreviewTooLargeError,
+  PreviewUnsupportedError,
+} from "./client";
 export type {
   ApiClientOptions,
   ImportStarterContentPayload,

packages/core/api/schema.test.ts

@@ -117,6 +117,108 @@ describe("ApiClient schema fallback", () => {
       expect(res).toEqual({ issues: [] });
     });
   });
+
+  // Agent template catalog is hit by the desktop create-agent picker.
+  // Installed desktop builds outlive any given server, so the shape MUST
+  // survive future field renames / wrapping without crashing. Each test
+  // here mirrors a concrete future drift we want to absorb.
+  describe("listAgentTemplates", () => {
+    it("falls back to [] when the body is null", async () => {
+      stubFetchJson(null);
+      const client = new ApiClient("https://api.example.test");
+      const tmpls = await client.listAgentTemplates();
+      expect(tmpls).toEqual([]);
+    });
+
+    it("defaults skills to [] when the field is missing from a template", async () => {
+      // Future server: drops `skills` because the picker no longer reads
+      // them. Picker code calls `template.skills.length` — must not throw.
+      stubFetchJson([{ slug: "x", name: "X" }]);
+      const client = new ApiClient("https://api.example.test");
+      const tmpls = await client.listAgentTemplates();
+      expect(tmpls).toHaveLength(1);
+      expect(tmpls[0]?.skills).toEqual([]);
+    });
+
+    it("accepts the bare-array shape (current contract)", async () => {
+      stubFetchJson([
+        { slug: "a", name: "A", description: "", skills: [] },
+        { slug: "b", name: "B", description: "", skills: [] },
+      ]);
+      const client = new ApiClient("https://api.example.test");
+      const tmpls = await client.listAgentTemplates();
+      expect(tmpls.map((t) => t.slug)).toEqual(["a", "b"]);
+    });
+
+    it("accepts a future {templates: [...]} envelope without breaking", async () => {
+      // Server migrates to a paginated envelope. We unwrap so the picker
+      // keeps working on the older bare-array consumer.
+      stubFetchJson({
+        templates: [{ slug: "a", name: "A", description: "", skills: [] }],
+        total: 1,
+      });
+      const client = new ApiClient("https://api.example.test");
+      const tmpls = await client.listAgentTemplates();
+      expect(tmpls).toHaveLength(1);
+      expect(tmpls[0]?.slug).toBe("a");
+    });
+  });
+
+  describe("getAgentTemplate", () => {
+    it("falls back to a minimal record carrying the requested slug", async () => {
+      // Slug is part of the URL the user clicked — the fallback round-
+      // trips it so the page header still makes sense after a parse miss.
+      stubFetchJson({ wrong: "shape" });
+      const client = new ApiClient("https://api.example.test");
+      const detail = await client.getAgentTemplate("code-reviewer");
+      expect(detail.slug).toBe("code-reviewer");
+      expect(detail.skills).toEqual([]);
+      expect(detail.instructions).toBe("");
+    });
+
+    it("defaults instructions to '' when the field is missing", async () => {
+      stubFetchJson({
+        slug: "code-reviewer",
+        name: "Code Reviewer",
+        description: "",
+        skills: [],
+      });
+      const client = new ApiClient("https://api.example.test");
+      const detail = await client.getAgentTemplate("code-reviewer");
+      expect(detail.instructions).toBe("");
+    });
+  });
+
+  describe("createAgentFromTemplate", () => {
+    it("falls back to an empty agent when the response is malformed", async () => {
+      // The agent was created server-side even though the client can't
+      // parse the response — UI code reads `agent.id === ""` and skips
+      // the navigation step rather than landing on `/agents/`.
+      stubFetchJson({ unexpected: "shape" });
+      const client = new ApiClient("https://api.example.test");
+      const resp = await client.createAgentFromTemplate({
+        template_slug: "x",
+        name: "X",
+        runtime_id: "rt-1",
+      });
+      expect(resp.agent.id).toBe("");
+      expect(resp.imported_skill_ids).toEqual([]);
+      expect(resp.reused_skill_ids).toEqual([]);
+    });
+
+    it("defaults imported_skill_ids / reused_skill_ids to [] when missing", async () => {
+      stubFetchJson({ agent: { id: "agent-1" } });
+      const client = new ApiClient("https://api.example.test");
+      const resp = await client.createAgentFromTemplate({
+        template_slug: "x",
+        name: "X",
+        runtime_id: "rt-1",
+      });
+      expect(resp.agent.id).toBe("agent-1");
+      expect(resp.imported_skill_ids).toEqual([]);
+      expect(resp.reused_skill_ids).toEqual([]);
+    });
+  });
 });
 
 // Direct tests for the helper, decoupled from any specific endpoint —

packages/core/api/schemas.ts

@@ -1,5 +1,13 @@
 import { z } from "zod";
-import type { Attachment, ListIssuesResponse, TimelineEntry } from "../types";
+import type {
+  Agent,
+  AgentTemplate,
+  AgentTemplateSummary,
+  Attachment,
+  CreateAgentFromTemplateResponse,
+  ListIssuesResponse,
+  TimelineEntry,
+} from "../types";
 
 // ---------------------------------------------------------------------------
 // Schemas for the highest-risk API endpoints — those whose responses drive
@@ -169,3 +177,132 @@ export const SubscribersListSchema = z.array(SubscriberSchema);
 export const ChildIssuesResponseSchema = z.object({
   issues: z.array(IssueSchema).default([]),
 }).loose();
+
+// ---------------------------------------------------------------------------
+// Workspace dashboard schemas
+//
+// The dashboard hits three independent rollup endpoints. Each returns a flat
+// array, and every field is consumed by chart / KPI math — a missing number
+// silently degrades to NaN downstream, so we coerce missing numbers to 0.
+// String fields stay lenient (no enum narrowing) to survive future model /
+// agent ID drift.
+// ---------------------------------------------------------------------------
+
+const DashboardUsageDailySchema = z.object({
+  date: z.string(),
+  model: z.string(),
+  input_tokens: z.number().default(0),
+  output_tokens: z.number().default(0),
+  cache_read_tokens: z.number().default(0),
+  cache_write_tokens: z.number().default(0),
+  task_count: z.number().default(0),
+}).loose();
+
+export const DashboardUsageDailyListSchema = z.array(DashboardUsageDailySchema);
+
+const DashboardUsageByAgentSchema = z.object({
+  agent_id: z.string(),
+  model: z.string(),
+  input_tokens: z.number().default(0),
+  output_tokens: z.number().default(0),
+  cache_read_tokens: z.number().default(0),
+  cache_write_tokens: z.number().default(0),
+  task_count: z.number().default(0),
+}).loose();
+
+export const DashboardUsageByAgentListSchema = z.array(DashboardUsageByAgentSchema);
+
+const DashboardAgentRunTimeSchema = z.object({
+  agent_id: z.string(),
+  total_seconds: z.number().default(0),
+  task_count: z.number().default(0),
+  failed_count: z.number().default(0),
+}).loose();
+
+export const DashboardAgentRunTimeListSchema = z.array(DashboardAgentRunTimeSchema);
+
+// ---------------------------------------------------------------------------
+// Agent template catalog — `/api/agent-templates*` and the
+// create-from-template response. The desktop app's create-agent picker
+// reaches these endpoints, and a future server change to the template shape
+// would white-screen older installed builds (#2192 pattern) without these
+// parsers. Lenient by the same rules as IssueSchema above: arrays default to
+// `[]`, optional fields stay optional, `.loose()` lets unknown fields pass
+// through unchanged.
+// ---------------------------------------------------------------------------
+
+const AgentTemplateSkillRefSchema = z.object({
+  source_url: z.string(),
+  cached_name: z.string().default(""),
+  cached_description: z.string().default(""),
+}).loose();
+
+const AgentTemplateSummarySchemaBase = z.object({
+  slug: z.string(),
+  name: z.string(),
+  description: z.string().default(""),
+  category: z.string().optional(),
+  icon: z.string().optional(),
+  accent: z.string().optional(),
+  // skills MUST default to [] — picker code reads `template.skills.length`
+  // and `.map(...)`, both of which crash on `undefined`. The most common
+  // future drift (field renamed / wrapped) lands here.
+  skills: z.array(AgentTemplateSkillRefSchema).default([]),
+}).loose();
+
+export const AgentTemplateSummarySchema = AgentTemplateSummarySchemaBase;
+
+// List endpoint historically returns a bare array. Server could legitimately
+// migrate to `{templates: [...]}` later — we accept either shape so an old
+// desktop survives the upgrade.
+export const AgentTemplateSummaryListSchema = z.union([
+  z.array(AgentTemplateSummarySchemaBase),
+  z.object({ templates: z.array(AgentTemplateSummarySchemaBase).default([]) })
+    .loose()
+    .transform((v) => v.templates),
+]);
+
+export const EMPTY_AGENT_TEMPLATE_SUMMARY_LIST: AgentTemplateSummary[] = [];
+
+export const AgentTemplateSchema = AgentTemplateSummarySchemaBase.extend({
+  // Detail-only field. Default "" so a malformed detail still renders the
+  // header + skill list; the user just sees an empty Instructions block.
+  instructions: z.string().default(""),
+}).loose();
+
+// Used as the parse fallback for `GET /api/agent-templates/:slug`. Slug comes
+// from the URL, so we round-trip the requested one back into the fallback
+// at the call site (see `getAgentTemplate` in client.ts).
+export const EMPTY_AGENT_TEMPLATE_DETAIL: AgentTemplate = {
+  slug: "",
+  name: "",
+  description: "",
+  skills: [],
+  instructions: "",
+};
+
+// `agent` is a full Agent record — schematising every field would duplicate
+// a 50-field interface and bit-rot fast. We keep it loose and require only
+// `id`, the one field the create-from-template flow consumes (used to
+// navigate to the new agent's detail page). Downstream code already
+// optional-chains the rest.
+const MinimalAgentSchema = z.object({
+  id: z.string(),
+}).loose();
+
+export const CreateAgentFromTemplateResponseSchema = z.object({
+  agent: MinimalAgentSchema,
+  imported_skill_ids: z.array(z.string()).default([]),
+  reused_skill_ids: z.array(z.string()).default([]),
+}).loose();
+
+// Fallback when the success response fails to parse. The agent server-side
+// has likely been created already, so we can't pretend nothing happened —
+// the caller (`create-agent-dialog.tsx`) is responsible for noticing
+// `agent.id === ""` and skipping navigation while keeping the list
+// invalidation, so the user finds their new agent in the list.
+export const EMPTY_CREATE_AGENT_FROM_TEMPLATE_RESPONSE: CreateAgentFromTemplateResponse = {
+  agent: { id: "" } as Agent,
+  imported_skill_ids: [],
+  reused_skill_ids: [],
+};

packages/core/chat/mutations.ts

@@ -64,6 +64,45 @@ export function useMarkChatSessionRead() {
   });
 }
 
+/**
+ * Renames a chat session. Optimistically swaps the title in the cached
+ * list so the dropdown reflects the new label immediately; rolls back on
+ * error. The matching `chat:session_updated` WS event keeps other
+ * tabs/devices in sync — see use-realtime-sync.ts.
+ */
+export function useUpdateChatSession() {
+  const qc = useQueryClient();
+  const wsId = useWorkspaceId();
+
+  return useMutation({
+    mutationFn: (data: { sessionId: string; title: string }) => {
+      logger.info("updateChatSession.start", {
+        sessionId: data.sessionId,
+        titleLength: data.title.length,
+      });
+      return api.updateChatSession(data.sessionId, { title: data.title });
+    },
+    onMutate: async ({ sessionId, title }) => {
+      await qc.cancelQueries({ queryKey: chatKeys.sessions(wsId) });
+
+      const prevSessions = qc.getQueryData<ChatSession[]>(chatKeys.sessions(wsId));
+
+      const patch = (old?: ChatSession[]) =>
+        old?.map((s) => (s.id === sessionId ? { ...s, title } : s));
+      qc.setQueryData<ChatSession[]>(chatKeys.sessions(wsId), patch);
+
+      return { prevSessions };
+    },
+    onError: (err, vars, ctx) => {
+      logger.error("updateChatSession.error.rollback", { sessionId: vars.sessionId, err });
+      if (ctx?.prevSessions) qc.setQueryData(chatKeys.sessions(wsId), ctx.prevSessions);
+    },
+    onSettled: () => {
+      qc.invalidateQueries({ queryKey: chatKeys.sessions(wsId) });
+    },
+  });
+}
+
 /**
  * Hard-deletes a chat session. Optimistically removes the row from the
  * sessions list so the dropdown updates instantly; rolls back on error.

packages/core/dashboard/index.ts

@@ -0,0 +1 @@
+export * from "./queries";

packages/core/dashboard/queries.ts

@@ -0,0 +1,72 @@
+import { queryOptions } from "@tanstack/react-query";
+import { api } from "../api";
+
+// Workspace dashboard query options. All three endpoints share the same
+// (wsId, days, projectId) key shape so workspace switching, time-range
+// changes, and the project filter each invalidate the cache cleanly.
+//
+// The cache key includes `wsId` explicitly: TanStack Query already isolates
+// per workspace via the key, but threading wsId into the queryFn lets
+// callers fail fast (return [] on empty wsId) instead of issuing a request
+// the server would reject.
+//
+// `projectId` is normalised to `null` (not undefined / "all") so the
+// queryKey shape is stable across renders even when the dropdown sits on
+// "all projects".
+
+export const dashboardKeys = {
+  all: (wsId: string) => ["dashboard", wsId] as const,
+  daily: (wsId: string, days: number, projectId: string | null) =>
+    [...dashboardKeys.all(wsId), "daily", days, projectId] as const,
+  byAgent: (wsId: string, days: number, projectId: string | null) =>
+    [...dashboardKeys.all(wsId), "by-agent", days, projectId] as const,
+  agentRuntime: (wsId: string, days: number, projectId: string | null) =>
+    [...dashboardKeys.all(wsId), "agent-runtime", days, projectId] as const,
+};
+
+// 60s staleTime matches the per-runtime usage queries — the data is rollup-
+// driven on the server (5-min rollup cadence) and the dashboard isn't a
+// real-time view, so background refetches every minute are plenty.
+const STALE_TIME = 60 * 1000;
+
+export function dashboardUsageDailyOptions(
+  wsId: string,
+  days: number,
+  projectId: string | null,
+) {
+  return queryOptions({
+    queryKey: dashboardKeys.daily(wsId, days, projectId),
+    queryFn: () =>
+      api.getDashboardUsageDaily({ days, project_id: projectId ?? undefined }),
+    enabled: !!wsId,
+    staleTime: STALE_TIME,
+  });
+}
+
+export function dashboardUsageByAgentOptions(
+  wsId: string,
+  days: number,
+  projectId: string | null,
+) {
+  return queryOptions({
+    queryKey: dashboardKeys.byAgent(wsId, days, projectId),
+    queryFn: () =>
+      api.getDashboardUsageByAgent({ days, project_id: projectId ?? undefined }),
+    enabled: !!wsId,
+    staleTime: STALE_TIME,
+  });
+}
+
+export function dashboardAgentRunTimeOptions(
+  wsId: string,
+  days: number,
+  projectId: string | null,
+) {
+  return queryOptions({
+    queryKey: dashboardKeys.agentRuntime(wsId, days, projectId),
+    queryFn: () =>
+      api.getDashboardAgentRunTime({ days, project_id: projectId ?? undefined }),
+    enabled: !!wsId,
+    staleTime: STALE_TIME,
+  });
+}

packages/core/issues/delete-cache.ts

@@ -0,0 +1,166 @@
+import type { QueryClient, QueryKey } from "@tanstack/react-query";
+import {
+  agentActivityKeys,
+  agentRunCountsKeys,
+  agentTaskSnapshotKeys,
+  agentTasksKeys,
+} from "../agents/queries";
+import { labelKeys } from "../labels/queries";
+import type { Issue, ListIssuesCache } from "../types";
+import { findIssueLocation, removeIssueFromBuckets } from "./cache-helpers";
+import { issueKeys } from "./queries";
+
+export type DeletedIssueCacheMetadata = {
+  parentIssueIds: string[];
+};
+
+function collectParentId(
+  parentIssueIds: Set<string>,
+  parentId: string | null | undefined,
+) {
+  if (parentId) parentIssueIds.add(parentId);
+}
+
+function collectParentFromListCache(
+  parentIssueIds: Set<string>,
+  data: ListIssuesCache | undefined,
+  issueId: string,
+) {
+  const parentId = data
+    ? findIssueLocation(data, issueId)?.issue.parent_issue_id
+    : undefined;
+  collectParentId(parentIssueIds, parentId);
+}
+
+function parentIdFromChildrenKey(key: QueryKey) {
+  const parentId = key[key.length - 1];
+  return typeof parentId === "string" ? parentId : null;
+}
+
+export function collectDeletedIssueCacheMetadata(
+  qc: QueryClient,
+  wsId: string,
+  issueId: string,
+): DeletedIssueCacheMetadata {
+  const parentIssueIds = new Set<string>();
+
+  const detail = qc.getQueryData<Issue>(issueKeys.detail(wsId, issueId));
+  collectParentId(parentIssueIds, detail?.parent_issue_id);
+
+  collectParentFromListCache(
+    parentIssueIds,
+    qc.getQueryData<ListIssuesCache>(issueKeys.list(wsId)),
+    issueId,
+  );
+
+  for (const [, data] of qc.getQueriesData<ListIssuesCache>({
+    queryKey: issueKeys.myAll(wsId),
+  })) {
+    collectParentFromListCache(parentIssueIds, data, issueId);
+  }
+
+  for (const [key, data] of qc.getQueriesData<Issue[]>({
+    queryKey: [...issueKeys.all(wsId), "children"],
+  })) {
+    const child = data?.find((issue) => issue.id === issueId);
+    if (!child) continue;
+    collectParentId(parentIssueIds, child.parent_issue_id);
+    collectParentId(parentIssueIds, parentIdFromChildrenKey(key));
+  }
+
+  return { parentIssueIds: Array.from(parentIssueIds) };
+}
+
+export function pruneDeletedIssueFromListCaches(
+  qc: QueryClient,
+  wsId: string,
+  issueId: string,
+) {
+  qc.setQueryData<ListIssuesCache>(issueKeys.list(wsId), (old) =>
+    old ? removeIssueFromBuckets(old, issueId) : old,
+  );
+
+  for (const [key] of qc.getQueriesData<ListIssuesCache>({
+    queryKey: issueKeys.myAll(wsId),
+  })) {
+    qc.setQueryData<ListIssuesCache>(key, (old) =>
+      old ? removeIssueFromBuckets(old, issueId) : old,
+    );
+  }
+}
+
+export function pruneDeletedIssueFromParentChildrenCaches(
+  qc: QueryClient,
+  wsId: string,
+  issueId: string,
+  metadata: DeletedIssueCacheMetadata,
+) {
+  for (const parentId of metadata.parentIssueIds) {
+    qc.setQueryData<Issue[]>(issueKeys.children(wsId, parentId), (old) =>
+      old?.filter((issue) => issue.id !== issueId),
+    );
+  }
+}
+
+export function invalidateDeletedIssueParentCaches(
+  qc: QueryClient,
+  wsId: string,
+  metadata: DeletedIssueCacheMetadata,
+) {
+  if (metadata.parentIssueIds.length === 0) return;
+  for (const parentId of metadata.parentIssueIds) {
+    qc.invalidateQueries({ queryKey: issueKeys.children(wsId, parentId) });
+  }
+  qc.invalidateQueries({ queryKey: issueKeys.childProgress(wsId) });
+}
+
+export function invalidateDeletedIssueDependentCaches(
+  qc: QueryClient,
+  wsId: string,
+) {
+  qc.invalidateQueries({ queryKey: agentTaskSnapshotKeys.list(wsId) });
+  qc.invalidateQueries({ queryKey: agentActivityKeys.last30d(wsId) });
+  qc.invalidateQueries({ queryKey: agentRunCountsKeys.last30d(wsId) });
+  qc.invalidateQueries({ queryKey: agentTasksKeys.all(wsId) });
+}
+
+export function invalidateIssueScopedCaches(
+  qc: QueryClient,
+  wsId: string,
+  issueId: string,
+) {
+  qc.invalidateQueries({ queryKey: issueKeys.timeline(issueId) });
+  qc.invalidateQueries({ queryKey: issueKeys.reactions(issueId) });
+  qc.invalidateQueries({ queryKey: issueKeys.subscribers(issueId) });
+  qc.invalidateQueries({ queryKey: issueKeys.usage(issueId) });
+  qc.invalidateQueries({ queryKey: issueKeys.attachments(issueId) });
+  qc.invalidateQueries({ queryKey: issueKeys.tasks(issueId) });
+  qc.invalidateQueries({ queryKey: issueKeys.children(wsId, issueId) });
+  qc.invalidateQueries({ queryKey: labelKeys.byIssue(wsId, issueId) });
+}
+
+export function cleanupDeletedIssueCaches(
+  qc: QueryClient,
+  wsId: string,
+  issueId: string,
+  metadata = collectDeletedIssueCacheMetadata(qc, wsId, issueId),
+) {
+  pruneDeletedIssueFromListCaches(qc, wsId, issueId);
+  pruneDeletedIssueFromParentChildrenCaches(qc, wsId, issueId, metadata);
+  invalidateDeletedIssueParentCaches(qc, wsId, metadata);
+
+  qc.removeQueries({ queryKey: issueKeys.detail(wsId, issueId) });
+  qc.removeQueries({ queryKey: issueKeys.timeline(issueId) });
+  qc.removeQueries({ queryKey: issueKeys.reactions(issueId) });
+  qc.removeQueries({ queryKey: issueKeys.subscribers(issueId) });
+  qc.removeQueries({ queryKey: issueKeys.usage(issueId) });
+  qc.removeQueries({ queryKey: issueKeys.attachments(issueId) });
+  qc.removeQueries({ queryKey: issueKeys.tasks(issueId) });
+  qc.removeQueries({ queryKey: issueKeys.children(wsId, issueId) });
+  qc.removeQueries({ queryKey: labelKeys.byIssue(wsId, issueId) });
+
+  qc.invalidateQueries({ queryKey: issueKeys.childProgress(wsId) });
+  qc.invalidateQueries({ queryKey: issueKeys.list(wsId) });
+  qc.invalidateQueries({ queryKey: issueKeys.myAll(wsId) });
+  invalidateDeletedIssueDependentCaches(qc, wsId);
+}

packages/core/issues/mutations.ts

@@ -11,9 +11,17 @@ import {
   findIssueLocation,
   getBucket,
   patchIssueInBuckets,
-  removeIssueFromBuckets,
   setBucket,
 } from "./cache-helpers";
+import {
+  cleanupDeletedIssueCaches,
+  collectDeletedIssueCacheMetadata,
+  invalidateDeletedIssueDependentCaches,
+  invalidateDeletedIssueParentCaches,
+  invalidateIssueScopedCaches,
+  pruneDeletedIssueFromListCaches,
+  pruneDeletedIssueFromParentChildrenCaches,
+} from "./delete-cache";
 import { useWorkspaceId } from "../hooks";
 import { useRecentIssuesStore } from "./stores";
 import type { Issue, IssueReaction, IssueStatus } from "../types";
@@ -217,24 +225,56 @@ export function useDeleteIssue() {
   return useMutation({
     mutationFn: (id: string) => api.deleteIssue(id),
     onMutate: async (id) => {
-      await qc.cancelQueries({ queryKey: issueKeys.list(wsId) });
-      const prevList = qc.getQueryData<ListIssuesCache>(issueKeys.list(wsId));
-      const deleted = prevList ? findIssueLocation(prevList, id)?.issue : undefined;
-      qc.setQueryData<ListIssuesCache>(issueKeys.list(wsId), (old) =>
-        old ? removeIssueFromBuckets(old, id) : old,
+      await Promise.all([
+        qc.cancelQueries({ queryKey: issueKeys.list(wsId) }),
+        qc.cancelQueries({ queryKey: issueKeys.myAll(wsId) }),
+      ]);
+      const metadata = collectDeletedIssueCacheMetadata(qc, wsId, id);
+      await Promise.all(
+        metadata.parentIssueIds.map((parentId) =>
+          qc.cancelQueries({ queryKey: issueKeys.children(wsId, parentId) }),
+        ),
       );
+      const prevList = qc.getQueryData<ListIssuesCache>(issueKeys.list(wsId));
+      const prevMyLists = qc.getQueriesData<ListIssuesCache>({
+        queryKey: issueKeys.myAll(wsId),
+      });
+      const prevDetail = qc.getQueryData<Issue>(issueKeys.detail(wsId, id));
+      const prevChildren = new Map<string, Issue[] | undefined>();
+      for (const parentId of metadata.parentIssueIds) {
+        prevChildren.set(
+          parentId,
+          qc.getQueryData<Issue[]>(issueKeys.children(wsId, parentId)),
+        );
+      }
+
+      pruneDeletedIssueFromListCaches(qc, wsId, id);
+      pruneDeletedIssueFromParentChildrenCaches(qc, wsId, id, metadata);
       qc.removeQueries({ queryKey: issueKeys.detail(wsId, id) });
-      return { prevList, parentIssueId: deleted?.parent_issue_id };
+      return { id, metadata, prevList, prevMyLists, prevDetail, prevChildren };
     },
     onError: (_err, _id, ctx) => {
       if (ctx?.prevList) qc.setQueryData(issueKeys.list(wsId), ctx.prevList);
+      if (ctx?.prevMyLists) {
+        for (const [key, snapshot] of ctx.prevMyLists) {
+          qc.setQueryData(key, snapshot);
+        }
+      }
+      if (ctx?.prevDetail) {
+        qc.setQueryData(issueKeys.detail(wsId, ctx.id), ctx.prevDetail);
+      }
+      if (ctx?.prevChildren) {
+        for (const [parentId, snapshot] of ctx.prevChildren) {
+          qc.setQueryData(issueKeys.children(wsId, parentId), snapshot);
+        }
+      }
+    },
+    onSuccess: (_data, id, ctx) => {
+      cleanupDeletedIssueCaches(qc, wsId, id, ctx?.metadata);
     },
     onSettled: (_data, _err, _id, ctx) => {
       qc.invalidateQueries({ queryKey: issueKeys.list(wsId) });
-      if (ctx?.parentIssueId) {
-        qc.invalidateQueries({ queryKey: issueKeys.children(wsId, ctx.parentIssueId) });
-        qc.invalidateQueries({ queryKey: issueKeys.childProgress(wsId) });
-      }
+      if (ctx?.metadata) invalidateDeletedIssueParentCaches(qc, wsId, ctx.metadata);
     },
   });
 }
@@ -309,57 +349,92 @@ export function useBatchDeleteIssues() {
   return useMutation({
     mutationFn: (ids: string[]) => api.batchDeleteIssues(ids),
     onMutate: async (ids) => {
-      await qc.cancelQueries({ queryKey: issueKeys.list(wsId) });
-      const prevList = qc.getQueryData<ListIssuesCache>(issueKeys.list(wsId));
+      await Promise.all([
+        qc.cancelQueries({ queryKey: issueKeys.list(wsId) }),
+        qc.cancelQueries({ queryKey: issueKeys.myAll(wsId) }),
+      ]);
+      const metadataById = new Map(
+        ids.map((id) => [
+          id,
+          collectDeletedIssueCacheMetadata(qc, wsId, id),
+        ]),
+      );
       const parentIssueIds = new Set<string>();
-      if (prevList) {
-        for (const id of ids) {
-          const loc = findIssueLocation(prevList, id);
-          if (loc?.issue.parent_issue_id) parentIssueIds.add(loc.issue.parent_issue_id);
+      for (const metadata of metadataById.values()) {
+        for (const parentId of metadata.parentIssueIds) {
+          parentIssueIds.add(parentId);
         }
       }
-      // Children cache may be the only place sub-issues live when the user
-      // operates from a parent's detail page. Collect affected parents and
-      // optimistically filter the deleted ids out of each children cache so
-      // the row disappears immediately, mirroring the list-cache behaviour.
-      const idSet = new Set(ids);
-      const childrenCaches = qc.getQueriesData<Issue[]>({
-        queryKey: [...issueKeys.all(wsId), "children"],
+      await Promise.all(
+        Array.from(parentIssueIds).map((parentId) =>
+          qc.cancelQueries({ queryKey: issueKeys.children(wsId, parentId) }),
+        ),
+      );
+      const prevList = qc.getQueryData<ListIssuesCache>(issueKeys.list(wsId));
+      const prevMyLists = qc.getQueriesData<ListIssuesCache>({
+        queryKey: issueKeys.myAll(wsId),
       });
       const prevChildren = new Map<string, Issue[] | undefined>();
-      for (const [key, data] of childrenCaches) {
-        if (!data?.some((c) => idSet.has(c.id))) continue;
-        const parentId = key[key.length - 1];
-        if (typeof parentId !== "string") continue;
-        parentIssueIds.add(parentId);
-        prevChildren.set(parentId, data);
-        qc.setQueryData<Issue[]>(issueKeys.children(wsId, parentId), (old) =>
-          old?.filter((c) => !idSet.has(c.id)),
+      for (const parentId of parentIssueIds) {
+        prevChildren.set(
+          parentId,
+          qc.getQueryData<Issue[]>(issueKeys.children(wsId, parentId)),
         );
       }
-      qc.setQueryData<ListIssuesCache>(issueKeys.list(wsId), (old) => {
-        if (!old) return old;
-        let next = old;
-        for (const id of ids) next = removeIssueFromBuckets(next, id);
-        return next;
-      });
-      return { prevList, prevChildren, parentIssueIds };
+
+      for (const id of ids) {
+        const metadata = metadataById.get(id);
+        pruneDeletedIssueFromListCaches(qc, wsId, id);
+        if (metadata) {
+          pruneDeletedIssueFromParentChildrenCaches(qc, wsId, id, metadata);
+        }
+      }
+      return { prevList, prevMyLists, prevChildren, parentIssueIds, metadataById };
     },
     onError: (_err, _ids, ctx) => {
       if (ctx?.prevList) qc.setQueryData(issueKeys.list(wsId), ctx.prevList);
+      if (ctx?.prevMyLists) {
+        for (const [key, snapshot] of ctx.prevMyLists) {
+          qc.setQueryData(key, snapshot);
+        }
+      }
       if (ctx?.prevChildren) {
         for (const [parentId, snapshot] of ctx.prevChildren) {
           qc.setQueryData(issueKeys.children(wsId, parentId), snapshot);
         }
       }
     },
+    onSuccess: (data, ids, ctx) => {
+      if (data.deleted === ids.length) {
+        for (const id of ids) {
+          cleanupDeletedIssueCaches(qc, wsId, id, ctx?.metadataById.get(id));
+        }
+        return;
+      }
+
+      if (ctx?.prevList) qc.setQueryData(issueKeys.list(wsId), ctx.prevList);
+      if (ctx?.prevMyLists) {
+        for (const [key, snapshot] of ctx.prevMyLists) {
+          qc.setQueryData(key, snapshot);
+        }
+      }
+      if (ctx?.prevChildren) {
+        for (const [parentId, snapshot] of ctx.prevChildren) {
+          qc.setQueryData(issueKeys.children(wsId, parentId), snapshot);
+        }
+      }
+      for (const id of ids) {
+        invalidateIssueScopedCaches(qc, wsId, id);
+      }
+      qc.invalidateQueries({ queryKey: issueKeys.all(wsId) });
+      invalidateDeletedIssueDependentCaches(qc, wsId);
+    },
     onSettled: (_data, _err, _ids, ctx) => {
       qc.invalidateQueries({ queryKey: issueKeys.list(wsId) });
       if (ctx?.parentIssueIds && ctx.parentIssueIds.size > 0) {
-        for (const parentId of ctx.parentIssueIds) {
-          qc.invalidateQueries({ queryKey: issueKeys.children(wsId, parentId) });
-        }
-        qc.invalidateQueries({ queryKey: issueKeys.childProgress(wsId) });
+        invalidateDeletedIssueParentCaches(qc, wsId, {
+          parentIssueIds: Array.from(ctx.parentIssueIds),
+        });
       }
     },
   });

packages/core/issues/ws-updaters.test.ts

@@ -1,17 +1,34 @@
 import { beforeEach, describe, expect, it } from "vitest";
 import { QueryClient } from "@tanstack/react-query";
-import { onIssueLabelsChanged } from "./ws-updaters";
+import {
+  agentActivityKeys,
+  agentRunCountsKeys,
+  agentTaskSnapshotKeys,
+  agentTasksKeys,
+} from "../agents/queries";
+import { onIssueDeleted, onIssueLabelsChanged } from "./ws-updaters";
 import { issueKeys } from "./queries";
 import { labelKeys } from "../labels/queries";
 import type {
+  AgentActivityBucket,
+  AgentRunCount,
+  AgentTask,
+  Attachment,
   Issue,
+  IssueReaction,
   IssueLabelsResponse,
+  IssueSubscriber,
+  IssueUsageSummary,
   Label,
   ListIssuesCache,
+  TimelineEntry,
 } from "../types";
 
 const WS_ID = "ws-1";
 const ISSUE_ID = "issue-1";
+const OTHER_ISSUE_ID = "issue-2";
+const PARENT_ISSUE_ID = "parent-1";
+const AGENT_ID = "agent-1";
 
 const labelA: Label = {
   id: "label-a",
@@ -53,6 +70,47 @@ const baseIssue: Issue = {
   updated_at: "2025-01-01T00:00:00Z",
 };
 
+const parentedIssue: Issue = {
+  ...baseIssue,
+  parent_issue_id: PARENT_ISSUE_ID,
+};
+
+const otherIssue: Issue = {
+  ...baseIssue,
+  id: OTHER_ISSUE_ID,
+  identifier: "MUL-2",
+  title: "Other",
+};
+
+function makeListCache(...issues: Issue[]): ListIssuesCache {
+  return {
+    byStatus: {
+      todo: { issues, total: issues.length },
+    },
+  };
+}
+
+function makeTask(issueId = ISSUE_ID): AgentTask {
+  return {
+    id: `task-${issueId}`,
+    agent_id: AGENT_ID,
+    runtime_id: "runtime-1",
+    issue_id: issueId,
+    status: "completed",
+    priority: 0,
+    dispatched_at: null,
+    started_at: "2025-01-01T00:00:00Z",
+    completed_at: "2025-01-01T00:01:00Z",
+    result: null,
+    error: null,
+    created_at: "2025-01-01T00:00:00Z",
+  };
+}
+
+function expectInvalidated(qc: QueryClient, queryKey: readonly unknown[]) {
+  expect(qc.getQueryState(queryKey)?.isInvalidated).toBe(true);
+}
+
 describe("onIssueLabelsChanged", () => {
   let qc: QueryClient;
 
@@ -93,3 +151,243 @@ describe("onIssueLabelsChanged", () => {
     expect(detail?.labels).toEqual([labelB]);
   });
 });
+
+describe("onIssueDeleted", () => {
+  let qc: QueryClient;
+
+  beforeEach(() => {
+    qc = new QueryClient();
+  });
+
+  it("removes every cache entry scoped directly to the deleted issue", () => {
+    qc.setQueryData<Issue>(issueKeys.detail(WS_ID, ISSUE_ID), baseIssue);
+    qc.setQueryData<TimelineEntry[]>(issueKeys.timeline(ISSUE_ID), [
+      {
+        type: "activity",
+        id: "activity-1",
+        actor_type: "member",
+        actor_id: "user-1",
+        action: "created",
+        created_at: "2025-01-01T00:00:00Z",
+      },
+    ]);
+    qc.setQueryData<IssueReaction[]>(issueKeys.reactions(ISSUE_ID), [
+      {
+        id: "reaction-1",
+        issue_id: ISSUE_ID,
+        actor_type: "member",
+        actor_id: "user-1",
+        emoji: "+1",
+        created_at: "2025-01-01T00:00:00Z",
+      },
+    ]);
+    qc.setQueryData<IssueSubscriber[]>(issueKeys.subscribers(ISSUE_ID), [
+      {
+        issue_id: ISSUE_ID,
+        user_type: "member",
+        user_id: "user-1",
+        reason: "manual",
+        created_at: "2025-01-01T00:00:00Z",
+      },
+    ]);
+    qc.setQueryData<IssueUsageSummary>(issueKeys.usage(ISSUE_ID), {
+      total_input_tokens: 10,
+      total_output_tokens: 20,
+      total_cache_read_tokens: 0,
+      total_cache_write_tokens: 0,
+      task_count: 1,
+    });
+    qc.setQueryData<Attachment[]>(issueKeys.attachments(ISSUE_ID), [
+      {
+        id: "attachment-1",
+        workspace_id: WS_ID,
+        issue_id: ISSUE_ID,
+        comment_id: null,
+        chat_session_id: null,
+        chat_message_id: null,
+        uploader_type: "member",
+        uploader_id: "user-1",
+        filename: "evidence.png",
+        url: "s3://bucket/evidence.png",
+        download_url: "https://example.test/evidence.png",
+        content_type: "image/png",
+        size_bytes: 1,
+        created_at: "2025-01-01T00:00:00Z",
+      },
+    ]);
+    qc.setQueryData<AgentTask[]>(issueKeys.tasks(ISSUE_ID), [makeTask()]);
+    qc.setQueryData<Issue[]>(issueKeys.children(WS_ID, ISSUE_ID), [otherIssue]);
+    qc.setQueryData<IssueLabelsResponse>(labelKeys.byIssue(WS_ID, ISSUE_ID), {
+      labels: [labelA],
+    });
+
+    qc.setQueryData<Issue>(issueKeys.detail(WS_ID, OTHER_ISSUE_ID), otherIssue);
+    qc.setQueryData<TimelineEntry[]>(issueKeys.timeline(OTHER_ISSUE_ID), []);
+    qc.setQueryData<IssueLabelsResponse>(
+      labelKeys.byIssue(WS_ID, OTHER_ISSUE_ID),
+      { labels: [labelB] },
+    );
+
+    onIssueDeleted(qc, WS_ID, ISSUE_ID);
+
+    expect(qc.getQueryData(issueKeys.detail(WS_ID, ISSUE_ID))).toBeUndefined();
+    expect(qc.getQueryData(issueKeys.timeline(ISSUE_ID))).toBeUndefined();
+    expect(qc.getQueryData(issueKeys.reactions(ISSUE_ID))).toBeUndefined();
+    expect(qc.getQueryData(issueKeys.subscribers(ISSUE_ID))).toBeUndefined();
+    expect(qc.getQueryData(issueKeys.usage(ISSUE_ID))).toBeUndefined();
+    expect(qc.getQueryData(issueKeys.attachments(ISSUE_ID))).toBeUndefined();
+    expect(qc.getQueryData(issueKeys.tasks(ISSUE_ID))).toBeUndefined();
+    expect(qc.getQueryData(issueKeys.children(WS_ID, ISSUE_ID))).toBeUndefined();
+    expect(qc.getQueryData(labelKeys.byIssue(WS_ID, ISSUE_ID))).toBeUndefined();
+
+    expect(qc.getQueryData(issueKeys.detail(WS_ID, OTHER_ISSUE_ID))).toEqual(
+      otherIssue,
+    );
+    expect(qc.getQueryData(issueKeys.timeline(OTHER_ISSUE_ID))).toEqual([]);
+    expect(qc.getQueryData(labelKeys.byIssue(WS_ID, OTHER_ISSUE_ID))).toEqual({
+      labels: [labelB],
+    });
+  });
+
+  it("removes the deleted issue from workspace and my-issues list caches immediately", () => {
+    const myFilter = { assignee_id: AGENT_ID };
+    qc.setQueryData<ListIssuesCache>(
+      issueKeys.list(WS_ID),
+      makeListCache(baseIssue, otherIssue),
+    );
+    qc.setQueryData<ListIssuesCache>(
+      issueKeys.myList(WS_ID, "assigned", myFilter),
+      makeListCache(baseIssue, otherIssue),
+    );
+
+    onIssueDeleted(qc, WS_ID, ISSUE_ID);
+
+    const list = qc.getQueryData<ListIssuesCache>(issueKeys.list(WS_ID));
+    const myList = qc.getQueryData<ListIssuesCache>(
+      issueKeys.myList(WS_ID, "assigned", myFilter),
+    );
+    expect(list?.byStatus.todo?.issues.map((i) => i.id)).toEqual([
+      OTHER_ISSUE_ID,
+    ]);
+    expect(list?.byStatus.todo?.total).toBe(1);
+    expect(myList?.byStatus.todo?.issues.map((i) => i.id)).toEqual([
+      OTHER_ISSUE_ID,
+    ]);
+    expect(myList?.byStatus.todo?.total).toBe(1);
+    expectInvalidated(qc, issueKeys.list(WS_ID));
+    expectInvalidated(qc, issueKeys.myList(WS_ID, "assigned", myFilter));
+  });
+
+  it("invalidates parent progress when the parent id only exists in detail cache", () => {
+    qc.setQueryData<Issue>(
+      issueKeys.detail(WS_ID, ISSUE_ID),
+      parentedIssue,
+    );
+    qc.setQueryData<Issue[]>(issueKeys.children(WS_ID, PARENT_ISSUE_ID), [
+      parentedIssue,
+      otherIssue,
+    ]);
+    qc.setQueryData(issueKeys.childProgress(WS_ID), new Map());
+
+    onIssueDeleted(qc, WS_ID, ISSUE_ID);
+
+    const parentChildren = qc.getQueryData<Issue[]>(
+      issueKeys.children(WS_ID, PARENT_ISSUE_ID),
+    );
+    expect(parentChildren?.map((i) => i.id)).toEqual([OTHER_ISSUE_ID]);
+    expectInvalidated(qc, issueKeys.children(WS_ID, PARENT_ISSUE_ID));
+    expectInvalidated(qc, issueKeys.childProgress(WS_ID));
+  });
+
+  it("invalidates parent progress when the deleted issue is only present in a children cache", () => {
+    qc.setQueryData<Issue[]>(issueKeys.children(WS_ID, PARENT_ISSUE_ID), [
+      parentedIssue,
+      otherIssue,
+    ]);
+    qc.setQueryData(issueKeys.childProgress(WS_ID), new Map());
+
+    onIssueDeleted(qc, WS_ID, ISSUE_ID);
+
+    const parentChildren = qc.getQueryData<Issue[]>(
+      issueKeys.children(WS_ID, PARENT_ISSUE_ID),
+    );
+    expect(parentChildren?.map((i) => i.id)).toEqual([OTHER_ISSUE_ID]);
+    expectInvalidated(qc, issueKeys.children(WS_ID, PARENT_ISSUE_ID));
+    expectInvalidated(qc, issueKeys.childProgress(WS_ID));
+  });
+
+  it("invalidates parent progress when the parent id only exists in a my-issues cache", () => {
+    const myFilter = { assignee_id: AGENT_ID };
+    qc.setQueryData<ListIssuesCache>(
+      issueKeys.myList(WS_ID, "assigned", myFilter),
+      makeListCache(parentedIssue, otherIssue),
+    );
+    qc.setQueryData<Issue[]>(issueKeys.children(WS_ID, PARENT_ISSUE_ID), [
+      otherIssue,
+    ]);
+    qc.setQueryData(issueKeys.childProgress(WS_ID), new Map());
+
+    onIssueDeleted(qc, WS_ID, ISSUE_ID);
+
+    const myList = qc.getQueryData<ListIssuesCache>(
+      issueKeys.myList(WS_ID, "assigned", myFilter),
+    );
+    expect(myList?.byStatus.todo?.issues.map((i) => i.id)).toEqual([
+      OTHER_ISSUE_ID,
+    ]);
+    expectInvalidated(qc, issueKeys.children(WS_ID, PARENT_ISSUE_ID));
+    expectInvalidated(qc, issueKeys.childProgress(WS_ID));
+  });
+
+  it("invalidates child progress when the deleted issue is itself a parent", () => {
+    qc.setQueryData<Issue>(issueKeys.detail(WS_ID, ISSUE_ID), baseIssue);
+    qc.setQueryData<Issue[]>(issueKeys.children(WS_ID, ISSUE_ID), [
+      {
+        ...otherIssue,
+        parent_issue_id: ISSUE_ID,
+      },
+    ]);
+    qc.setQueryData(
+      issueKeys.childProgress(WS_ID),
+      new Map([[ISSUE_ID, { done: 0, total: 1 }]]),
+    );
+
+    onIssueDeleted(qc, WS_ID, ISSUE_ID);
+
+    expect(qc.getQueryData(issueKeys.children(WS_ID, ISSUE_ID))).toBeUndefined();
+    expectInvalidated(qc, issueKeys.childProgress(WS_ID));
+  });
+
+  it("invalidates agent task and activity caches that can reference the deleted issue", () => {
+    qc.setQueryData<AgentTask[]>(
+      agentTaskSnapshotKeys.list(WS_ID),
+      [makeTask()],
+    );
+    qc.setQueryData<AgentActivityBucket[]>(
+      agentActivityKeys.last30d(WS_ID),
+      [
+        {
+          agent_id: AGENT_ID,
+          bucket_at: "2025-01-01T00:00:00Z",
+          task_count: 1,
+          failed_count: 0,
+        },
+      ],
+    );
+    qc.setQueryData<AgentRunCount[]>(agentRunCountsKeys.last30d(WS_ID), [
+      { agent_id: AGENT_ID, run_count: 1 },
+    ]);
+    qc.setQueryData<AgentTask[]>(agentTasksKeys.detail(WS_ID, AGENT_ID), [
+      makeTask(),
+    ]);
+    qc.setQueryData<AgentTask[]>(issueKeys.tasks(ISSUE_ID), [makeTask()]);
+
+    onIssueDeleted(qc, WS_ID, ISSUE_ID);
+
+    expectInvalidated(qc, agentTaskSnapshotKeys.list(WS_ID));
+    expectInvalidated(qc, agentActivityKeys.last30d(WS_ID));
+    expectInvalidated(qc, agentRunCountsKeys.last30d(WS_ID));
+    expectInvalidated(qc, agentTasksKeys.detail(WS_ID, AGENT_ID));
+    expect(qc.getQueryData(issueKeys.tasks(ISSUE_ID))).toBeUndefined();
+  });
+});

packages/core/issues/ws-updaters.ts

@@ -5,8 +5,8 @@ import {
   addIssueToBuckets,
   findIssueLocation,
   patchIssueInBuckets,
-  removeIssueFromBuckets,
 } from "./cache-helpers";
+import { cleanupDeletedIssueCaches } from "./delete-cache";
 import type { Issue, IssueLabelsResponse, Label } from "../types";
 import type { ListIssuesCache } from "../types";
 
@@ -107,21 +107,5 @@ export function onIssueDeleted(
   wsId: string,
   issueId: string,
 ) {
-  // Look up the issue before removing it to check for parent_issue_id
-  const listData = qc.getQueryData<ListIssuesCache>(issueKeys.list(wsId));
-  const deleted = listData ? findIssueLocation(listData, issueId)?.issue : undefined;
-
-  qc.setQueryData<ListIssuesCache>(issueKeys.list(wsId), (old) =>
-    old ? removeIssueFromBuckets(old, issueId) : old,
-  );
-  qc.invalidateQueries({ queryKey: issueKeys.myAll(wsId) });
-  qc.removeQueries({ queryKey: issueKeys.detail(wsId, issueId) });
-  qc.removeQueries({ queryKey: issueKeys.timeline(issueId) });
-  qc.removeQueries({ queryKey: issueKeys.reactions(issueId) });
-  qc.removeQueries({ queryKey: issueKeys.subscribers(issueId) });
-  qc.removeQueries({ queryKey: issueKeys.children(wsId, issueId) });
-  if (deleted?.parent_issue_id) {
-    qc.invalidateQueries({ queryKey: issueKeys.children(wsId, deleted.parent_issue_id) });
-    qc.invalidateQueries({ queryKey: issueKeys.childProgress(wsId) });
-  }
+  cleanupDeletedIssueCaches(qc, wsId, issueId);
 }

packages/core/modals/store.ts

@@ -7,6 +7,7 @@ type ModalType =
   | "create-issue"
   | "quick-create-issue"
   | "create-project"
+  | "create-squad"
   | "feedback"
   | "issue-set-parent"
   | "issue-add-child"

packages/core/onboarding/index.ts

@@ -13,3 +13,8 @@ export {
 } from "./store";
 export { ONBOARDING_STEP_ORDER } from "./step-order";
 export { recommendTemplate, type AgentTemplateId } from "./recommend-template";
+export {
+  startOnboardingSession,
+  getOnboardingSessionId,
+  clearOnboardingSession,
+} from "./session";

packages/core/onboarding/session.test.ts

@@ -0,0 +1,59 @@
+import { afterEach, beforeEach, describe, expect, it } from "vitest";
+
+async function loadModule() {
+  // Reset module cache so each test starts with a clean in-memory id.
+  const vitest = await import("vitest");
+  vitest.vi.resetModules();
+  return import("./session");
+}
+
+beforeEach(() => {
+  // Provide a minimal in-memory localStorage so defaultStorage's
+  // typeof-window check passes and we exercise the persistence path.
+  const store = new Map<string, string>();
+  Object.defineProperty(globalThis, "window", {
+    value: {},
+    configurable: true,
+    writable: true,
+  });
+  Object.defineProperty(globalThis, "localStorage", {
+    value: {
+      getItem: (k: string) => (store.has(k) ? store.get(k)! : null),
+      setItem: (k: string, v: string) => store.set(k, v),
+      removeItem: (k: string) => store.delete(k),
+    },
+    configurable: true,
+    writable: true,
+  });
+});
+
+afterEach(() => {
+  // @ts-expect-error — test-only cleanup
+  delete globalThis.window;
+  // @ts-expect-error — test-only cleanup
+  delete globalThis.localStorage;
+});
+
+describe("onboarding session", () => {
+  it("startOnboardingSession returns a stable id within the same funnel", async () => {
+    const session = await loadModule();
+    const id1 = session.startOnboardingSession();
+    const id2 = session.startOnboardingSession();
+    expect(id1).toBe(id2);
+    expect(session.getOnboardingSessionId()).toBe(id1);
+  });
+
+  it("clearOnboardingSession resets so the next start gets a fresh id", async () => {
+    const session = await loadModule();
+    const first = session.startOnboardingSession();
+    session.clearOnboardingSession();
+    expect(session.getOnboardingSessionId()).toBeNull();
+    const second = session.startOnboardingSession();
+    expect(second).not.toBe(first);
+  });
+
+  it("getOnboardingSessionId returns null when no session has been started", async () => {
+    const session = await loadModule();
+    expect(session.getOnboardingSessionId()).toBeNull();
+  });
+});

packages/core/onboarding/session.ts

@@ -0,0 +1,63 @@
+// Onboarding session identifier — issued once per funnel entry and attached
+// to every onboarding event so PostHog can correlate the full funnel back to
+// a single `onboarding_started`. Solves the prior funnel-attribution gap
+// where `onboarding_completed` events fired from skip/invite paths (no
+// `onboarding_started` in their lineage) were indistinguishable from real
+// funnel completions when joining on `distinct_id` alone.
+//
+// The id is persisted to client storage because the funnel spans page
+// reloads (especially on web — desktop bundle install, OAuth redirects).
+// It's cleared on completion; entering onboarding again starts a fresh
+// session and a fresh id.
+
+import { createSafeId } from "../utils";
+import { defaultStorage } from "../platform/storage";
+
+const STORAGE_KEY = "multica_onboarding_session_id";
+
+// In-memory cache so the analytics wrapper doesn't hit storage on every
+// event. Storage is read once on first access and on every start/clear.
+let cached: string | null | undefined;
+
+function read(): string | null {
+  if (cached !== undefined) return cached;
+  cached = defaultStorage.getItem(STORAGE_KEY);
+  return cached;
+}
+
+/**
+ * Generate a new session id and persist it. Idempotent — calling twice in
+ * the same funnel returns the same id, so a re-mount of the onboarding
+ * shell can't accidentally split one funnel across two sessions.
+ *
+ * The expected fire site is the same place that emits `onboarding_started`.
+ */
+export function startOnboardingSession(): string {
+  const existing = read();
+  if (existing) return existing;
+  const id = createSafeId();
+  cached = id;
+  defaultStorage.setItem(STORAGE_KEY, id);
+  return id;
+}
+
+/**
+ * Read the current session id. Returns null when no onboarding session is
+ * in progress — the analytics wrapper omits the property in that case
+ * rather than emitting an empty string, so HogQL queries can filter
+ * `onboarding_session_id IS NOT NULL` to isolate real funnel events from
+ * skip/invite paths that legitimately have no session.
+ */
+export function getOnboardingSessionId(): string | null {
+  return read();
+}
+
+/**
+ * Clear the session. Called at the funnel terminus (after a successful
+ * `onboarding_completed`) so a returning user who somehow re-enters
+ * onboarding starts a fresh session.
+ */
+export function clearOnboardingSession(): void {
+  cached = null;
+  defaultStorage.removeItem(STORAGE_KEY);
+}

packages/core/onboarding/store.ts

@@ -1,6 +1,10 @@
 import { api } from "../api";
 import { useAuthStore } from "../auth";
 import { setPersonProperties } from "../analytics";
+import {
+  clearOnboardingSession,
+  getOnboardingSessionId,
+} from "./session";
 import type { OnboardingCompletionPath, QuestionnaireAnswers } from "./types";
 
 /**
@@ -16,7 +20,11 @@ import type { OnboardingCompletionPath, QuestionnaireAnswers } from "./types";
 export async function saveQuestionnaire(
   answers: Partial<QuestionnaireAnswers>,
 ): Promise<void> {
-  const user = await api.patchOnboarding({ questionnaire: answers });
+  const onboardingSessionId = getOnboardingSessionId() ?? undefined;
+  const user = await api.patchOnboarding({
+    questionnaire: answers,
+    onboarding_session_id: onboardingSessionId,
+  });
   useAuthStore.getState().setUser(user);
   // Mirror the three cohort signals into person properties so every
   // PostHog event on this user can be broken down by role / use_case /
@@ -44,11 +52,20 @@ export async function completeOnboarding(
   completionPath?: OnboardingCompletionPath,
   workspaceId?: string,
 ): Promise<void> {
-  await api.markOnboardingComplete(
-    completionPath || workspaceId
-      ? { completion_path: completionPath, workspace_id: workspaceId }
-      : undefined,
-  );
+  const onboardingSessionId = getOnboardingSessionId() ?? undefined;
+  const payload =
+    completionPath || workspaceId || onboardingSessionId
+      ? {
+          completion_path: completionPath,
+          workspace_id: workspaceId,
+          onboarding_session_id: onboardingSessionId,
+        }
+      : undefined;
+  await api.markOnboardingComplete(payload);
+  // Clear the session AFTER the server records it. The funnel terminus
+  // is over — any subsequent re-entry into onboarding starts a fresh
+  // session (and a fresh id), which is what we want.
+  clearOnboardingSession();
   await useAuthStore.getState().refreshMe();
 }
 

packages/core/package.json

@@ -47,6 +47,8 @@
     "./runtimes/mutations": "./runtimes/mutations.ts",
     "./runtimes/hooks": "./runtimes/hooks.ts",
     "./runtimes/custom-pricing-store": "./runtimes/custom-pricing-store.ts",
+    "./dashboard": "./dashboard/index.ts",
+    "./dashboard/queries": "./dashboard/queries.ts",
     "./agents": "./agents/index.ts",
     "./agents/queries": "./agents/queries.ts",
     "./agents/derive-presence": "./agents/derive-presence.ts",

packages/core/paths/consistency.test.ts

@@ -17,14 +17,17 @@ describe("paths.workspace() shape", () => {
     expect(new Set(parameterlessRoutes)).toEqual(
       new Set([
         "root",
+        "usage",
         "issues",
         "projects",
         "autopilots",
         "agents",
+        "squads",
         "inbox",
         "myIssues",
         "runtimes",
         "skills",
+        "squads",
         "settings",
       ]),
     );
@@ -35,14 +38,17 @@ describe("paths.workspace() shape", () => {
     // Check that none of the parameterless paths embed a leaked literal
     // and that their second URL segment matches the method name's kebab-case.
     const expectedSegments: Array<[string, string]> = [
+      ["usage", "usage"],
       ["issues", "issues"],
       ["projects", "projects"],
       ["autopilots", "autopilots"],
       ["agents", "agents"],
+      ["squads", "squads"],
       ["inbox", "inbox"],
       ["myIssues", "my-issues"],
       ["runtimes", "runtimes"],
       ["skills", "skills"],
+      ["squads", "squads"],
       ["settings", "settings"],
     ];
     const wsAsAny = ws as unknown as Record<string, () => string>;

packages/core/paths/paths.test.ts

@@ -4,7 +4,8 @@ import { paths, isGlobalPath } from "./paths";
 describe("paths.workspace(slug)", () => {
   const ws = paths.workspace("acme");
 
-  it("builds dashboard paths with slug prefix", () => {
+  it("builds workspace paths with slug prefix", () => {
+    expect(ws.usage()).toBe("/acme/usage");
     expect(ws.issues()).toBe("/acme/issues");
     expect(ws.issueDetail("abc-123")).toBe("/acme/issues/abc-123");
     expect(ws.projects()).toBe("/acme/projects");
@@ -17,6 +18,8 @@ describe("paths.workspace(slug)", () => {
     expect(ws.runtimes()).toBe("/acme/runtimes");
     expect(ws.skills()).toBe("/acme/skills");
     expect(ws.skillDetail("skl_123")).toBe("/acme/skills/skl_123");
+    expect(ws.squads()).toBe("/acme/squads");
+    expect(ws.squadDetail("sq_1")).toBe("/acme/squads/sq_1");
     expect(ws.settings()).toBe("/acme/settings");
   });
 

packages/core/paths/paths.ts

@@ -18,6 +18,7 @@ function workspaceScoped(slug: string) {
   const ws = `/${encode(slug)}`;
   return {
     root: () => `${ws}/issues`,
+    usage: () => `${ws}/usage`,
     issues: () => `${ws}/issues`,
     issueDetail: (id: string) => `${ws}/issues/${encode(id)}`,
     projects: () => `${ws}/projects`,
@@ -26,6 +27,8 @@ function workspaceScoped(slug: string) {
     autopilotDetail: (id: string) => `${ws}/autopilots/${encode(id)}`,
     agents: () => `${ws}/agents`,
     agentDetail: (id: string) => `${ws}/agents/${encode(id)}`,
+    squads: () => `${ws}/squads`,
+    squadDetail: (id: string) => `${ws}/squads/${encode(id)}`,
     inbox: () => `${ws}/inbox`,
     myIssues: () => `${ws}/my-issues`,
     runtimes: () => `${ws}/runtimes`,

packages/core/paths/reserved-slugs.ts

@@ -70,7 +70,7 @@ export const RESERVED_SLUGS: ReadonlySet<string> = new Set([
   "search",
   "members",
 
-  // Dashboard / workspace route segments
+  // Workspace route segments
   // Reserving each segment name prevents `/{slug}/{view}` from being visually
   // ambiguous (e.g. a workspace named `issues` would make `/issues/abc` mean two
   // things). `workspaces` covers the global `/workspaces/new` workspace-creation
@@ -79,8 +79,10 @@ export const RESERVED_SLUGS: ReadonlySet<string> = new Set([
   "projects",
   "autopilots",
   "agents",
+  "squads",
   "inbox",
   "my-issues",
+  "usage",
   "runtimes",
   "skills",
   "settings",

packages/core/realtime/use-realtime-sync.test.ts

@@ -0,0 +1,117 @@
+import { QueryClient } from "@tanstack/react-query";
+import { describe, expect, it, vi } from "vitest";
+import { chatKeys } from "../chat/queries";
+import type { ChatDonePayload, ChatMessage, ChatPendingTask } from "../types";
+import { applyChatDoneToCache } from "./use-realtime-sync";
+
+const sessionId = "session-1";
+const taskId = "task-1";
+const messagesKey = chatKeys.messages(sessionId);
+const pendingKey = chatKeys.pendingTask(sessionId);
+
+function createQueryClient() {
+  return new QueryClient({
+    defaultOptions: {
+      queries: { retry: false },
+    },
+  });
+}
+
+function userMessage(): ChatMessage {
+  return {
+    id: "msg-user",
+    chat_session_id: sessionId,
+    role: "user",
+    content: "hello",
+    task_id: null,
+    created_at: "2026-05-13T05:00:00Z",
+  };
+}
+
+function donePayload(overrides: Partial<ChatDonePayload> = {}): ChatDonePayload {
+  return {
+    chat_session_id: sessionId,
+    task_id: taskId,
+    message_id: "msg-assistant",
+    content: "done",
+    elapsed_ms: 1234,
+    created_at: "2026-05-13T05:00:02Z",
+    ...overrides,
+  };
+}
+
+describe("applyChatDoneToCache", () => {
+  it("writes the assistant message before clearing pending task", () => {
+    const qc = createQueryClient();
+    qc.setQueryData<ChatMessage[]>(messagesKey, [userMessage()]);
+    qc.setQueryData<ChatPendingTask>(pendingKey, {
+      task_id: taskId,
+      status: "running",
+    });
+
+    const setQueryData = vi.spyOn(qc, "setQueryData");
+
+    applyChatDoneToCache(qc, donePayload());
+
+    expect(setQueryData.mock.calls[0]?.[0]).toEqual(messagesKey);
+    expect(setQueryData.mock.calls[1]?.[0]).toEqual(pendingKey);
+    expect(qc.getQueryData<ChatPendingTask>(pendingKey)).toEqual({});
+    expect(qc.getQueryData<ChatMessage[]>(messagesKey)).toEqual([
+      userMessage(),
+      {
+        id: "msg-assistant",
+        chat_session_id: sessionId,
+        role: "assistant",
+        content: "done",
+        task_id: taskId,
+        created_at: "2026-05-13T05:00:02Z",
+        elapsed_ms: 1234,
+      },
+    ]);
+  });
+
+  it("does not duplicate a replayed chat done event", () => {
+    const qc = createQueryClient();
+    const assistant: ChatMessage = {
+      id: "msg-assistant",
+      chat_session_id: sessionId,
+      role: "assistant",
+      content: "done",
+      task_id: taskId,
+      created_at: "2026-05-13T05:00:02Z",
+      elapsed_ms: 1234,
+    };
+    qc.setQueryData<ChatMessage[]>(messagesKey, [userMessage(), assistant]);
+    qc.setQueryData<ChatPendingTask>(pendingKey, {
+      task_id: taskId,
+      status: "running",
+    });
+
+    applyChatDoneToCache(qc, donePayload());
+
+    expect(qc.getQueryData<ChatMessage[]>(messagesKey)).toEqual([
+      userMessage(),
+      assistant,
+    ]);
+    expect(qc.getQueryData<ChatPendingTask>(pendingKey)).toEqual({});
+  });
+
+  it("falls back to invalidation-only when older servers omit message fields", () => {
+    const qc = createQueryClient();
+    qc.setQueryData<ChatMessage[]>(messagesKey, [userMessage()]);
+    qc.setQueryData<ChatPendingTask>(pendingKey, {
+      task_id: taskId,
+      status: "running",
+    });
+
+    applyChatDoneToCache(
+      qc,
+      donePayload({ message_id: undefined, content: undefined }),
+    );
+
+    expect(qc.getQueryData<ChatMessage[]>(messagesKey)).toEqual([
+      userMessage(),
+    ]);
+    expect(qc.getQueryData<ChatPendingTask>(pendingKey)).toEqual({});
+  });
+});

packages/core/realtime/use-realtime-sync.ts

@@ -1,7 +1,7 @@
 "use client";
 
 import { useEffect, useRef } from "react";
-import { useQueryClient } from "@tanstack/react-query";
+import { useQueryClient, type QueryClient } from "@tanstack/react-query";
 import type { WSClient } from "../api/ws-client";
 import type { StoreApi, UseBoundStore } from "zustand";
 import type { AuthState } from "../auth/store";
@@ -62,6 +62,7 @@ import type {
   TaskFailedPayload,
   TaskCancelledPayload,
   ChatDonePayload,
+  ChatMessage,
   ChatPendingTask,
   InvitationCreatedPayload,
 } from "../types";
@@ -70,6 +71,42 @@ const chatWsLogger = createLogger("chat.ws");
 
 const logger = createLogger("realtime-sync");
 
+export function applyChatDoneToCache(
+  qc: QueryClient,
+  payload: ChatDonePayload,
+) {
+  const sessionId = payload.chat_session_id;
+  const taskId = payload.task_id;
+  const messageId = payload.message_id;
+  const content = payload.content;
+  if (messageId && content !== undefined) {
+    qc.setQueryData<ChatMessage[] | undefined>(
+      chatKeys.messages(sessionId),
+      (old) => {
+        if (!old) return old; // first fetch will pick it up
+        // Idempotent against reconnect replay.
+        if (old.some((m) => m.id === messageId)) return old;
+        const assistant: ChatMessage = {
+          id: messageId,
+          chat_session_id: sessionId,
+          role: "assistant",
+          content,
+          task_id: taskId,
+          created_at: payload.created_at ?? new Date().toISOString(),
+          elapsed_ms: payload.elapsed_ms ?? null,
+        };
+        return [...old, assistant];
+      },
+    );
+  }
+  // Replacement is in the messages list now; safe to drop pending.
+  qc.setQueryData(chatKeys.pendingTask(sessionId), {});
+  // Authoritative refetch reconciles redaction / migrations / clients
+  // that took the fallback branch above.
+  qc.invalidateQueries({ queryKey: chatKeys.messages(sessionId) });
+  qc.invalidateQueries({ queryKey: chatKeys.pendingTask(sessionId) });
+}
+
 export interface RealtimeSyncStores {
   authStore: UseBoundStore<StoreApi<AuthState>>;
 }
@@ -134,6 +171,14 @@ export function useRealtimeSync(
         const wsId = getCurrentWsId();
         if (wsId) qc.invalidateQueries({ queryKey: projectKeys.all(wsId) });
       },
+      squad: () => {
+        const wsId = getCurrentWsId();
+        if (wsId) {
+          qc.invalidateQueries({ queryKey: workspaceKeys.squads(wsId) });
+          // squad:deleted triggers assignee transfer — refresh issues too.
+          qc.invalidateQueries({ queryKey: issueKeys.all(wsId) });
+        }
+      },
       label: () => {
         // label:created/updated/deleted — also refresh issues, since each
         // issue carries a denormalized snapshot of its labels (rename/recolor
@@ -222,6 +267,7 @@ export function useRealtimeSync(
       "daemon:heartbeat",
       // Chat events are handled explicitly below; do not double-invalidate.
       "chat:message", "chat:done", "chat:session_read", "chat:session_deleted",
+      "chat:session_updated",
       // task:message stays out of the prefix path because it fires per
       // streamed message during a long run — invalidating the snapshot on
       // every message would flood the network. Specific chat handlers below
@@ -568,13 +614,21 @@ export function useRealtimeSync(
       chatWsLogger.info("chat:done (global)", {
         task_id: payload.task_id,
         chat_session_id: payload.chat_session_id,
+        has_message: !!payload.message_id,
       });
-      // Assistant message was just written and task flipped out of 'running'.
-      // Clear pending-task cache immediately so the live-timeline-vs-assistant
-      // race window collapses to zero — the subsequent refetch will confirm.
-      qc.setQueryData(chatKeys.pendingTask(payload.chat_session_id), {});
-      qc.invalidateQueries({ queryKey: chatKeys.messages(payload.chat_session_id) });
-      qc.invalidateQueries({ queryKey: chatKeys.pendingTask(payload.chat_session_id) });
+      // Inline-insert the assistant message into the messages cache BEFORE
+      // clearing pending-task. Both writes land in the same React render
+      // tick, so ChatMessageList sees `pendingAlreadyPersisted === true`
+      // and the live TimelineView unmounts only after AssistantMessage has
+      // mounted — no flicker window. This applies TkDodo's "combine
+      // setQueryData (active query) + invalidateQueries (others)" pattern
+      // (https://tkdodo.eu/blog/using-web-sockets-with-react-query).
+      //
+      // Falls back to invalidate-only when the server omits the message
+      // payload (older builds). Older clients hitting a newer server also
+      // work: they ignore the extra fields and rely on the invalidate
+      // below, which keeps the old behavior alive.
+      applyChatDoneToCache(qc, payload);
       invalidatePendingAggregate();
       // Assistant message just landed → has_unread may have flipped to true.
       invalidateSessionLists();
@@ -645,9 +699,12 @@ export function useRealtimeSync(
         task_id: payload.task_id,
         chat_session_id: payload.chat_session_id,
       });
-      qc.setQueryData(chatKeys.pendingTask(payload.chat_session_id), {});
-      qc.invalidateQueries({ queryKey: chatKeys.messages(payload.chat_session_id) });
-      qc.invalidateQueries({ queryKey: chatKeys.pendingTask(payload.chat_session_id) });
+      // `chat:done` (broadcast immediately before this event in CompleteTask)
+      // already wrote the assistant message into the messages cache and
+      // cleared `chatKeys.pendingTask`. This event is now only responsible
+      // for refreshing the per-user cross-session aggregate that drives the
+      // FAB indicator — `chat:done` is per-session and doesn't carry that
+      // information.
       invalidatePendingAggregate();
     });
 
@@ -676,6 +733,33 @@ export function useRealtimeSync(
       invalidateSessionLists();
     });
 
+    // chat:session_updated fires after the creator renames a session in
+    // any tab/device. Patch the cached row inline so the dropdown reflects
+    // the new title without a full sessions-list refetch.
+    const unsubChatSessionUpdated = ws.on("chat:session_updated", (p) => {
+      const payload = p as {
+        chat_session_id: string;
+        title?: string;
+        updated_at?: string;
+      };
+      chatWsLogger.info("chat:session_updated (global)", payload);
+      const id = getCurrentWsId();
+      if (!id) return;
+      const patch = (
+        old?: { id: string; title: string; updated_at: string }[],
+      ) =>
+        old?.map((s) =>
+          s.id === payload.chat_session_id
+            ? {
+                ...s,
+                title: payload.title ?? s.title,
+                updated_at: payload.updated_at ?? s.updated_at,
+              }
+            : s,
+        );
+      qc.setQueryData(chatKeys.sessions(id), patch);
+    });
+
     // chat:session_deleted fires after a hard delete. The originating tab has
     // already optimistically dropped the row via useDeleteChatSession; this
     // handler keeps OTHER tabs/devices in sync and also clears the active
@@ -736,6 +820,7 @@ export function useRealtimeSync(
       unsubTaskFailed();
       unsubChatSessionRead();
       unsubChatSessionDeleted();
+      unsubChatSessionUpdated();
       timers.forEach(clearTimeout);
       timers.clear();
     };

packages/core/types/agent.ts

@@ -168,6 +168,76 @@ export interface CreateAgentRequest {
   template?: string;
 }
 
+/** Agent template summary — fields needed by the picker grid. Does NOT
+ *  include `instructions` to keep the list payload small; the detail
+ *  endpoint or the create flow returns the full template body. */
+export interface AgentTemplateSummary {
+  slug: string;
+  name: string;
+  description: string;
+  /** Optional grouping for the picker UI ("Engineering" / "Writing" / …). */
+  category?: string;
+  /** Optional lucide-react icon name (e.g. "Search"). Frontend falls back
+   *  to a generic icon when empty. */
+  icon?: string;
+  /** Optional semantic color token for the icon badge — one of "info" /
+   *  "success" / "warning" / "primary" / "secondary". Frontend has a
+   *  static class map so Tailwind can JIT-scan all variants. */
+  accent?: string;
+  skills: AgentTemplateSkillRef[];
+}
+
+/** Full agent template — same as `AgentTemplateSummary` plus the
+ *  instructions block. Returned by `GET /api/agent-templates/:slug`. */
+export interface AgentTemplate extends AgentTemplateSummary {
+  instructions: string;
+}
+
+/** Skill reference inside an agent template. `source_url` is the upstream
+ *  GitHub / skills.sh URL fetched on create; `cached_*` mirror the upstream
+ *  frontmatter at template-author time and let the picker render without
+ *  HTTP fetches. */
+export interface AgentTemplateSkillRef {
+  source_url: string;
+  cached_name: string;
+  cached_description: string;
+}
+
+export interface CreateAgentFromTemplateRequest {
+  template_slug: string;
+  name: string;
+  runtime_id: string;
+  model?: string;
+  visibility?: AgentVisibility;
+  max_concurrent_tasks?: number;
+  /** Optional overrides applied to the template before creation. nil/omit
+   *  uses the template's own value. */
+  description?: string;
+  instructions?: string;
+  avatar_url?: string;
+  /** Workspace skill IDs attached **in addition to** the template's
+   *  skills. Server dedupes against template skills automatically. */
+  extra_skill_ids?: string[];
+}
+
+export interface CreateAgentFromTemplateResponse {
+  agent: Agent;
+  /** Skill IDs that were newly created in the workspace from upstream URLs. */
+  imported_skill_ids: string[];
+  /** Skill IDs that already existed in the workspace (same name) and were
+   *  reused rather than re-imported. The UI can surface this as a toast so
+   *  the user knows their pre-existing skill wasn't overwritten. */
+  reused_skill_ids: string[];
+}
+
+/** 422 body returned by `POST /api/agents/from-template` when one or more
+ *  template skill URLs cannot be reached. The transaction is rolled back —
+ *  no partial workspace state. */
+export interface CreateAgentFromTemplateFailure {
+  error: string;
+  failed_urls: string[];
+}
+
 export interface UpdateAgentRequest {
   name?: string;
   description?: string;
@@ -289,6 +359,44 @@ export interface RuntimeUsageByHour {
   task_count: number;
 }
 
+// One (date, model) bucket of token usage for the workspace dashboard.
+// Same shape as RuntimeUsage but workspace-scoped (no runtime_id, no
+// provider field on the wire) and optionally narrowed to a single project
+// on the server side. Cost stays client-side via the model pricing table.
+export interface DashboardUsageDaily {
+  date: string;
+  model: string;
+  input_tokens: number;
+  output_tokens: number;
+  cache_read_tokens: number;
+  cache_write_tokens: number;
+  task_count: number;
+}
+
+// Per-(agent, model) token totals for the workspace dashboard. Identical
+// wire shape to RuntimeUsageByAgent — the client folds by agent_id and
+// sums cost.
+export interface DashboardUsageByAgent {
+  agent_id: string;
+  model: string;
+  input_tokens: number;
+  output_tokens: number;
+  cache_read_tokens: number;
+  cache_write_tokens: number;
+  task_count: number;
+}
+
+// Per-agent total terminal-task run-time + counts. Powers the workspace
+// dashboard's "time by agent" list. failed_count is a subset of
+// task_count (failed tasks still contribute to total_seconds because
+// they consumed runtime to fail).
+export interface DashboardAgentRunTime {
+  agent_id: string;
+  total_seconds: number;
+  task_count: number;
+  failed_count: number;
+}
+
 export type RuntimeUpdateStatus =
   | "pending"
   | "running"

packages/core/types/events.ts

@@ -54,9 +54,13 @@ export type WSEventType =
   | "chat:done"
   | "chat:session_read"
   | "chat:session_deleted"
+  | "chat:session_updated"
   | "project:created"
   | "project:updated"
   | "project:deleted"
+  | "squad:created"
+  | "squad:updated"
+  | "squad:deleted"
   | "label:created"
   | "label:updated"
   | "label:deleted"
@@ -289,7 +293,16 @@ export interface ChatMessageEventPayload {
 export interface ChatDonePayload {
   chat_session_id: string;
   task_id: string;
+  /**
+   * Server populates these from the freshly-persisted assistant ChatMessage
+   * row so the WS handler can write it into the messages cache inline. Older
+   * servers (pre-#2123) only sent chat_session_id + task_id; treat every field
+   * below as optional and fall back to a refetch when absent.
+   */
+  message_id?: string;
   content?: string;
+  elapsed_ms?: number;
+  created_at?: string;
 }
 
 export interface ChatSessionReadPayload {

packages/core/types/inbox.ts

@@ -25,7 +25,7 @@ export interface InboxItem {
   workspace_id: string;
   recipient_type: "member" | "agent";
   recipient_id: string;
-  actor_type: "member" | "agent" | null;
+  actor_type: "member" | "agent" | "system" | null;
   actor_id: string | null;
   type: InboxItemType;
   severity: InboxSeverity;

packages/core/types/index.ts

@@ -11,6 +11,12 @@ export type {
   AgentRuntime,
   RuntimeDevice,
   CreateAgentRequest,
+  AgentTemplate,
+  AgentTemplateSummary,
+  AgentTemplateSkillRef,
+  CreateAgentFromTemplateRequest,
+  CreateAgentFromTemplateResponse,
+  CreateAgentFromTemplateFailure,
   UpdateAgentRequest,
   Skill,
   SkillSummary,
@@ -23,6 +29,9 @@ export type {
   RuntimeHourlyActivity,
   RuntimeUsageByAgent,
   RuntimeUsageByHour,
+  DashboardUsageDaily,
+  DashboardUsageByAgent,
+  DashboardAgentRunTime,
   RuntimeUpdate,
   RuntimeUpdateStatus,
   RuntimeModel,
@@ -91,3 +100,16 @@ export type {
   GetAutopilotResponse,
   ListAutopilotRunsResponse,
 } from "./autopilot";
+export type {
+  Squad,
+  SquadMember,
+  SquadMemberType,
+  SquadActivityLog,
+  SquadActivityOutcome,
+  CreateSquadRequest,
+  UpdateSquadRequest,
+  AddSquadMemberRequest,
+  RemoveSquadMemberRequest,
+  UpdateSquadMemberRoleRequest,
+  CreateSquadActivityLogRequest,
+} from "./squad";

packages/core/types/issue.ts

@@ -11,7 +11,7 @@ export type IssueStatus =
 
 export type IssuePriority = "urgent" | "high" | "medium" | "low" | "none";
 
-export type IssueAssigneeType = "member" | "agent";
+export type IssueAssigneeType = "member" | "agent" | "squad";
 
 export interface IssueReaction {
   id: string;

packages/core/types/squad.ts

@@ -0,0 +1,77 @@
+export type SquadMemberType = "agent" | "member";
+
+export type SquadActivityOutcome = "action" | "no_action" | "failed";
+
+export interface Squad {
+  id: string;
+  workspace_id: string;
+  name: string;
+  description: string;
+  instructions: string;
+  avatar_url: string | null;
+  leader_id: string;
+  creator_id: string;
+  created_at: string;
+  updated_at: string;
+  archived_at: string | null;
+  archived_by: string | null;
+}
+
+export interface SquadMember {
+  id: string;
+  squad_id: string;
+  member_type: SquadMemberType;
+  member_id: string;
+  role: string;
+  created_at: string;
+}
+
+export interface SquadActivityLog {
+  id: string;
+  squad_id: string;
+  issue_id: string;
+  trigger_comment_id: string | null;
+  leader_id: string;
+  outcome: SquadActivityOutcome;
+  details: unknown;
+  created_at: string;
+}
+
+export interface CreateSquadRequest {
+  name: string;
+  description?: string;
+  leader_id: string;
+}
+
+export interface UpdateSquadRequest {
+  name?: string;
+  description?: string;
+  instructions?: string;
+  leader_id?: string;
+  avatar_url?: string;
+}
+
+export interface AddSquadMemberRequest {
+  member_type: SquadMemberType;
+  member_id: string;
+  role?: string;
+}
+
+export interface RemoveSquadMemberRequest {
+  member_type: SquadMemberType;
+  member_id: string;
+}
+
+export interface UpdateSquadMemberRoleRequest {
+  member_type: SquadMemberType;
+  member_id: string;
+  role: string;
+}
+
+export interface CreateSquadActivityLogRequest {
+  squad_id: string;
+  issue_id: string;
+  trigger_comment_id?: string;
+  outcome: SquadActivityOutcome;
+  details?: unknown;
+}

packages/core/workspace/hooks.ts

@@ -2,12 +2,13 @@
 
 import { useQuery } from "@tanstack/react-query";
 import { useWorkspaceId } from "../hooks";
-import { memberListOptions, agentListOptions } from "./queries";
+import { memberListOptions, agentListOptions, squadListOptions } from "./queries";
 
 export function useActorName() {
   const wsId = useWorkspaceId();
   const { data: members = [] } = useQuery(memberListOptions(wsId));
   const { data: agents = [] } = useQuery(agentListOptions(wsId));
+  const { data: squads = [] } = useQuery(squadListOptions(wsId));
 
   const getMemberName = (userId: string) => {
     const m = members.find((m) => m.user_id === userId);
@@ -19,9 +20,16 @@ export function useActorName() {
     return a?.name ?? "Unknown Agent";
   };
 
+  const getSquadName = (squadId: string) => {
+    const s = squads.find((s) => s.id === squadId);
+    return s?.name ?? "Unknown Squad";
+  };
+
   const getActorName = (type: string, id: string) => {
     if (type === "member") return getMemberName(id);
     if (type === "agent") return getAgentName(id);
+    if (type === "squad") return getSquadName(id);
+    if (type === "system") return "Multica";
     return "System";
   };
 
@@ -38,8 +46,9 @@ export function useActorName() {
   const getActorAvatarUrl = (type: string, id: string): string | null => {
     if (type === "member") return members.find((m) => m.user_id === id)?.avatar_url ?? null;
     if (type === "agent") return agents.find((a) => a.id === id)?.avatar_url ?? null;
+    if (type === "squad") return squads.find((s) => s.id === id)?.avatar_url ?? null;
     return null;
   };
 
-  return { getMemberName, getAgentName, getActorName, getActorInitials, getActorAvatarUrl };
+  return { getMemberName, getAgentName, getSquadName, getActorName, getActorInitials, getActorAvatarUrl };
 }

packages/core/workspace/queries.ts

@@ -1,6 +1,6 @@
 import { queryOptions } from "@tanstack/react-query";
 import { api } from "../api";
-import type { Agent, Workspace } from "../types";
+import type { Agent, Squad, Workspace } from "../types";
 
 export const workspaceKeys = {
   all: (wsId: string) => ["workspaces", wsId] as const,
@@ -9,6 +9,7 @@ export const workspaceKeys = {
   invitations: (wsId: string) => ["workspaces", wsId, "invitations"] as const,
   myInvitations: () => ["invitations", "mine"] as const,
   agents: (wsId: string) => ["workspaces", wsId, "agents"] as const,
+  squads: (wsId: string) => ["workspaces", wsId, "squads"] as const,
   skills: (wsId: string) => ["workspaces", wsId, "skills"] as const,
   assigneeFrequency: (wsId: string) => ["workspaces", wsId, "assignee-frequency"] as const,
 };
@@ -43,6 +44,14 @@ export function agentListOptions(wsId: string) {
   });
 }
 
+export function squadListOptions(wsId: string) {
+  return queryOptions<Squad[]>({
+    queryKey: workspaceKeys.squads(wsId),
+    queryFn: () => api.listSquads(),
+    enabled: !!wsId,
+  });
+}
+
 export function skillListOptions(wsId: string) {
   return queryOptions({
     queryKey: workspaceKeys.skills(wsId),

packages/ui/components/common/actor-avatar.tsx

@@ -1,14 +1,17 @@
 "use client";
 
 import { useState, useEffect } from "react";
-import { Bot } from "lucide-react";
+import { Bot, Users } from "lucide-react";
 import { cn } from "@multica/ui/lib/utils";
+import { MulticaIcon } from "./multica-icon";
 
 interface ActorAvatarProps {
   name: string;
   initials: string;
   avatarUrl?: string | null;
   isAgent?: boolean;
+  isSystem?: boolean;
+  isSquad?: boolean;
   size?: number;
   className?: string;
 }
@@ -18,12 +21,13 @@ function ActorAvatar({
   initials,
   avatarUrl,
   isAgent,
+  isSystem,
+  isSquad,
   size = 20,
   className,
 }: ActorAvatarProps) {
   const [imgError, setImgError] = useState(false);
 
-  // Reset error state when URL changes (e.g. user uploads new avatar)
   useEffect(() => {
     setImgError(false);
   }, [avatarUrl]);
@@ -32,7 +36,10 @@ function ActorAvatar({
     <div
       data-slot="avatar"
       className={cn(
-        "inline-flex shrink-0 items-center justify-center rounded-full font-medium overflow-hidden",
+        "inline-flex shrink-0 items-center justify-center font-medium overflow-hidden",
+        // Squads (a group, non-human) get a square tile so they don't read as
+        // a single person; everyone else stays round.
+        isSquad ? "rounded-md" : "rounded-full",
         "bg-muted text-muted-foreground",
         className
       )}
@@ -46,8 +53,12 @@ function ActorAvatar({
           className="h-full w-full object-cover"
           onError={() => setImgError(true)}
         />
+      ) : isSystem ? (
+        <MulticaIcon noSpin style={{ width: size * 0.55, height: size * 0.55 }} />
       ) : isAgent ? (
         <Bot style={{ width: size * 0.55, height: size * 0.55 }} />
+      ) : isSquad ? (
+        <Users style={{ width: size * 0.55, height: size * 0.55 }} />
       ) : (
         initials
       )}

packages/ui/components/common/file-upload-button.tsx

@@ -2,6 +2,7 @@
 
 import { useRef } from "react";
 import { Paperclip } from "lucide-react";
+import { useTranslation } from "react-i18next";
 import { cn } from "@multica/ui/lib/utils";
 
 interface FileUploadButtonProps {
@@ -18,7 +19,9 @@ function FileUploadButton({
   className,
   size = "default",
 }: FileUploadButtonProps) {
+  const { t } = useTranslation("ui");
   const inputRef = useRef<HTMLInputElement>(null);
+  const attachLabel = t(($) => $.attach_file);
 
   const handleChange = (e: React.ChangeEvent<HTMLInputElement>) => {
     const file = e.target.files?.[0];
@@ -36,8 +39,8 @@ function FileUploadButton({
         type="button"
         onClick={() => inputRef.current?.click()}
         disabled={disabled}
-        aria-label="Attach file"
-        title="Attach file"
+        aria-label={attachLabel}
+        title={attachLabel}
         className={cn(
           "inline-flex items-center justify-center rounded-full text-muted-foreground hover:bg-accent hover:text-foreground transition-colors disabled:opacity-50 disabled:pointer-events-none",
           btnSize,

packages/ui/components/ui/pagination.tsx

@@ -1,4 +1,5 @@
 import * as React from "react"
+import { useTranslation } from "react-i18next"
 
 import { cn } from "@multica/ui/lib/utils"
 import { Button } from "@multica/ui/components/ui/button"
@@ -67,9 +68,10 @@ function PaginationPrevious({
   text = "Previous",
   ...props
 }: React.ComponentProps<typeof PaginationLink> & { text?: string }) {
+  const { t } = useTranslation("ui")
   return (
     <PaginationLink
-      aria-label="Go to previous page"
+      aria-label={t(($) => $.pagination_previous)}
       size="default"
       className={cn("pl-1.5!", className)}
       {...props}
@@ -85,9 +87,10 @@ function PaginationNext({
   text = "Next",
   ...props
 }: React.ComponentProps<typeof PaginationLink> & { text?: string }) {
+  const { t } = useTranslation("ui")
   return (
     <PaginationLink
-      aria-label="Go to next page"
+      aria-label={t(($) => $.pagination_next)}
       size="default"
       className={cn("pr-1.5!", className)}
       {...props}

packages/ui/components/ui/sidebar.tsx

@@ -4,6 +4,7 @@ import * as React from "react"
 import { mergeProps } from "@base-ui/react/merge-props"
 import { useRender } from "@base-ui/react/use-render"
 import { cva, type VariantProps } from "class-variance-authority"
+import { useTranslation } from "react-i18next"
 
 import { useIsMobile } from "@multica/ui/hooks/use-mobile"
 import { cn } from "@multica/ui/lib/utils"
@@ -265,6 +266,7 @@ function SidebarTrigger({
   ...props
 }: React.ComponentProps<typeof Button>) {
   const { toggleSidebar } = useSidebar()
+  const { t } = useTranslation("ui")
 
   return (
     <Button
@@ -280,13 +282,15 @@ function SidebarTrigger({
       {...props}
     >
       <PanelLeftIcon />
-      <span className="sr-only">Toggle Sidebar</span>
+      <span className="sr-only">{t(($) => $.toggle_sidebar)}</span>
     </Button>
   )
 }
 
 function SidebarRail({ className, ...props }: React.ComponentProps<"button">) {
   const { toggleSidebar, setWidth, setIsResizing } = useSidebar()
+  const { t } = useTranslation("ui")
+  const toggleLabel = t(($) => $.toggle_sidebar)
   const didDragRef = React.useRef(false)
   const dragRef = React.useRef<{ startX: number; startWidth: number } | null>(null)
 
@@ -330,11 +334,11 @@ function SidebarRail({ className, ...props }: React.ComponentProps<"button">) {
     <button
       data-sidebar="rail"
       data-slot="sidebar-rail"
-      aria-label="Toggle Sidebar"
+      aria-label={toggleLabel}
       tabIndex={-1}
       onClick={handleClick}
       onMouseDown={onMouseDown}
-      title="Toggle Sidebar"
+      title={toggleLabel}
       className={cn(
         "absolute inset-y-0 z-20 hidden w-4 transition-all ease-linear group-data-[side=left]:-right-4 group-data-[side=right]:left-0 after:absolute after:inset-y-0 after:start-1/2 after:w-[2px] hover:after:bg-sidebar-border sm:flex ltr:-translate-x-1/2 rtl:-translate-x-1/2",
         "in-data-[side=left]:cursor-col-resize in-data-[side=right]:cursor-col-resize",

packages/ui/hooks/use-scroll-fade.ts

@@ -43,11 +43,19 @@ export function useScrollFade(
     el.addEventListener("scroll", update, { passive: true });
     const ro = new ResizeObserver(update);
     ro.observe(el);
+    // ResizeObserver only fires on the container's own box. When children
+    // grow inside a flex/auto-height parent (e.g. async-loaded list items,
+    // collapsibles), scrollHeight changes but clientHeight does not — the
+    // mask would stay "none" until the user scrolls. MutationObserver on
+    // childList catches those content insertions.
+    const mo = new MutationObserver(update);
+    mo.observe(el, { childList: true, subtree: true });
 
     return () => {
       cancelAnimationFrame(frame);
       el.removeEventListener("scroll", update);
       ro.disconnect();
+      mo.disconnect();
     };
   }, [ref, update]);
 

packages/ui/markdown/CodeBlock.tsx

@@ -1,6 +1,7 @@
 import * as React from 'react'
 import { codeToHtml, bundledLanguages, type BundledLanguage } from 'shiki'
 import { Copy, Check } from "lucide-react"
+import { useTranslation } from "react-i18next"
 import { Button } from "@multica/ui/components/ui/button"
 import { Tooltip, TooltipTrigger, TooltipContent } from "@multica/ui/components/ui/tooltip"
 import { cn } from '@multica/ui/lib/utils'
@@ -61,6 +62,7 @@ export function CodeBlock({
   className,
   mode = 'full'
 }: CodeBlockProps): React.JSX.Element {
+  const { t } = useTranslation("ui")
   const [highlighted, setHighlighted] = React.useState<string | null>(null)
   const [isLoading, setIsLoading] = React.useState(true)
   const [copied, setCopied] = React.useState(false)
@@ -178,7 +180,7 @@ export function CodeBlock({
       {/* Language label + copy button */}
       <div className="flex items-center justify-between px-3 py-1.5 bg-muted/50 border-b text-xs">
         <span className="text-muted-foreground font-medium uppercase tracking-wide">
-          {resolvedLang !== 'text' ? resolvedLang : 'plain text'}
+          {resolvedLang !== 'text' ? resolvedLang : t(($) => $.plain_text)}
         </span>
         <Tooltip>
           <TooltipTrigger
@@ -188,7 +190,7 @@ export function CodeBlock({
                 size="icon-xs"
                 onClick={handleCopy}
                 className="opacity-0 group-hover:opacity-100 transition-opacity text-muted-foreground hover:text-foreground"
-                aria-label="Copy code"
+                aria-label={t(($) => $.copy_code)}
               >
                 {copied ? (
                   <Check className="size-3.5 text-success" />
@@ -198,7 +200,7 @@ export function CodeBlock({
               </Button>
             }
           />
-          <TooltipContent>Copy code</TooltipContent>
+          <TooltipContent>{t(($) => $.copy_code)}</TooltipContent>
         </Tooltip>
       </div>
 

packages/ui/package.json

@@ -17,6 +17,7 @@
     "./hooks/*": "./hooks/*.ts",
     "./lib/utils": "./lib/utils.ts",
     "./lib/data-table": "./lib/data-table.ts",
+    "./i18n-types": "./types/i18next.ts",
     "./styles/tokens.css": "./styles/tokens.css",
     "./styles/base.css": "./styles/base.css"
   },
@@ -52,8 +53,10 @@
     "vaul": "^1.1.2"
   },
   "peerDependencies": {
+    "i18next": "catalog:",
     "react": "catalog:",
-    "react-dom": "catalog:"
+    "react-dom": "catalog:",
+    "react-i18next": "catalog:"
   },
   "devDependencies": {
     "@multica/tsconfig": "workspace:*",

packages/ui/types/i18next.ts

@@ -0,0 +1,35 @@
+import "i18next";
+
+// Local slice of the i18next augmentation that owns the `ui` namespace.
+// The base augmentation lives in packages/views/i18n/resources-types.ts and
+// declares everything else; this file contributes only the `ui` entry via
+// declaration merging on the global `I18nResources` interface so
+// packages/ui can typecheck the selector form standalone without depending
+// on @multica/views.
+//
+// When both files are loaded together (in a consumer's typecheck program),
+// the two augmentations compose: views contributes common/auth/... and ui
+// contributes `ui`. No properties overlap, so the merge is conflict-free.
+//
+// The resource shape is mirrored from packages/views/locales/{en,zh-Hans}/ui.json.
+// Drift between the JSON and these types is not caught by the locale parity
+// test — if you add a key to ui.json, mirror it here.
+declare global {
+  interface I18nResources {
+    ui: {
+      attach_file: string;
+      toggle_sidebar: string;
+      pagination_previous: string;
+      pagination_next: string;
+      copy_code: string;
+      plain_text: string;
+    };
+  }
+}
+
+declare module "i18next" {
+  interface CustomTypeOptions {
+    resources: I18nResources;
+    enableSelector: true;
+  }
+}

packages/views/agents/components/agent-overview-pane.tsx

@@ -208,7 +208,7 @@ export function AgentOverviewPane({
   );
 }
 
-// Centred, max-width container shared by every config tab. `h-full flex
+// Padded, full-width container shared by every config tab. `h-full flex
 // flex-col` lets a tab opt into "fill the viewport" by giving its root
 // element `flex-1 min-h-0` (Instructions does this so the editor expands
 // instead of pushing the Save row off-screen). Tabs that don't opt in
@@ -216,6 +216,6 @@ export function AgentOverviewPane({
 // list) still scrolls via the parent's overflow-y-auto.
 function TabContent({ children }: { children: React.ReactNode }) {
   return (
-    <div className="mx-auto flex h-full max-w-2xl flex-col p-4 md:p-6">{children}</div>
+    <div className="flex h-full flex-col p-4 md:p-6">{children}</div>
   );
 }

packages/views/agents/components/agents-page.tsx

@@ -280,35 +280,24 @@ export function AgentsPage() {
     if (view === "archived" && archivedCount === 0) setView("active");
   }, [view, archivedCount]);
 
-  const handleCreate = async (data: CreateAgentRequest) => {
+  const handleCreate = async (data: CreateAgentRequest): Promise<Agent> => {
     const agent = await api.createAgent(data);
-    let cachedAgent = agent;
-    // When duplicating, carry the source agent's skill assignments over.
-    // Skills aren't part of CreateAgentRequest (they're managed via
-    // setAgentSkills) so the create endpoint can't take them inline; we
-    // do a follow-up call. Failure here doesn't abort the duplicate —
-    // the agent already exists and the user can re-attach skills from
-    // the detail page.
-    if (duplicateTemplate?.skills.length) {
-      try {
-        await api.setAgentSkills(agent.id, {
-          skill_ids: duplicateTemplate.skills.map((s) => s.id),
-        });
-        cachedAgent = { ...agent, skills: duplicateTemplate.skills };
-      } catch {
-        // Surfaced softly; the agent itself is fine.
-      }
-    }
+    // Skill follow-up is now owned by the dialog (it reads the user's
+    // form selection, which already includes the duplicate source's
+    // skills as a default when applicable). The dialog will call
+    // setAgentSkills after we return; we just have to surface the
+    // created agent so it can.
     qc.setQueryData<Agent[]>(workspaceKeys.agents(wsId), (current = []) => {
-      const exists = current.some((a) => a.id === cachedAgent.id);
+      const exists = current.some((a) => a.id === agent.id);
       return exists
-        ? current.map((a) => (a.id === cachedAgent.id ? cachedAgent : a))
-        : [...current, cachedAgent];
+        ? current.map((a) => (a.id === agent.id ? agent : a))
+        : [...current, agent];
     });
     setShowCreate(false);
     setDuplicateTemplate(null);
     navigation.push(paths.agentDetail(agent.id));
     qc.invalidateQueries({ queryKey: workspaceKeys.agents(wsId) });
+    return agent;
   };
 
   const handleDuplicate = useCallback((agent: Agent) => {
@@ -467,6 +456,7 @@ export function AgentsPage() {
           members={members}
           currentUserId={currentUser?.id ?? null}
           template={duplicateTemplate}
+          existingAgentNames={agents.map((a) => a.name)}
           onClose={() => {
             setShowCreate(false);
             setDuplicateTemplate(null);

packages/views/agents/components/avatar-picker.tsx

@@ -0,0 +1,143 @@
+"use client";
+
+import { useRef, useState } from "react";
+import { Camera, ImagePlus, Loader2, X } from "lucide-react";
+import { toast } from "sonner";
+import { api } from "@multica/core/api";
+import { useFileUpload } from "@multica/core/hooks/use-file-upload";
+import { cn } from "@multica/ui/lib/utils";
+import { useT } from "../../i18n";
+
+interface AvatarPickerProps {
+  /** Current avatar URL. null when nothing chosen yet. */
+  value: string | null;
+  /** Fires after a successful upload — the parent stashes the URL for the
+   *  create call. Re-fires with null when the user clears the choice. */
+  onChange: (url: string | null) => void;
+  /** Pixel size of the square. Defaults to 56 (h-14 / w-14), which lines
+   *  up vertically with the Name + Description stack in the create-agent
+   *  form so the two read as a single visual row. */
+  size?: number;
+}
+
+/**
+ * Compact avatar picker — a single square that lives next to the Name
+ * input in the create-agent form. Mirrors the visual language of
+ * agent-detail-inspector.tsx (Camera overlay on hover, file input behind
+ * the scenes), so users who've configured an avatar elsewhere in the app
+ * recognise the affordance immediately.
+ *
+ * No avatar yet → dashed placeholder with an ImagePlus icon.
+ * Has avatar    → image fills the square, hover dims it with a Camera
+ *                 overlay for "click to change". A small × in the corner
+ *                 clears the choice.
+ */
+export function AvatarPicker({ value, onChange, size = 56 }: AvatarPickerProps) {
+  const { t } = useT("agents");
+  const fileInputRef = useRef<HTMLInputElement>(null);
+  const { upload, uploading } = useFileUpload(api);
+  const [previewError, setPreviewError] = useState(false);
+
+  const handleFile = async (e: React.ChangeEvent<HTMLInputElement>) => {
+    const file = e.target.files?.[0];
+    if (!file) return;
+    e.target.value = ""; // allow re-selecting the same file
+    if (!file.type.startsWith("image/")) {
+      toast.error(t(($) => $.create_dialog.avatar.select_image_toast));
+      return;
+    }
+    try {
+      const result = await upload(file);
+      if (!result) return;
+      setPreviewError(false);
+      onChange(result.link);
+    } catch (err) {
+      toast.error(
+        err instanceof Error
+          ? err.message
+          : t(($) => $.create_dialog.avatar.upload_failed_toast),
+      );
+    }
+  };
+
+  const hasValue = !!value && !previewError;
+  const dimensionStyle = { width: size, height: size };
+
+  return (
+    <div className="relative shrink-0" style={dimensionStyle}>
+      <button
+        type="button"
+        onClick={() => fileInputRef.current?.click()}
+        disabled={uploading}
+        className={cn(
+          "group relative h-full w-full overflow-hidden rounded-lg outline-none transition-colors",
+          "focus-visible:ring-2 focus-visible:ring-ring",
+          hasValue
+            ? "border bg-muted"
+            : "border border-dashed bg-muted/40 hover:bg-muted",
+        )}
+        aria-label={
+          hasValue
+            ? t(($) => $.create_dialog.avatar.change_aria)
+            : t(($) => $.create_dialog.avatar.upload_aria)
+        }
+        style={dimensionStyle}
+      >
+        {hasValue ? (
+          <img
+            src={value ?? undefined}
+            alt=""
+            className="h-full w-full object-cover"
+            onError={() => setPreviewError(true)}
+          />
+        ) : (
+          <div className="flex h-full w-full items-center justify-center text-muted-foreground">
+            {uploading ? (
+              <Loader2 className="h-5 w-5 animate-spin" />
+            ) : (
+              <ImagePlus className="h-5 w-5" />
+            )}
+          </div>
+        )}
+
+        {/* Hover overlay only when there's already an image — otherwise the
+            placeholder icon already invites the click. */}
+        {hasValue && (
+          <div className="absolute inset-0 flex items-center justify-center bg-black/40 opacity-0 transition-opacity group-hover:opacity-100">
+            {uploading ? (
+              <Loader2 className="h-4 w-4 animate-spin text-white" />
+            ) : (
+              <Camera className="h-4 w-4 text-white" />
+            )}
+          </div>
+        )}
+      </button>
+
+      {/* Tiny X to clear, only shown when there's a value. Positioned just
+          outside the avatar's top-right corner so it doesn't cover the
+          image. */}
+      {hasValue && !uploading && (
+        <button
+          type="button"
+          onClick={(e) => {
+            e.stopPropagation();
+            onChange(null);
+            setPreviewError(false);
+          }}
+          className="absolute -right-1.5 -top-1.5 flex h-5 w-5 items-center justify-center rounded-full border bg-background text-muted-foreground shadow-sm transition-colors hover:bg-muted hover:text-foreground"
+          aria-label={t(($) => $.create_dialog.avatar.remove_aria)}
+        >
+          <X className="h-3 w-3" />
+        </button>
+      )}
+
+      <input
+        ref={fileInputRef}
+        type="file"
+        accept="image/*"
+        className="hidden"
+        onChange={handleFile}
+      />
+    </div>
+  );
+}

packages/views/agents/components/create-agent-dialog.test.tsx

@@ -1,13 +1,24 @@
 // @vitest-environment jsdom
 
-import { describe, it, expect, vi, beforeEach } from "vitest";
+import { describe, it, expect, vi, beforeEach, afterEach } from "vitest";
 import { QueryClient, QueryClientProvider } from "@tanstack/react-query";
-import { render, screen, fireEvent } from "@testing-library/react";
+import { render, screen, fireEvent, cleanup } from "@testing-library/react";
 import type { Agent, MemberWithUser, RuntimeDevice } from "@multica/core/types";
 import { I18nProvider } from "@multica/core/i18n/react";
+import { WorkspaceSlugProvider } from "@multica/core/paths";
+import { NavigationProvider, type NavigationAdapter } from "../../navigation";
 import enCommon from "../../locales/en/common.json";
 import enAgents from "../../locales/en/agents.json";
 
+const navigationStub: NavigationAdapter = {
+  push: vi.fn(),
+  replace: vi.fn(),
+  back: vi.fn(),
+  pathname: "/",
+  searchParams: new URLSearchParams(),
+  getShareableUrl: (path: string) => path,
+};
+
 const TEST_RESOURCES = { en: { common: enCommon, agents: enAgents } };
 
 vi.mock("@multica/core/hooks", () => ({
@@ -120,22 +131,42 @@ function renderDialog(runtimes: RuntimeDevice[], template?: Agent) {
   render(
     <I18nProvider locale="en" resources={TEST_RESOURCES}>
       <QueryClientProvider client={queryClient}>
-        <CreateAgentDialog
-          runtimes={runtimes}
-          members={members}
-          currentUserId={ME}
-          template={template}
-          onClose={onClose}
-          onCreate={onCreate}
-        />
+        <WorkspaceSlugProvider slug="test-ws">
+        <NavigationProvider value={navigationStub}>
+          <CreateAgentDialog
+            runtimes={runtimes}
+            members={members}
+            currentUserId={ME}
+            template={template}
+            onClose={onClose}
+            onCreate={onCreate}
+          />
+        </NavigationProvider>
+        </WorkspaceSlugProvider>
       </QueryClientProvider>
     </I18nProvider>,
   );
+  // Without a `template`, the dialog opens on the blank-vs-template
+  // chooser. These tests target the manual form's runtime picker, so
+  // advance through the chooser to the form. Duplicate mode jumps
+  // straight to the form and doesn't render the chooser.
+  if (!template) {
+    fireEvent.click(screen.getByText(enAgents.create_dialog.chooser.blank_title));
+  }
   return { onCreate, onClose };
 }
 
 describe("CreateAgentDialog runtime visibility gate", () => {
   beforeEach(() => vi.clearAllMocks());
+  // Base UI Dialog renders into a portal on document.body and leaves
+  // focus-guard / inert wrapper divs around after the React tree unmounts.
+  // The auto-cleanup from @testing-library/react drops the container but
+  // not the portal residue, so two-tests-in-a-row queries see double
+  // matches ("All", "My Runtime"). Force cleanup + wipe body between tests.
+  afterEach(() => {
+    cleanup();
+    document.body.innerHTML = "";
+  });
 
   it("disables another member's private runtime in the picker", () => {
     const mine = makeRuntime({ id: "rt-mine", name: "My Runtime", owner_id: ME, visibility: "private" });

packages/views/agents/components/instructions-editor.tsx

@@ -0,0 +1,116 @@
+"use client";
+
+import { useRef, useState } from "react";
+import { ChevronDown, FileText, X } from "lucide-react";
+import { ContentEditor, type ContentEditorRef } from "../../editor";
+import { Button } from "@multica/ui/components/ui/button";
+import { cn } from "@multica/ui/lib/utils";
+import { useT } from "../../i18n";
+
+interface InstructionsEditorProps {
+  /** Markdown source. Used both as default value when expanded and as
+   *  preview text when collapsed. */
+  value: string;
+  /** Fires on every keystroke (debounced inside ContentEditor). */
+  onChange: (value: string) => void;
+  /** Optional placeholder override. Defaults to the i18n "click to write"
+   *  copy; the create dialog passes the duplicate-specific string for
+   *  agents being cloned. */
+  placeholder?: string;
+}
+
+/**
+ * Collapsible Instructions field for the create-agent dialog. Stays compact
+ * until the user wants to write — most agents only need instructions when
+ * they're being authored carefully, not on every quick-create.
+ *
+ * Two states:
+ *   collapsed → small clickable card, shows a preview of `value` (or the
+ *               placeholder when empty). One click expands.
+ *   expanded  → full ContentEditor (markdown, bubble menu, mention support,
+ *               attachment upload). "Collapse" button on the right of the
+ *               header tucks it back; value is preserved.
+ */
+export function InstructionsEditor({
+  value,
+  onChange,
+  placeholder,
+}: InstructionsEditorProps) {
+  const { t } = useT("agents");
+  const [expanded, setExpanded] = useState(false);
+  const editorRef = useRef<ContentEditorRef>(null);
+
+  const label = t(($) => $.create_dialog.instructions.label);
+  const resolvedPlaceholder =
+    placeholder ?? t(($) => $.create_dialog.instructions.placeholder_blank);
+
+  const expand = () => {
+    setExpanded(true);
+    // Focus on next tick so the editor mounts first.
+    setTimeout(() => editorRef.current?.focus(), 0);
+  };
+
+  if (!expanded) {
+    return (
+      <div>
+        <div className="text-xs font-medium uppercase tracking-wider text-muted-foreground">
+          {label}
+        </div>
+        <button
+          type="button"
+          onClick={expand}
+          className="mt-1.5 flex w-full items-start gap-2.5 rounded-lg border bg-card px-3 py-3 text-left transition-colors hover:border-primary/40 hover:bg-accent/40"
+        >
+          <FileText className="mt-0.5 h-3.5 w-3.5 shrink-0 text-muted-foreground" />
+          <div className="min-w-0 flex-1">
+            {value.trim() ? (
+              // Preview: first 2 lines of markdown, ellipsised.
+              <div className="line-clamp-2 whitespace-pre-wrap text-sm text-foreground/80">
+                {value}
+              </div>
+            ) : (
+              <div className="text-sm text-muted-foreground">{resolvedPlaceholder}</div>
+            )}
+          </div>
+          <ChevronDown className="h-3.5 w-3.5 shrink-0 text-muted-foreground/40" />
+        </button>
+      </div>
+    );
+  }
+
+  return (
+    <div>
+      <div className="flex items-center justify-between">
+        <div className="text-xs font-medium uppercase tracking-wider text-muted-foreground">
+          {label}
+        </div>
+        <Button
+          type="button"
+          variant="ghost"
+          size="sm"
+          onClick={() => setExpanded(false)}
+          className="h-6 gap-1 px-2 text-xs"
+        >
+          <X className="h-3 w-3" />
+          {t(($) => $.create_dialog.instructions.collapse)}
+        </Button>
+      </div>
+      <div
+        className={cn(
+          "mt-1.5 rounded-lg border bg-card",
+          "focus-within:border-primary/40",
+        )}
+      >
+        <ContentEditor
+          ref={editorRef}
+          defaultValue={value}
+          onUpdate={onChange}
+          placeholder={t(($) => $.create_dialog.instructions.editor_placeholder)}
+          className="min-h-[160px] max-h-[320px] overflow-y-auto px-3 py-2.5 text-sm"
+          showBubbleMenu={true}
+          disableMentions={true}
+        />
+      </div>
+    </div>
+  );
+}

packages/views/agents/components/skill-add-dialog.tsx

@@ -1,10 +1,9 @@
 "use client";
 
-import { useState } from "react";
-import { FileText, Search } from "lucide-react";
+import { useMemo, useState } from "react";
 import { useQuery, useQueryClient } from "@tanstack/react-query";
 import { toast } from "sonner";
-import type { Agent } from "@multica/core/types";
+import type { Agent, SkillSummary } from "@multica/core/types";
 import { api } from "@multica/core/api";
 import { useWorkspaceId } from "@multica/core/hooks";
 import {
@@ -20,18 +19,19 @@ import {
   DialogHeader,
   DialogTitle,
 } from "@multica/ui/components/ui/dialog";
-import { Input } from "@multica/ui/components/ui/input";
 import { useT } from "../../i18n";
+import { SkillPickerList } from "./skill-picker-list";
 
 /**
- * Single source of truth for "attach a workspace skill to this agent".
- * Used by both:
- *   - SkillsTab — full surface, "Add skill" button
- *   - Inspector → SkillAttach — inline dashed `+ Attach` chip
+ * "Attach workspace skills to this agent." Multi-select with explicit
+ * Confirm — earlier iterations attached on a single row click, which
+ * meant the user couldn't tick several skills at once and the dialog
+ * closed before they could review their choice.
  *
- * Owns the workspace-skill list query, the "what's still attachable" filter,
- * the API call, and the optimistic invalidation. Callers only manage the
- * open/close state — they don't repeat the attach logic.
+ * Already-attached skills are filtered out of the list entirely (vs.
+ * showing them disabled). When there are no remaining workspace skills
+ * to attach, the empty-state copy explains why, and the Confirm button
+ * is naturally disabled because nothing can be selected.
  */
 export function SkillAddDialog({
   agent,
@@ -45,34 +45,44 @@ export function SkillAddDialog({
   const { t } = useT("agents");
   const wsId = useWorkspaceId();
   const qc = useQueryClient();
-  const { data: workspaceSkills = [] } = useQuery(skillListOptions(wsId));
+  const { data: workspaceSkills = [], isLoading } = useQuery(skillListOptions(wsId));
   const [saving, setSaving] = useState(false);
-  const [query, setQuery] = useState("");
+  const [selectedIds, setSelectedIds] = useState<Set<string>>(new Set());
 
-  const agentSkillIds = new Set(agent.skills.map((s) => s.id));
-  const availableSkills = workspaceSkills.filter(
-    (s) => !agentSkillIds.has(s.id),
+  const attachedIds = useMemo(
+    () => new Set(agent.skills.map((s) => s.id)),
+    [agent.skills],
+  );
+  // Hide attached skills outright — the dialog is for adding new ones.
+  // If a user wants to see what's already on the agent, the SkillsTab
+  // list above shows it.
+  const availableSkills = useMemo(
+    () => workspaceSkills.filter((s) => !attachedIds.has(s.id)),
+    [workspaceSkills, attachedIds],
   );
-  const trimmedQuery = query.trim().toLowerCase();
-  const filteredSkills = trimmedQuery
-    ? availableSkills.filter((s) => {
-        const name = s.name.toLowerCase();
-        const description = s.description?.toLowerCase() ?? "";
-        return (
-          name.includes(trimmedQuery) || description.includes(trimmedQuery)
-        );
-      })
-    : availableSkills;
 
   const handleOpenChange = (v: boolean) => {
-    if (!v) setQuery("");
+    if (!v) setSelectedIds(new Set());
     onOpenChange(v);
   };
 
-  const handleAdd = async (skillId: string) => {
+  const handleToggle = (skill: SkillSummary) => {
+    setSelectedIds((prev) => {
+      const next = new Set(prev);
+      if (next.has(skill.id)) next.delete(skill.id);
+      else next.add(skill.id);
+      return next;
+    });
+  };
+
+  const handleConfirm = async () => {
+    if (selectedIds.size === 0) return;
     setSaving(true);
     try {
-      const newIds = [...agent.skills.map((s) => s.id), skillId];
+      const newIds = [
+        ...agent.skills.map((s) => s.id),
+        ...selectedIds,
+      ];
       await api.setAgentSkills(agent.id, { skill_ids: newIds });
       qc.invalidateQueries({ queryKey: workspaceKeys.agents(wsId) });
       handleOpenChange(false);
@@ -83,65 +93,46 @@ export function SkillAddDialog({
     }
   };
 
-  const showSearch = availableSkills.length > 0;
-  const noMatch = showSearch && filteredSkills.length === 0;
+  const count = selectedIds.size;
 
   return (
     <Dialog open={open} onOpenChange={handleOpenChange}>
       <DialogContent className="max-w-md">
         <DialogHeader>
-          <DialogTitle className="text-sm">{t(($) => $.tab_body.skills.add_dialog_title)}</DialogTitle>
+          <DialogTitle className="text-sm">
+            {t(($) => $.tab_body.skills.add_dialog_title)}
+          </DialogTitle>
           <DialogDescription className="text-xs">
             {t(($) => $.tab_body.skills.add_dialog_description)}
           </DialogDescription>
         </DialogHeader>
-        {showSearch && (
-          <div className="relative">
-            <Search className="pointer-events-none absolute left-2.5 top-1/2 h-3.5 w-3.5 -translate-y-1/2 text-muted-foreground" />
-            <Input
-              autoFocus
-              value={query}
-              onChange={(e) => setQuery(e.target.value)}
-              placeholder={t(($) => $.tab_body.skills.add_dialog_search_placeholder)}
-              aria-label={t(($) => $.tab_body.skills.add_dialog_search_placeholder)}
-              className="pl-7"
-            />
-          </div>
-        )}
-        <div className="max-h-64 space-y-1 overflow-y-auto">
-          {filteredSkills.map((skill) => (
-            <button
-              key={skill.id}
-              onClick={() => handleAdd(skill.id)}
-              disabled={saving}
-              className="flex w-full items-center gap-2.5 rounded-md px-3 py-2 text-left text-sm transition-colors hover:bg-accent/50"
-            >
-              <FileText className="h-3.5 w-3.5 shrink-0 text-muted-foreground" />
-              <div className="min-w-0 flex-1">
-                <div className="truncate font-medium">{skill.name}</div>
-                {skill.description && (
-                  <div className="truncate text-xs text-muted-foreground">
-                    {skill.description}
-                  </div>
-                )}
-              </div>
-            </button>
-          ))}
-          {availableSkills.length === 0 && (
-            <p className="py-6 text-center text-xs text-muted-foreground">
-              {t(($) => $.tab_body.skills.add_dialog_empty)}
-            </p>
-          )}
-          {noMatch && (
-            <p className="py-6 text-center text-xs text-muted-foreground">
-              {t(($) => $.tab_body.skills.add_dialog_no_match)}
-            </p>
-          )}
-        </div>
+
+        <SkillPickerList
+          skills={availableSkills}
+          selectedIds={selectedIds}
+          onToggle={handleToggle}
+          loading={isLoading}
+          emptyMessage={
+            workspaceSkills.length === 0
+              ? t(($) => $.tab_body.skills.add_dialog_empty)
+              : t(($) => $.tab_body.skills.add_dialog_empty_partial)
+          }
+        />
+
         <DialogFooter>
           <Button variant="ghost" onClick={() => handleOpenChange(false)}>
             {t(($) => $.tab_body.skills.add_dialog_cancel)}
           </Button>
+          <Button
+            onClick={handleConfirm}
+            disabled={count === 0 || saving}
+          >
+            {saving
+              ? t(($) => $.tab_body.skills.add_dialog_saving)
+              : count > 0
+                ? t(($) => $.tab_body.skills.add_dialog_confirm, { count })
+                : t(($) => $.tab_body.skills.add_dialog_confirm_default)}
+          </Button>
         </DialogFooter>
       </DialogContent>
     </Dialog>

packages/views/agents/components/skill-multi-select.tsx

@@ -0,0 +1,104 @@
+"use client";
+
+import { useState } from "react";
+import { ChevronDown, Plus, X } from "lucide-react";
+import { useQuery } from "@tanstack/react-query";
+import type { SkillSummary } from "@multica/core/types";
+import { useWorkspaceId } from "@multica/core/hooks";
+import { skillListOptions } from "@multica/core/workspace/queries";
+import { Button } from "@multica/ui/components/ui/button";
+import { useT } from "../../i18n";
+import { SkillPickerList } from "./skill-picker-list";
+
+interface SkillMultiSelectProps {
+  /** Currently-selected skill IDs (controlled). */
+  selectedIds: ReadonlySet<string>;
+  /** Replaces the selection on every toggle. */
+  onChange: (next: Set<string>) => void;
+}
+
+/**
+ * Multi-select wrapper for the create-agent form. Collapsed by default;
+ * expands into a SkillPickerList configured for toggle behaviour
+ * (click adds to / removes from the local selection set).
+ *
+ * Shares its visual surface with SkillAddDialog via SkillPickerList —
+ * one component owns search + row rendering + indicators, so a tweak
+ * to either appears identically in both flows.
+ */
+export function SkillMultiSelect({
+  selectedIds,
+  onChange,
+}: SkillMultiSelectProps) {
+  const { t } = useT("agents");
+  const wsId = useWorkspaceId();
+  const { data: workspaceSkills = [], isLoading } = useQuery(skillListOptions(wsId));
+  const [expanded, setExpanded] = useState(selectedIds.size > 0);
+
+  const label = t(($) => $.create_dialog.skills_section.label);
+
+  const toggle = (skill: SkillSummary) => {
+    const next = new Set(selectedIds);
+    if (next.has(skill.id)) next.delete(skill.id);
+    else next.add(skill.id);
+    onChange(next);
+  };
+
+  if (!expanded) {
+    return (
+      <div>
+        <div className="text-xs font-medium uppercase tracking-wider text-muted-foreground">
+          {label}
+        </div>
+        <button
+          type="button"
+          onClick={() => setExpanded(true)}
+          className="mt-1.5 flex w-full items-center gap-2.5 rounded-lg border bg-card px-3 py-3 text-left transition-colors hover:border-primary/40 hover:bg-accent/40"
+        >
+          <Plus className="h-3.5 w-3.5 shrink-0 text-muted-foreground" />
+          <div className="min-w-0 flex-1 truncate text-sm text-muted-foreground">
+            {selectedIds.size > 0
+              ? t(($) => $.create_dialog.skills_section.selected, {
+                  count: selectedIds.size,
+                })
+              : t(($) => $.create_dialog.skills_section.placeholder)}
+          </div>
+          <ChevronDown className="h-3.5 w-3.5 shrink-0 text-muted-foreground/40" />
+        </button>
+      </div>
+    );
+  }
+
+  return (
+    <div>
+      <div className="flex items-center justify-between">
+        <div className="text-xs font-medium uppercase tracking-wider text-muted-foreground">
+          {label}
+          {selectedIds.size > 0 ? (
+            <span className="ml-2 text-foreground/60">({selectedIds.size})</span>
+          ) : null}
+        </div>
+        <Button
+          type="button"
+          variant="ghost"
+          size="sm"
+          onClick={() => setExpanded(false)}
+          className="h-6 gap-1 px-2 text-xs"
+        >
+          <X className="h-3 w-3" />
+          {t(($) => $.create_dialog.skills_section.collapse)}
+        </Button>
+      </div>
+
+      <div className="mt-1.5">
+        <SkillPickerList
+          skills={workspaceSkills}
+          selectedIds={selectedIds}
+          onToggle={toggle}
+          loading={isLoading}
+          emptyMessage={t(($) => $.create_dialog.skills_section.list_empty_multi)}
+        />
+      </div>
+    </div>
+  );
+}

packages/views/agents/components/skill-picker-list.tsx

@@ -0,0 +1,141 @@
+"use client";
+
+import { useState } from "react";
+import { FileText, Search } from "lucide-react";
+import type { SkillSummary } from "@multica/core/types";
+import { Checkbox } from "@multica/ui/components/ui/checkbox";
+import { Input } from "@multica/ui/components/ui/input";
+import { cn } from "@multica/ui/lib/utils";
+import { useT } from "../../i18n";
+
+interface SkillPickerListProps {
+  /** Skills to show. Callers filter (e.g. exclude already-attached
+   *  skills in SkillAddDialog) before passing — this component just
+   *  renders the rows. */
+  skills: readonly SkillSummary[];
+
+  /** Currently-toggled rows. Selected rows get a checked Checkbox and a
+   *  subtle background; click toggles. */
+  selectedIds: ReadonlySet<string>;
+
+  /** Fires on every row click. Caller updates `selectedIds`. */
+  onToggle: (skill: SkillSummary) => void;
+
+  /** Show the search input at the top. Default true. */
+  searchable?: boolean;
+
+  /** Loading state for the skills query. */
+  loading?: boolean;
+
+  /** Caller-supplied empty / no-match copy. Falls back to generic i18n
+   *  strings when omitted — the dialog and the create-form pass their
+   *  own flavour-specific copy. */
+  emptyMessage?: string;
+  noMatchMessage?: string;
+
+  /** Outer-wrapper className. Defaults to `w-full`; callers pass
+   *  e.g. `max-w-md` to constrain width. */
+  className?: string;
+}
+
+/**
+ * Headless multi-select list of workspace skills. Used by both
+ * SkillAddDialog (filtered to unattached skills) and SkillMultiSelect
+ * (create-form selection). One surface owns row layout, the search
+ * input, empty/loading states, and the shadcn Checkbox indicator, so
+ * tweaks land in one place.
+ *
+ * Rows truncate the name + description columns inside `flex-1 min-w-0`
+ * so long text doesn't push the Checkbox out of view.
+ */
+export function SkillPickerList({
+  skills,
+  selectedIds,
+  onToggle,
+  searchable = true,
+  loading = false,
+  emptyMessage,
+  noMatchMessage,
+  className,
+}: SkillPickerListProps) {
+  const { t } = useT("agents");
+  const [query, setQuery] = useState("");
+
+  const trimmedQuery = query.trim().toLowerCase();
+  const filtered = trimmedQuery
+    ? skills.filter((s) => {
+        const name = s.name.toLowerCase();
+        const description = s.description?.toLowerCase() ?? "";
+        return name.includes(trimmedQuery) || description.includes(trimmedQuery);
+      })
+    : skills;
+
+  const resolvedEmpty =
+    emptyMessage ?? t(($) => $.create_dialog.skills_section.list_empty_default);
+  const resolvedNoMatch =
+    noMatchMessage ?? t(($) => $.create_dialog.skills_section.list_no_match);
+
+  return (
+    <div className={cn("w-full overflow-hidden rounded-lg border bg-card", className)}>
+      {searchable && skills.length > 0 && (
+        <div className="border-b p-2">
+          <div className="relative">
+            <Search className="pointer-events-none absolute left-2.5 top-1/2 h-3.5 w-3.5 -translate-y-1/2 text-muted-foreground" />
+            <Input
+              value={query}
+              onChange={(e) => setQuery(e.target.value)}
+              placeholder={t(($) => $.create_dialog.skills_section.search_placeholder)}
+              className="h-8 pl-7 text-xs"
+            />
+          </div>
+        </div>
+      )}
+
+      <div className="max-h-64 space-y-0.5 overflow-y-auto p-1.5">
+        {loading ? (
+          <div className="py-6 text-center text-xs text-muted-foreground">
+            {t(($) => $.create_dialog.skills_section.list_loading)}
+          </div>
+        ) : skills.length === 0 ? (
+          <div className="py-6 text-center text-xs text-muted-foreground">{resolvedEmpty}</div>
+        ) : filtered.length === 0 ? (
+          <div className="py-6 text-center text-xs text-muted-foreground">{resolvedNoMatch}</div>
+        ) : (
+          filtered.map((skill) => {
+            const isSelected = selectedIds.has(skill.id);
+            return (
+              <button
+                key={skill.id}
+                type="button"
+                onClick={() => onToggle(skill)}
+                aria-pressed={isSelected}
+                className={cn(
+                  "flex w-full items-center gap-2.5 rounded-md px-2.5 py-2 text-left transition-colors",
+                  isSelected ? "bg-accent" : "hover:bg-accent/50",
+                )}
+              >
+                {/* Indicator only — the wrapping <button> handles clicks,
+                    so the Checkbox is non-interactive on its own. We
+                    pass `checked` so the visual matches the row state. */}
+                <Checkbox
+                  checked={isSelected}
+                  tabIndex={-1}
+                  className="pointer-events-none"
+                />
+                <FileText className="h-3.5 w-3.5 shrink-0 text-muted-foreground" />
+                <div className="min-w-0 flex-1">
+                  <div className="truncate text-sm font-medium">{skill.name}</div>
+                  {skill.description ? (
+                    <div className="truncate text-xs text-muted-foreground">
+                      {skill.description}
+                    </div>
+                  ) : null}
+                </div>
+              </button>
+            );
+          })
+        )}
+      </div>
+    </div>
+  );
+}

packages/views/agents/components/tabs/skills-tab.tsx

@@ -24,17 +24,16 @@ export function SkillsTab({
   const qc = useQueryClient();
   const wsId = useWorkspaceId();
   // Same query the SkillAddDialog uses (TanStack Query dedupes by key, so
-  // this isn't an extra request) — used here only to grey out the "Add skill"
-  // button when there's nothing left to attach.
+  // this isn't an extra request) — used here only to grey out the "Add
+  // skill" button when the workspace has zero skills total. When skills
+  // exist but are all already attached, we still open the dialog: it
+  // filters out attached skills and renders a localised "no more skills
+  // to add" empty state, which is more useful than a mysterious
+  // greyed-out button.
   const { data: workspaceSkills = [] } = useQuery(skillListOptions(wsId));
   const [removing, setRemoving] = useState(false);
   const [showAdd, setShowAdd] = useState(false);
 
-  const agentSkillIds = new Set(agent.skills.map((s) => s.id));
-  const availableCount = workspaceSkills.filter(
-    (s) => !agentSkillIds.has(s.id),
-  ).length;
-
   const handleRemove = async (skillId: string) => {
     setRemoving(true);
     try {
@@ -60,7 +59,7 @@ export function SkillsTab({
           variant="outline"
           size="sm"
           onClick={() => setShowAdd(true)}
-          disabled={availableCount === 0}
+          disabled={workspaceSkills.length === 0}
           className="shrink-0"
         >
           <Plus className="h-3 w-3" />
@@ -84,7 +83,7 @@ export function SkillsTab({
           <p className="mt-1 max-w-xs text-center text-xs text-muted-foreground">
             {t(($) => $.tab_body.skills.empty_hint)}
           </p>
-          {availableCount > 0 && (
+          {workspaceSkills.length > 0 && (
             <Button
               onClick={() => setShowAdd(true)}
               size="sm"

packages/views/agents/components/template-detail.tsx

@@ -0,0 +1,173 @@
+"use client";
+
+import { Check, ChevronRight, Loader2 } from "lucide-react";
+import { useQuery } from "@tanstack/react-query";
+import { agentTemplateDetailOptions } from "@multica/core/agents/queries";
+import type { AgentTemplateSummary } from "@multica/core/types";
+import { Button } from "@multica/ui/components/ui/button";
+import { cn } from "@multica/ui/lib/utils";
+import { useT } from "../../i18n";
+import { getAccentClass, getTemplateIcon } from "./template-picker";
+
+interface TemplateDetailProps {
+  template: AgentTemplateSummary;
+  /** Fired when the user clicks "Use this template" — the dialog calls
+   *  the create API and navigates to the new agent. */
+  onUse: (template: AgentTemplateSummary) => void;
+  /** True while the parent's create request is in flight; we disable the
+   *  Use button so the user can't double-click. */
+  creating?: boolean;
+  /** Upstream URLs the server reported as unreachable on the most recent
+   *  create attempt. Surfaces an inline error banner so the user knows
+   *  *why* Create didn't navigate. The detail step is the only place
+   *  this banner can render — `quickCreateFromTemplate` fires from here
+   *  and never advances to a different step on failure. */
+  failedURLs?: readonly string[] | null;
+}
+
+/**
+ * Step 3 of the create-agent flow: a read-only preview of the picked
+ * template — instructions, skill list with cached descriptions, and a
+ * "Use this template" CTA at the bottom. Clicking Use kicks off a
+ * one-shot create with default settings (no form step in between).
+ *
+ * Instructions come from the lazy-fetched detail endpoint (the picker
+ * only carries the summary). Cached through TanStack Query keyed by
+ * slug with `staleTime: Infinity`, so navigating back and forth between
+ * picker and detail doesn't re-fetch. Visual rhythm matches the picker
+ * card so the transition feels seamless.
+ */
+export function TemplateDetail({
+  template,
+  onUse,
+  creating = false,
+  failedURLs,
+}: TemplateDetailProps) {
+  const { t } = useT("agents");
+  const { data: detail, isLoading, error } = useQuery(
+    agentTemplateDetailOptions(template.slug),
+  );
+
+  const Icon = getTemplateIcon(template.icon);
+  const accentClass = getAccentClass(template.accent);
+
+  return (
+    <>
+      <div className="flex-1 overflow-y-auto">
+        <div className="mx-auto max-w-3xl p-6">
+          {/* failedURLs banner — sits above the header so it's the first
+              thing the user sees after the spinner clears on a 422. */}
+          {failedURLs && failedURLs.length > 0 && (
+            <div className="mb-5 rounded-lg border border-destructive/40 bg-destructive/5 p-3 text-sm">
+              <div className="font-medium text-destructive">
+                {t(($) => $.create_dialog.template_failure.title)}
+              </div>
+              <div className="mt-1 text-xs text-muted-foreground">
+                {t(($) => $.create_dialog.template_failure.body)}
+              </div>
+              <ul className="mt-2 space-y-0.5 text-xs">
+                {failedURLs.map((u) => (
+                  <li key={u} className="break-all font-mono">
+                    {u}
+                  </li>
+                ))}
+              </ul>
+            </div>
+          )}
+
+          {/* Header: icon + name + description. Same rhythm as the picker
+              card so the user reads the transition as "the same item,
+              expanded". */}
+          <div className="flex items-start gap-3">
+            <div className={cn("flex h-12 w-12 shrink-0 items-center justify-center rounded-lg", accentClass)}>
+              <Icon className="h-6 w-6" />
+            </div>
+            <div className="min-w-0 flex-1">
+              <h2 className="text-lg font-semibold">{template.name}</h2>
+              <p className="mt-0.5 text-sm text-muted-foreground">{template.description}</p>
+              {template.category ? (
+                <div className="mt-2 inline-flex items-center rounded-full bg-muted px-2 py-0.5 text-[10px] font-medium uppercase tracking-wider text-muted-foreground">
+                  {template.category}
+                </div>
+              ) : null}
+            </div>
+          </div>
+
+          {/* Skill list — always visible (summary has cached descriptions) */}
+          <section className="mt-6">
+            <h3 className="text-xs font-semibold uppercase tracking-wider text-muted-foreground">
+              {t(($) => $.create_dialog.template_detail.skill_count, {
+                count: template.skills.length,
+              })}
+            </h3>
+            <ul className="mt-3 space-y-2">
+              {template.skills.map((s) => (
+                <li
+                  key={s.source_url}
+                  className="rounded-lg border bg-card px-3 py-2.5"
+                >
+                  <div className="flex items-center gap-2">
+                    <Check className="h-4 w-4 text-success" />
+                    <span className="font-mono text-xs font-medium">{s.cached_name}</span>
+                  </div>
+                  {s.cached_description ? (
+                    <p className="mt-1 ml-6 text-xs text-muted-foreground">
+                      {s.cached_description}
+                    </p>
+                  ) : null}
+                </li>
+              ))}
+            </ul>
+          </section>
+
+          {/* Instructions — lazy fetch + loading/error states */}
+          <section className="mt-6">
+            <h3 className="text-xs font-semibold uppercase tracking-wider text-muted-foreground">
+              {t(($) => $.create_dialog.template_detail.instructions_label)}
+            </h3>
+            <div className="mt-3 rounded-lg border bg-muted/30 px-4 py-3">
+              {isLoading ? (
+                <div className="flex items-center gap-2 text-xs text-muted-foreground">
+                  <Loader2 className="h-3.5 w-3.5 animate-spin" />
+                  {t(($) => $.create_dialog.template_detail.instructions_loading)}
+                </div>
+              ) : error ? (
+                <div className="text-xs text-destructive">
+                  {error instanceof Error
+                    ? error.message
+                    : t(($) => $.create_dialog.template_detail.load_failed)}
+                </div>
+              ) : (
+                <pre className="max-h-60 overflow-y-auto whitespace-pre-wrap text-xs leading-relaxed text-foreground/80">
+                  {detail?.instructions ?? ""}
+                </pre>
+              )}
+            </div>
+          </section>
+        </div>
+      </div>
+
+      {/* Sticky CTA footer — click Use kicks off the create API call;
+          parent shows a creating spinner and navigates on success. */}
+      <div className="flex items-center justify-end gap-2 border-t bg-background px-5 py-3">
+        <Button
+          onClick={() => onUse(template)}
+          disabled={creating}
+          className="gap-1.5"
+        >
+          {creating ? (
+            <>
+              <Loader2 className="h-4 w-4 animate-spin" />
+              {t(($) => $.create_dialog.template_detail.creating)}
+            </>
+          ) : (
+            <>
+              {t(($) => $.create_dialog.template_detail.use)}
+              <ChevronRight className="h-4 w-4" />
+            </>
+          )}
+        </Button>
+      </div>
+    </>
+  );
+}

packages/views/agents/components/template-picker.tsx

@@ -0,0 +1,203 @@
+"use client";
+
+import { useMemo } from "react";
+import {
+  Brush,
+  ChevronRight,
+  FileText,
+  FlaskConical,
+  LayoutDashboard,
+  ListChecks,
+  Loader2,
+  Megaphone,
+  Palette,
+  PenLine,
+  Presentation,
+  Search,
+  Sparkles,
+} from "lucide-react";
+import type { LucideIcon } from "lucide-react";
+import { useQuery } from "@tanstack/react-query";
+import { agentTemplateListOptions } from "@multica/core/agents/queries";
+import type { AgentTemplateSummary } from "@multica/core/types";
+import { cn } from "@multica/ui/lib/utils";
+import { useT } from "../../i18n";
+
+interface TemplatePickerProps {
+  /** Fired when a template card is clicked. The dialog advances to the
+   *  detail step (which shows instructions + skills + Use button). */
+  onSelect: (template: AgentTemplateSummary) => void;
+}
+
+/**
+ * Step 2 of the create-agent flow: a 2-column grid of template cards,
+ * grouped by `category`. Clicking a card moves to the detail step.
+ *
+ * Templates are a static catalog (workspace-independent, only changes on
+ * server deploy), so the catalog is loaded through TanStack Query with
+ * `staleTime: Infinity` — re-opening the picker hits the cache instantly
+ * and there's no per-mount refetch.
+ *
+ * Icons and accent colors come from the template JSON itself (`icon` is a
+ * lucide-react name, `accent` is a Multica semantic token). Resolved
+ * through static maps (ICONS / ACCENTS) so Tailwind can JIT-scan every
+ * class variant — dynamic `bg-${accent}/10` strings would silently not
+ * generate.
+ */
+export function TemplatePicker({ onSelect }: TemplatePickerProps) {
+  const { t } = useT("agents");
+  const { data: templates = [], isLoading, error } = useQuery(
+    agentTemplateListOptions(),
+  );
+
+  // Group by category. Templates without a category fall into the
+  // localised "Other" bucket so they still render. Preserves the load
+  // order within each group for deterministic UI (matches the
+  // alphabetic-by-filename order the loader uses on the server).
+  const otherCategory = t(($) => $.create_dialog.template_picker.other_category);
+  const groups = useMemo(() => {
+    const byCategory = new Map<string, AgentTemplateSummary[]>();
+    for (const tmpl of templates) {
+      const key = tmpl.category?.trim() ? tmpl.category : otherCategory;
+      if (!byCategory.has(key)) byCategory.set(key, []);
+      byCategory.get(key)!.push(tmpl);
+    }
+    return Array.from(byCategory.entries());
+  }, [templates, otherCategory]);
+
+  if (isLoading) {
+    return (
+      <div className="flex flex-1 items-center justify-center">
+        <Loader2 className="h-5 w-5 animate-spin text-muted-foreground" />
+      </div>
+    );
+  }
+  if (error) {
+    return (
+      <div className="flex flex-1 items-center justify-center p-6">
+        <div className="text-sm text-destructive">
+          {error instanceof Error
+            ? error.message
+            : t(($) => $.create_dialog.template_picker.load_failed)}
+        </div>
+      </div>
+    );
+  }
+  if (templates.length === 0) {
+    return (
+      <div className="flex flex-1 items-center justify-center p-6">
+        <div className="text-sm text-muted-foreground">
+          {t(($) => $.create_dialog.template_picker.empty)}
+        </div>
+      </div>
+    );
+  }
+
+  return (
+    <div className="flex-1 overflow-y-auto">
+      <div className="mx-auto max-w-5xl space-y-6 p-6">
+        {groups.map(([category, tmpls]) => (
+          <section key={category}>
+            <h2 className="sticky top-0 z-10 -mx-6 border-b bg-background px-6 py-2 text-xs font-semibold uppercase tracking-wider text-muted-foreground">
+              {category}
+            </h2>
+            <div className="grid grid-cols-1 gap-3 pt-3 md:grid-cols-2">
+              {tmpls.map((tmpl) => (
+                <TemplateCard
+                  key={tmpl.slug}
+                  template={tmpl}
+                  onClick={() => onSelect(tmpl)}
+                />
+              ))}
+            </div>
+          </section>
+        ))}
+      </div>
+    </div>
+  );
+}
+
+interface TemplateCardProps {
+  template: AgentTemplateSummary;
+  onClick: () => void;
+}
+
+function TemplateCard({ template, onClick }: TemplateCardProps) {
+  const { t } = useT("agents");
+  const Icon = ICONS[template.icon ?? ""] ?? FileText;
+  const accentClass = ACCENTS[template.accent ?? ""] ?? ACCENTS.muted;
+
+  return (
+    <button
+      type="button"
+      onClick={onClick}
+      className="group flex items-start gap-3 rounded-lg border bg-card p-4 text-left transition-colors hover:border-primary/40 hover:bg-accent/40"
+    >
+      <div
+        className={cn(
+          "flex h-10 w-10 shrink-0 items-center justify-center rounded-lg",
+          accentClass,
+        )}
+      >
+        <Icon className="h-5 w-5" />
+      </div>
+      <div className="min-w-0 flex-1">
+        <div className="flex items-center gap-1">
+          <span className="truncate text-sm font-semibold">{template.name}</span>
+          <ChevronRight className="ml-auto h-4 w-4 shrink-0 text-muted-foreground/40 transition-transform group-hover:translate-x-0.5 group-hover:text-muted-foreground" />
+        </div>
+        <p className="mt-1 line-clamp-2 text-xs text-muted-foreground">
+          {template.description}
+        </p>
+        <div className="mt-2.5 inline-flex items-center gap-1 rounded-full bg-muted px-2 py-0.5 text-[10px] font-medium text-muted-foreground">
+          {t(($) => $.create_dialog.template_card.skills, {
+            count: template.skills.length,
+          })}
+        </div>
+      </div>
+    </button>
+  );
+}
+
+// --- Static maps so Tailwind's JIT scanner picks up every variant ---
+
+/** Lucide icon name → component. Add new entries when shipping templates
+ *  that use icons not yet listed here. Unknown names fall back to FileText. */
+const ICONS: Record<string, LucideIcon> = {
+  Search,
+  Palette,
+  FileText,
+  FlaskConical,
+  Sparkles,
+  ListChecks,
+  Brush,
+  PenLine,
+  Megaphone,
+  Presentation,
+  LayoutDashboard,
+};
+
+/** Semantic accent → Tailwind class string. The class strings are written
+ *  out verbatim so JIT scans them; dynamic `bg-${name}/10` would not be
+ *  generated. Mirrors the conventions in runtime-columns.tsx /
+ *  usage-section.tsx (existing uses of these tokens). */
+const DEFAULT_ACCENT = "bg-muted text-muted-foreground";
+
+const ACCENTS: Record<string, string> = {
+  info: "bg-info/10 text-info",
+  success: "bg-success/10 text-success",
+  warning: "bg-warning/10 text-warning",
+  primary: "bg-primary/10 text-primary",
+  secondary: "bg-secondary text-secondary-foreground",
+  muted: DEFAULT_ACCENT,
+};
+
+/** Exposed for the detail / form steps so they can render the same icon
+ *  badge as the picker card. Keeps visual continuity across steps. */
+export function getTemplateIcon(iconName: string | undefined): LucideIcon {
+  return ICONS[iconName ?? ""] ?? FileText;
+}
+
+export function getAccentClass(accent: string | undefined): string {
+  return ACCENTS[accent ?? ""] ?? DEFAULT_ACCENT;
+}

packages/views/chat/components/chat-window.tsx

@@ -3,7 +3,7 @@
 import React, { useCallback, useEffect, useMemo, useRef, useState } from "react";
 import { useQuery, useQueryClient } from "@tanstack/react-query";
 import { motion } from "motion/react";
-import { Minus, Maximize2, Minimize2, ChevronDown, ChevronRight, Plus, Check, Trash2 } from "lucide-react";
+import { Minus, Maximize2, Minimize2, ChevronDown, ChevronRight, Plus, Check, Trash2, Pencil } from "lucide-react";
 import { Button } from "@multica/ui/components/ui/button";
 import { Tooltip, TooltipTrigger, TooltipContent } from "@multica/ui/components/ui/tooltip";
 import {
@@ -46,6 +46,7 @@ import {
   useCreateChatSession,
   useDeleteChatSession,
   useMarkChatSessionRead,
+  useUpdateChatSession,
 } from "@multica/core/chat/mutations";
 import { useChatStore } from "@multica/core/chat";
 import { ChatMessageList, ChatMessageSkeleton } from "./chat-message-list";
@@ -210,6 +211,12 @@ export function ChatWindow() {
   // session-dropdown's existing localized `window.untitled` fallback kicks
   // in. A follow-up task may back-fill the real title from the first user
   // message — until then this keeps the session list scannable across locales.
+  //
+  // NOTE: ensureSession does NOT flip `activeSessionId` itself. Callers must
+  // seed `chatKeys.messages(sessionId)` in the Query cache BEFORE calling
+  // `setActiveSession(sessionId)`, otherwise the first useQuery subscription
+  // for the new key reports `isLoading: true` and renders ChatMessageSkeleton
+  // for one frame (the "new-chat first-message" white flash).
   const sessionPromiseRef = useRef<Promise<string | null> | null>(null);
   const ensureSession = useCallback(
     async (titleSeed: string): Promise<string | null> => {
@@ -223,7 +230,6 @@ export function ChatWindow() {
             agent_id: activeAgent.id,
             title: titleSeed.slice(0, 50),
           });
-          setActiveSession(session.id);
           return session.id;
         } finally {
           sessionPromiseRef.current = null;
@@ -232,16 +238,25 @@ export function ChatWindow() {
       sessionPromiseRef.current = promise;
       return promise;
     },
-    [activeSessionId, activeAgent, createSession, setActiveSession],
+    [activeSessionId, activeAgent, createSession],
   );
 
   const handleUploadFile = useCallback(
     async (file: File) => {
       const sessionId = await ensureSession("");
       if (!sessionId) return null;
+      // Prime the messages cache as empty before flipping activeSessionId so
+      // ChatMessageList mounts directly (no Skeleton frame). Skip the write
+      // when an entry already exists — a concurrent handleSend may have
+      // seeded an optimistic message we must not clobber.
+      qc.setQueryData<ChatMessage[]>(
+        chatKeys.messages(sessionId),
+        (old) => old ?? [],
+      );
+      setActiveSession(sessionId);
       return uploadWithToast(file, { chatSessionId: sessionId });
     },
-    [ensureSession, uploadWithToast],
+    [ensureSession, uploadWithToast, qc, setActiveSession],
   );
 
   const handleSend = useCallback(
@@ -287,6 +302,12 @@ export function ChatWindow() {
         task_id: null,
         created_at: sentAt,
       };
+      // Seed cache BEFORE flipping activeSessionId. If we set the active
+      // session first, useQuery's first subscription to the new key sees no
+      // cached data and renders ChatMessageSkeleton for one frame — the
+      // "new-chat first-message" white flash. Priming the cache first means
+      // the very first read after activeSessionId flips hits data
+      // synchronously and ChatMessageList mounts directly.
       qc.setQueryData<ChatMessage[]>(
         chatKeys.messages(sessionId),
         (old) => (old ? [...old, optimistic] : [optimistic]),
@@ -301,6 +322,9 @@ export function ChatWindow() {
         status: "queued",
         created_at: sentAt,
       });
+      // Cache primed → safe to publish the new active session. Idempotent
+      // when the session was already active (existing-conversation send).
+      setActiveSession(sessionId);
       apiLogger.debug("sendChatMessage.optimistic", { sessionId, optimisticId: optimistic.id });
 
       const result = await api.sendChatMessage(sessionId, finalContent, attachmentIds);
@@ -325,6 +349,7 @@ export function ChatWindow() {
       anchorCandidate,
       ensureSession,
       qc,
+      setActiveSession,
     ],
   );
 
@@ -710,7 +735,12 @@ function SessionDropdown({
 
   const [showArchived, setShowArchived] = useState(false);
   const [pendingDelete, setPendingDelete] = useState<ChatSession | null>(null);
+  // Inline rename: only one row can be in edit mode at a time. We track the
+  // session id (not the full session) so a stale closure can't overwrite a
+  // newer rename pulled in via WS.
+  const [renamingId, setRenamingId] = useState<string | null>(null);
   const deleteSession = useDeleteChatSession();
+  const updateSession = useUpdateChatSession();
   const setActiveSession = useChatStore((s) => s.setActiveSession);
   const formatTimeAgo = useFormatTimeAgo();
 
@@ -749,14 +779,35 @@ function SessionDropdown({
     });
   };
 
+  const handleSubmitRename = (sessionId: string, raw: string) => {
+    const trimmed = raw.trim();
+    const current = sessions.find((s) => s.id === sessionId);
+    setRenamingId(null);
+    // No-op submits (unchanged or blank) skip the network round-trip — the
+    // server would reject a blank title anyway, and an unchanged title would
+    // just bump updated_at for no user-visible reason.
+    if (!trimmed || trimmed === current?.title) return;
+    updateSession.mutate({ sessionId, title: trimmed });
+  };
+
   const renderRow = (session: ChatSession) => {
     const isCurrent = session.id === activeSessionId;
     const agent = agentById.get(session.agent_id) ?? null;
     const isRunning = inFlightSessionIds.has(session.id);
+    const isRenaming = renamingId === session.id;
     return (
       <DropdownMenuItem
         key={session.id}
-        onClick={() => onSelectSession(session)}
+        // While renaming we don't want a row click to select the session
+        // OR close the menu — the user is editing text, not navigating.
+        // closeOnClick=false keeps the dropdown open across input clicks
+        // / button clicks inside the row; the normal "click row → switch
+        // session → close menu" flow is unchanged when isRenaming=false.
+        closeOnClick={!isRenaming}
+        onClick={() => {
+          if (isRenaming) return;
+          onSelectSession(session);
+        }}
         className="group flex min-w-0 items-center gap-2"
       >
         {agent ? (
@@ -771,45 +822,84 @@ function SessionDropdown({
           <span className="size-6 shrink-0" />
         )}
         <div className="min-w-0 flex-1">
-          <div className="truncate text-sm">
-            {session.title?.trim() || t(($) => $.window.untitled)}
-          </div>
-          <div className="truncate text-xs text-muted-foreground/70">
-            {formatTimeAgo(session.updated_at)}
-          </div>
+          {isRenaming ? (
+            <SessionRenameInput
+              initialValue={session.title ?? ""}
+              onSubmit={(value) => handleSubmitRename(session.id, value)}
+              onCancel={() => setRenamingId(null)}
+            />
+          ) : (
+            <>
+              <div className="truncate text-sm">
+                {session.title?.trim() || t(($) => $.window.untitled)}
+              </div>
+              <div className="truncate text-xs text-muted-foreground/70">
+                {formatTimeAgo(session.updated_at)}
+              </div>
+            </>
+          )}
         </div>
         {/* Right-edge status pip: in-flight wins over unread because
          *  "still working" is more actionable than "has reply" — and
          *  the two rarely coexist in practice (the unread flag fires
          *  on chat_message write, by which point the task has just
          *  finished). Same pip shape as unread for visual rhythm,
-         *  amber + pulse to read as activity. */}
-        {isRunning ? (
+         *  amber + pulse to read as activity.
+         *
+         *  Hidden while renaming so the inline input has room to
+         *  breathe and trailing pips don't visually trail off-screen
+         *  next to the editor caret. */}
+        {!isRenaming && isRunning ? (
           <span
             aria-label={t(($) => $.window.running)}
             title={t(($) => $.window.running)}
             className="size-1.5 shrink-0 rounded-full bg-amber-500 animate-pulse"
           />
-        ) : session.has_unread ? (
+        ) : !isRenaming && session.has_unread ? (
           <span
             aria-label={t(($) => $.window.unread)}
             title={t(($) => $.window.unread)}
             className="size-1.5 shrink-0 rounded-full bg-brand"
           />
         ) : null}
-        {isCurrent && <Check className="size-3.5 text-muted-foreground shrink-0" />}
-        <button
-          type="button"
-          onClick={(e) => {
-            e.stopPropagation();
-            e.preventDefault();
-            setPendingDelete(session);
-          }}
-          className="shrink-0 rounded p-1 text-muted-foreground opacity-0 transition-opacity hover:bg-destructive/10 hover:text-destructive focus-visible:opacity-100 group-hover:opacity-100"
-          aria-label={t(($) => $.session_history.row_delete_aria)}
-        >
-          <Trash2 className="size-3.5" />
-        </button>
+        {!isRenaming && isCurrent && (
+          <Check className="size-3.5 text-muted-foreground shrink-0" />
+        )}
+        {!isRenaming && (
+          <>
+            <button
+              type="button"
+              // preventDefault is what tells Base UI's Menu.Item to skip
+              // its close-on-click; stopPropagation prevents the row's
+              // onClick from also firing (which would switch sessions).
+              // onPointerDown is stopped too so the menu's typeahead /
+              // focus tracking doesn't pre-empt the click.
+              onPointerDown={(e) => e.stopPropagation()}
+              onClick={(e) => {
+                e.stopPropagation();
+                e.preventDefault();
+                setRenamingId(session.id);
+              }}
+              className="shrink-0 rounded p-1 text-muted-foreground opacity-0 transition-opacity hover:bg-accent hover:text-foreground focus-visible:opacity-100 group-hover:opacity-100"
+              aria-label={t(($) => $.session_history.row_rename_aria)}
+              title={t(($) => $.session_history.row_rename_aria)}
+            >
+              <Pencil className="size-3.5" />
+            </button>
+            <button
+              type="button"
+              onClick={(e) => {
+                e.stopPropagation();
+                e.preventDefault();
+                setPendingDelete(session);
+              }}
+              className="shrink-0 rounded p-1 text-muted-foreground opacity-0 transition-opacity hover:bg-destructive/10 hover:text-destructive focus-visible:opacity-100 group-hover:opacity-100"
+              aria-label={t(($) => $.session_history.row_delete_aria)}
+            >
+              <Trash2 className="size-3.5" />
+            </button>
+          </>
+        )}
       </DropdownMenuItem>
     );
   };
@@ -926,6 +1016,86 @@ function SessionDropdown({
   );
 }
 
+/**
+ * Inline editor for a session title. Mounts focused with the existing
+ * title pre-selected so the user can either replace it outright or arrow
+ * into the existing text. Enter commits, Escape cancels, a real click
+ * outside the input also commits.
+ *
+ * We do NOT commit on the input's `blur` event: Base UI's Menu uses
+ * focus-follows-cursor (hovering a sibling row drags DOM focus there),
+ * so a blur handler would fire on every mouse-move and "save" the user's
+ * half-typed title without them clicking anywhere. Instead a document-
+ * level `pointerdown` listener — registered in capture phase so it runs
+ * before Base UI's outside-click close handler — commits when the user
+ * actually clicks outside the input.
+ */
+function SessionRenameInput({
+  initialValue,
+  onSubmit,
+  onCancel,
+}: {
+  initialValue: string;
+  onSubmit: (value: string) => void;
+  onCancel: () => void;
+}) {
+  const { t } = useT("chat");
+  const [value, setValue] = useState(initialValue);
+  const inputRef = useRef<HTMLInputElement>(null);
+  // Hold the latest value + callback in refs so the mount-only effect's
+  // listener always sees fresh state without re-subscribing on every
+  // keystroke (which would briefly leave a window where pointerdown isn't
+  // observed).
+  const valueRef = useRef(value);
+  valueRef.current = value;
+  const onSubmitRef = useRef(onSubmit);
+  onSubmitRef.current = onSubmit;
+
+  useEffect(() => {
+    inputRef.current?.focus();
+    inputRef.current?.select();
+
+    const handlePointerDown = (e: PointerEvent) => {
+      const input = inputRef.current;
+      if (!input) return;
+      if (input.contains(e.target as Node)) return;
+      onSubmitRef.current(valueRef.current);
+    };
+    // Capture phase — Base UI registers its own outside-click handler in
+    // bubble; running first lets us commit before the menu starts to
+    // close (and unmount this component).
+    document.addEventListener("pointerdown", handlePointerDown, true);
+    return () => {
+      document.removeEventListener("pointerdown", handlePointerDown, true);
+    };
+  }, []);
+
+  return (
+    <input
+      ref={inputRef}
+      type="text"
+      value={value}
+      maxLength={200}
+      aria-label={t(($) => $.session_history.row_rename_aria)}
+      onChange={(e) => setValue(e.target.value)}
+      onClick={(e) => e.stopPropagation()}
+      onPointerDown={(e) => e.stopPropagation()}
+      onKeyDown={(e) => {
+        // Stop the menu from stealing arrow / typeahead / space input.
+        e.stopPropagation();
+        if (e.key === "Enter") {
+          e.preventDefault();
+          onSubmit(value);
+        } else if (e.key === "Escape") {
+          e.preventDefault();
+          onCancel();
+        }
+      }}
+      className="w-full rounded-sm bg-background px-1 py-0.5 text-sm outline-none ring-1 ring-border focus-visible:ring-brand"
+    />
+  );
+}
+
 function useFormatTimeAgo(): (dateStr: string) => string {
   const { t } = useT("chat");
   return (dateStr: string) => {

packages/views/common/actor-avatar.tsx

@@ -53,6 +53,8 @@ export function ActorAvatar({
       initials={getActorInitials(actorType, actorId)}
       avatarUrl={getActorAvatarUrl(actorType, actorId)}
       isAgent={actorType === "agent"}
+      isSystem={actorType === "system"}
+      isSquad={actorType === "squad"}
       size={size}
       className={className}
     />

packages/views/common/timezone-select.tsx

@@ -0,0 +1,112 @@
+"use client";
+
+import {
+  Select,
+  SelectContent,
+  SelectItem,
+  SelectTrigger,
+  SelectValue,
+} from "@multica/ui/components/ui/select";
+
+// Common IANA zones surfaced as quick picks. Used as the fallback option set
+// when Intl.supportedValuesOf is not available, and promoted to the top of
+// the list when it is.
+const COMMON_TIMEZONES = [
+  "UTC",
+  "America/Los_Angeles",
+  "America/Denver",
+  "America/Chicago",
+  "America/New_York",
+  "America/Sao_Paulo",
+  "Europe/London",
+  "Europe/Berlin",
+  "Europe/Paris",
+  "Europe/Moscow",
+  "Africa/Cairo",
+  "Asia/Dubai",
+  "Asia/Kolkata",
+  "Asia/Bangkok",
+  "Asia/Shanghai",
+  "Asia/Singapore",
+  "Asia/Tokyo",
+  "Australia/Sydney",
+  "Pacific/Auckland",
+];
+
+export function browserTimezone(): string {
+  try {
+    const tz = Intl.DateTimeFormat().resolvedOptions().timeZone;
+    return tz || "UTC";
+  } catch {
+    return "UTC";
+  }
+}
+
+type IntlWithSupportedValues = typeof Intl & {
+  supportedValuesOf?: (key: "timeZone") => string[];
+};
+
+function supportedTimezones(): string[] {
+  try {
+    const supported = (Intl as IntlWithSupportedValues).supportedValuesOf?.(
+      "timeZone",
+    );
+    return supported && supported.length > 0 ? supported : COMMON_TIMEZONES;
+  } catch {
+    return COMMON_TIMEZONES;
+  }
+}
+
+export function timezoneOptions(current: string): string[] {
+  const browser = browserTimezone();
+  return Array.from(
+    new Set([current, browser, ...COMMON_TIMEZONES, ...supportedTimezones()]),
+  ).filter(Boolean);
+}
+
+// Shared single-select timezone picker. Surfaces the browser-resolved zone
+// with a translated suffix (passed in by the caller — the picker itself stays
+// i18n-namespace agnostic), followed by a curated set of common IANA zones
+// and everything Intl.supportedValuesOf exposes.
+export function TimezoneSelect({
+  value,
+  onValueChange,
+  browserSuffix,
+  disabled,
+  triggerClassName,
+}: {
+  value: string;
+  onValueChange: (next: string) => void;
+  browserSuffix: string;
+  disabled?: boolean;
+  triggerClassName?: string;
+}) {
+  const browser = browserTimezone();
+  const options = timezoneOptions(value);
+  const render = (tz: string) =>
+    tz === browser ? `${tz}${browserSuffix}` : tz;
+
+  return (
+    <Select
+      value={value}
+      disabled={disabled}
+      onValueChange={(next) => {
+        if (next) onValueChange(next);
+      }}
+    >
+      <SelectTrigger
+        size="sm"
+        className={triggerClassName ?? "w-full rounded-md font-mono text-xs"}
+      >
+        <SelectValue>{render(value)}</SelectValue>
+      </SelectTrigger>
+      <SelectContent align="start" className="max-h-72">
+        {options.map((tz) => (
+          <SelectItem key={tz} value={tz} className="font-mono text-xs">
+            {render(tz)}
+          </SelectItem>
+        ))}
+      </SelectContent>
+    </Select>
+  );
+}

packages/views/dashboard/components/dashboard-page.tsx

@@ -0,0 +1,558 @@
+"use client";
+
+import { useMemo, useState } from "react";
+import { BarChart3, FolderKanban } from "lucide-react";
+import { useQuery } from "@tanstack/react-query";
+import { Skeleton } from "@multica/ui/components/ui/skeleton";
+import {
+  Select,
+  SelectContent,
+  SelectItem,
+  SelectTrigger,
+  SelectValue,
+} from "@multica/ui/components/ui/select";
+import { useWorkspaceId } from "@multica/core/hooks";
+import { agentListOptions } from "@multica/core/workspace/queries";
+import { projectListOptions } from "@multica/core/projects/queries";
+import {
+  dashboardUsageDailyOptions,
+  dashboardUsageByAgentOptions,
+  dashboardAgentRunTimeOptions,
+} from "@multica/core/dashboard";
+import { useCustomPricingStore } from "@multica/core/runtimes/custom-pricing-store";
+import { PageHeader } from "../../layout/page-header";
+import { KpiCard } from "../../runtimes/components/shared";
+import { DailyCostChart, DailyTokensChart } from "../../runtimes/components/charts";
+import { ProjectIcon } from "../../projects/components/project-icon";
+import { ActorAvatar } from "../../common/actor-avatar";
+import {
+  TimezoneSelect,
+  browserTimezone,
+} from "../../common/timezone-select";
+import { formatTokens } from "../../runtimes/utils";
+import { useT } from "../../i18n";
+import {
+  aggregateAgentTokens,
+  aggregateDailyCost,
+  aggregateDailyTokens,
+  computeDailyTotals,
+  formatDuration,
+  mergeAgentDashboardRows,
+  type AgentDashboardRow,
+} from "../utils";
+
+// One-place source of truth for the period selector. Matches the runtime
+// detail page so users see the same three options across the dashboards.
+const TIME_RANGES = [
+  { label: "7d", days: 7 },
+  { label: "30d", days: 30 },
+  { label: "90d", days: 90 },
+] as const;
+type TimeRange = (typeof TIME_RANGES)[number]["days"];
+
+// Sentinel for "no project filter" — kept distinct from the empty string
+// so it survives a refactor that ever lets a project be slug-keyed.
+const ALL_PROJECTS = "__all__";
+
+// Stable references — `data ?? []` would create a new empty array on
+// every render while the query is loading, which breaks useMemo's
+// reference-equality dep check and trips the exhaustive-deps lint rule.
+const EMPTY_DAILY: import("@multica/core/types").DashboardUsageDaily[] = [];
+const EMPTY_BY_AGENT: import("@multica/core/types").DashboardUsageByAgent[] = [];
+const EMPTY_RUNTIME: import("@multica/core/types").DashboardAgentRunTime[] = [];
+
+function fmtMoney(n: number): string {
+  if (n >= 100) return `$${n.toFixed(0)}`;
+  return `$${n.toFixed(2)}`;
+}
+
+// Local segmented control — same visual language the runtime usage section
+// uses for its period / tab toggles. shadcn's Tabs is wired for full tab
+// pages with ARIA semantics the compact toolbar pill doesn't need.
+function Segmented<T extends string | number>({
+  value,
+  onChange,
+  options,
+}: {
+  value: T;
+  onChange: (v: T) => void;
+  options: readonly { label: string; value: T }[];
+}) {
+  return (
+    <div className="inline-flex items-center gap-0.5 rounded-md bg-muted p-0.5">
+      {options.map((o) => (
+        <button
+          key={String(o.value)}
+          type="button"
+          onClick={() => onChange(o.value)}
+          className={`rounded-sm px-2.5 py-1 text-xs font-medium transition-colors ${
+            o.value === value
+              ? "bg-background text-foreground shadow-sm"
+              : "text-muted-foreground hover:text-foreground"
+          }`}
+        >
+          {o.label}
+        </button>
+      ))}
+    </div>
+  );
+}
+
+/**
+ * Workspace + project token / run-time dashboard.
+ *
+ * Lives at `/{slug}/dashboard`. Three independent rollups (daily cost,
+ * per-agent tokens, per-agent run-time) feed four KPI tiles, a daily cost
+ * chart, and a combined "by agent" list. A project dropdown narrows every
+ * query to one project; the period selector applies to all three.
+ *
+ * Cost math runs client-side via the runtimes utils — keeps the dashboard
+ * and the runtime page using one pricing table.
+ */
+export function DashboardPage() {
+  const { t } = useT("usage");
+  const { t: tRuntimes } = useT("runtimes");
+  const wsId = useWorkspaceId();
+  const [days, setDays] = useState<TimeRange>(30);
+  const [projectValue, setProjectValue] = useState<string>(ALL_PROJECTS);
+  // Default to the browser's resolved zone so day-boundary buckets match the
+  // user's local clock on first render. Pure client-state — the rollup queries
+  // are zone-agnostic today; this is the UI affordance the user can pin.
+  const [timezone, setTimezone] = useState<string>(() => browserTimezone());
+
+  // The user can save model prices from the runtimes page; re-render when
+  // they do so the dashboard reflects the new rates.
+  useCustomPricingStore((s) => s.pricings);
+
+  const { data: projects = [] } = useQuery(projectListOptions(wsId));
+  const { data: agents = [] } = useQuery(agentListOptions(wsId));
+
+  // Validate the picked project against the current workspace's list. A
+  // stale UUID — left over from a project that's been deleted, or from the
+  // previous workspace after a switch — would silently filter all three
+  // queries to empty rows while the dropdown still reads "All projects".
+  // Derive the effective filter so the API call matches the user-visible
+  // selection.
+  const projectId = useMemo(() => {
+    if (projectValue === ALL_PROJECTS) return null;
+    return projects.some((p) => p.id === projectValue) ? projectValue : null;
+  }, [projectValue, projects]);
+
+  const dailyQuery = useQuery(dashboardUsageDailyOptions(wsId, days, projectId));
+  const byAgentQuery = useQuery(dashboardUsageByAgentOptions(wsId, days, projectId));
+  const runTimeQuery = useQuery(dashboardAgentRunTimeOptions(wsId, days, projectId));
+
+  const dailyUsage = dailyQuery.data ?? EMPTY_DAILY;
+  const byAgentUsage = byAgentQuery.data ?? EMPTY_BY_AGENT;
+  const runTimeRows = runTimeQuery.data ?? EMPTY_RUNTIME;
+
+  const isLoading =
+    dailyQuery.isLoading || byAgentQuery.isLoading || runTimeQuery.isLoading;
+
+  // Three independent rollups, but the empty-state is one decision — only
+  // show "no data yet" when ALL three came back empty so a project with
+  // tokens but no runs doesn't look broken.
+  const hasNoData =
+    !isLoading &&
+    dailyUsage.length === 0 &&
+    byAgentUsage.length === 0 &&
+    runTimeRows.length === 0;
+
+  // Cost / token math — re-derived when usage, days, or pricings change.
+  const totals = useMemo(() => computeDailyTotals(dailyUsage), [dailyUsage]);
+  const dailyCost = useMemo(() => aggregateDailyCost(dailyUsage), [dailyUsage]);
+  const dailyTokens = useMemo(() => aggregateDailyTokens(dailyUsage), [dailyUsage]);
+  const agentTokenRows = useMemo(
+    () => aggregateAgentTokens(byAgentUsage),
+    [byAgentUsage],
+  );
+
+  // Run-time totals — taskCount + failedCount summed for the KPI row.
+  const runTimeTotals = useMemo(() => {
+    let totalSeconds = 0;
+    let taskCount = 0;
+    let failedCount = 0;
+    for (const r of runTimeRows) {
+      totalSeconds += r.total_seconds;
+      taskCount += r.task_count;
+      failedCount += r.failed_count;
+    }
+    return { totalSeconds, taskCount, failedCount };
+  }, [runTimeRows]);
+
+  const agentRows = useMemo(
+    () => mergeAgentDashboardRows(agentTokenRows, runTimeRows),
+    [agentTokenRows, runTimeRows],
+  );
+
+  return (
+    <div className="flex h-full flex-col">
+      {/* h-auto + min-h-12 + flex-wrap: the toolbar (project filter, range
+          switch, timezone select) overflows the single h-12 row on narrow
+          and medium widths once the timezone picker is added — letting the
+          right cluster wrap underneath keeps every control reachable
+          without an off-screen bleed. Wider viewports still render the
+          original single row. */}
+      <PageHeader className="h-auto min-h-12 flex-wrap justify-between gap-y-1.5 px-5 py-1.5 sm:py-0">
+        <div className="flex min-w-0 items-center gap-2">
+          <BarChart3 className="h-4 w-4 shrink-0 text-muted-foreground" />
+          <h1 className="truncate text-sm font-medium">{t(($) => $.title)}</h1>
+        </div>
+        <div className="flex flex-wrap items-center gap-2">
+          <ProjectFilter
+            projects={projects}
+            value={projectValue}
+            onChange={setProjectValue}
+          />
+          <Segmented
+            value={days}
+            onChange={setDays}
+            options={TIME_RANGES.map((r) => ({ label: r.label, value: r.days }))}
+          />
+          <TimezoneSelect
+            value={timezone}
+            onValueChange={setTimezone}
+            browserSuffix={tRuntimes(($) => $.detail.timezone_browser_suffix)}
+            triggerClassName="rounded-md font-mono text-xs"
+          />
+        </div>
+      </PageHeader>
+
+      <div className="flex-1 overflow-y-auto">
+        <div className="mx-auto max-w-6xl space-y-5 p-6">
+          <p className="text-xs text-muted-foreground">{t(($) => $.subtitle)}</p>
+
+          {isLoading ? (
+            <DashboardSkeleton />
+          ) : hasNoData ? (
+            <DashboardEmpty />
+          ) : (
+            <>
+              {/* KPI row — same 3-divide-x card grid the runtime usage
+                  section uses, expanded to four tiles. */}
+              <div className="grid grid-cols-1 divide-y rounded-lg border bg-card sm:grid-cols-2 sm:divide-x sm:divide-y-0 lg:grid-cols-4">
+                <KpiCard
+                  label={t(($) => $.kpi.cost_label, { days })}
+                  value={fmtMoney(totals.cost)}
+                />
+                <KpiCard
+                  label={t(($) => $.kpi.tokens_label, { days })}
+                  value={formatTokens(
+                    totals.input + totals.output + totals.cacheRead + totals.cacheWrite,
+                  )}
+                  hint={t(($) => $.kpi.tokens_hint, {
+                    input: formatTokens(totals.input),
+                    output: formatTokens(totals.output),
+                  })}
+                />
+                <KpiCard
+                  label={t(($) => $.kpi.run_time_label, { days })}
+                  value={formatDuration(
+                    runTimeTotals.totalSeconds,
+                    t(($) => $.duration.less_than_minute),
+                  )}
+                  hint={t(($) => $.kpi.run_time_hint, {
+                    tasks: runTimeTotals.taskCount,
+                  })}
+                />
+                <KpiCard
+                  label={t(($) => $.kpi.tasks_label, { days })}
+                  value={String(runTimeTotals.taskCount)}
+                  hint={t(($) => $.kpi.tasks_hint, {
+                    failed: runTimeTotals.failedCount,
+                  })}
+                  accent={runTimeTotals.failedCount > 0 ? "default" : "default"}
+                />
+              </div>
+
+              {/* Daily trend chart — toggle picks Cost vs Tokens axis,
+                  mirroring the runtime-detail Usage section so both
+                  surfaces share one chart language. */}
+              <DailyTrendBlock dailyCost={dailyCost} dailyTokens={dailyTokens} />
+
+              {/* Per-agent leaderboard — user picks the ranking metric;
+                  the progress bar and column emphasis follow the metric. */}
+              <Leaderboard
+                rows={agentRows}
+                agents={agents}
+                lessThanMinuteLabel={t(($) => $.duration.less_than_minute)}
+              />
+            </>
+          )}
+        </div>
+      </div>
+    </div>
+  );
+}
+
+function ProjectFilter({
+  projects,
+  value,
+  onChange,
+}: {
+  projects: { id: string; title: string; icon: string | null }[];
+  value: string;
+  onChange: (v: string) => void;
+}) {
+  const { t } = useT("usage");
+  const allLabel = t(($) => $.filter.all_projects);
+  const selected = projects.find((p) => p.id === value);
+  const selectedTitle =
+    value === ALL_PROJECTS ? allLabel : selected?.title ?? allLabel;
+
+  return (
+    <Select
+      value={value}
+      onValueChange={(v) => onChange(v ?? ALL_PROJECTS)}
+    >
+      <SelectTrigger size="sm" className="min-w-[180px]">
+        <SelectValue>
+          {() => (
+            <>
+              {selected ? (
+                <ProjectIcon project={selected} size="sm" />
+              ) : (
+                <FolderKanban className="h-3.5 w-3.5 shrink-0 text-muted-foreground" />
+              )}
+              <span className="truncate">{selectedTitle}</span>
+            </>
+          )}
+        </SelectValue>
+      </SelectTrigger>
+      {/* alignItemWithTrigger=false: the default aligns the *selected* item
+          to the trigger, which pushes "All projects" above the trigger and
+          clips it off-screen when the usage header sits at the top of the
+          viewport. Anchor the dropdown to the bottom of the trigger so
+          every entry stays reachable.
+          max-h-72: cap the dropdown so a long project list scrolls instead
+          of stretching to the bottom of the window. */}
+      <SelectContent align="start" alignItemWithTrigger={false} className="max-h-72">
+        <SelectItem value={ALL_PROJECTS}>
+          <FolderKanban className="h-3.5 w-3.5 shrink-0 text-muted-foreground" />
+          <span className="truncate">{allLabel}</span>
+        </SelectItem>
+        {projects.map((p) => (
+          <SelectItem key={p.id} value={p.id}>
+            <ProjectIcon project={p} size="sm" />
+            <span className="truncate">{p.title}</span>
+          </SelectItem>
+        ))}
+      </SelectContent>
+    </Select>
+  );
+}
+
+type DailyMetric = "cost" | "tokens";
+
+function DailyTrendBlock({
+  dailyCost,
+  dailyTokens,
+}: {
+  dailyCost: ReturnType<typeof aggregateDailyCost>;
+  dailyTokens: ReturnType<typeof aggregateDailyTokens>;
+}) {
+  const { t } = useT("usage");
+  const [metric, setMetric] = useState<DailyMetric>("tokens");
+
+  // Empty-state is per-metric so a workspace that recorded tokens but
+  // has no priced models (unmapped) still gets a real Tokens chart while
+  // its Cost view falls through to the empty-state — same convention as
+  // the runtimes-side DailyTab in usage-section.tsx.
+  const totalCost = dailyCost.reduce((sum, d) => sum + d.total, 0);
+  const totalTokens = dailyTokens.reduce(
+    (sum, d) => sum + d.input + d.output + d.cacheRead + d.cacheWrite,
+    0,
+  );
+  const isEmpty = metric === "cost" ? totalCost === 0 : totalTokens === 0;
+
+  return (
+    <div className="rounded-lg border bg-card p-4">
+      <div className="mb-3 flex flex-wrap items-center justify-between gap-3">
+        <h4 className="text-sm font-semibold">
+          {metric === "cost"
+            ? t(($) => $.daily.title_cost)
+            : t(($) => $.daily.title_tokens)}
+        </h4>
+        <Segmented
+          value={metric}
+          onChange={setMetric}
+          options={[
+            { label: t(($) => $.daily.metric_tokens), value: "tokens" as const },
+            { label: t(($) => $.daily.metric_cost), value: "cost" as const },
+          ]}
+        />
+      </div>
+      <div className="min-h-[240px]">
+        {isEmpty ? (
+          <div className="flex aspect-[3/1] flex-col items-center justify-center gap-2 rounded-md border border-dashed bg-muted/20 p-6 text-center">
+            <BarChart3 className="h-5 w-5 text-muted-foreground/50" />
+            <p className="text-xs text-muted-foreground">
+              {t(($) => $.daily.no_data)}
+            </p>
+          </div>
+        ) : metric === "cost" ? (
+          <DailyCostChart data={dailyCost} />
+        ) : (
+          <DailyTokensChart data={dailyTokens} />
+        )}
+      </div>
+    </div>
+  );
+}
+
+// Which metric ranks the leaderboard. Drives row order, progress bar
+// width, and which column header is emphasised — keeping the three in
+// lockstep so the user always sees what the ranking actually measures.
+type LeaderboardSort = "tokens" | "cost" | "time" | "tasks";
+
+const SORT_METRIC: Record<LeaderboardSort, (r: AgentDashboardRow) => number> = {
+  tokens: (r) => r.tokens,
+  cost: (r) => r.cost,
+  time: (r) => r.seconds,
+  tasks: (r) => r.taskCount,
+};
+
+function Leaderboard({
+  rows,
+  agents,
+  lessThanMinuteLabel,
+}: {
+  rows: AgentDashboardRow[];
+  agents: { id: string; name: string }[];
+  lessThanMinuteLabel: string;
+}) {
+  const { t } = useT("usage");
+  const [sortBy, setSortBy] = useState<LeaderboardSort>("tokens");
+
+  const sortOptions = useMemo(
+    () => [
+      { value: "tokens" as const, label: t(($) => $.leaderboard.header_tokens) },
+      { value: "cost" as const, label: t(($) => $.leaderboard.header_cost) },
+      { value: "time" as const, label: t(($) => $.leaderboard.header_time) },
+      { value: "tasks" as const, label: t(($) => $.leaderboard.header_tasks) },
+    ],
+    [t],
+  );
+
+  // Re-rank when the metric changes; keep the merged input untouched so
+  // upstream `mergeAgentDashboardRows`'s tiebreaker (run time desc) still
+  // applies inside an equal-bucket.
+  const sortedRows = useMemo(() => {
+    const metric = SORT_METRIC[sortBy];
+    return [...rows].sort((a, b) => metric(b) - metric(a));
+  }, [rows, sortBy]);
+
+  const maxValue = useMemo(() => {
+    const metric = SORT_METRIC[sortBy];
+    return sortedRows.reduce((m, r) => Math.max(m, metric(r)), 0);
+  }, [sortedRows, sortBy]);
+
+  // Active column gets foreground text; others stay muted. Helps the user
+  // see "this is what the bar is measuring" at a glance.
+  const colClass = (key: LeaderboardSort) =>
+    `text-right ${sortBy === key ? "text-foreground" : "text-muted-foreground"}`;
+
+  return (
+    <div className="rounded-lg border bg-card">
+      <div className="flex flex-wrap items-center justify-between gap-3 border-b px-4 pt-4 pb-3">
+        <h4 className="text-sm font-semibold">{t(($) => $.leaderboard.title)}</h4>
+        <div className="flex items-center gap-3">
+          <Segmented value={sortBy} onChange={setSortBy} options={sortOptions} />
+          <span className="text-xs text-muted-foreground">
+            {t(($) => $.leaderboard.caption, { count: rows.length })}
+          </span>
+        </div>
+      </div>
+      {sortedRows.length === 0 ? (
+        <p className="px-4 py-8 text-center text-xs text-muted-foreground">
+          {t(($) => $.leaderboard.no_data)}
+        </p>
+      ) : (
+        <>
+          <div className="grid grid-cols-[minmax(0,1.6fr)_minmax(0,1fr)_5rem_5rem_5rem_4rem] items-center gap-3 border-b px-4 py-2 text-xs font-medium text-muted-foreground">
+            <span>{t(($) => $.leaderboard.header_agent)}</span>
+            <span />
+            <span className={colClass("tokens")}>{t(($) => $.leaderboard.header_tokens)}</span>
+            <span className={colClass("cost")}>{t(($) => $.leaderboard.header_cost)}</span>
+            <span className={colClass("time")}>{t(($) => $.leaderboard.header_time)}</span>
+            <span className={colClass("tasks")}>{t(($) => $.leaderboard.header_tasks)}</span>
+          </div>
+          <div className="divide-y">
+            {sortedRows.map((row) => {
+              const agent = agents.find((a) => a.id === row.agentId);
+              const value = SORT_METRIC[sortBy](row);
+              const pct = maxValue > 0 ? (value / maxValue) * 100 : 0;
+              return (
+                <div
+                  key={row.agentId}
+                  className="grid grid-cols-[minmax(0,1.6fr)_minmax(0,1fr)_5rem_5rem_5rem_4rem] items-center gap-3 px-4 py-2"
+                >
+                  <div className="flex min-w-0 items-center gap-2">
+                    <ActorAvatar
+                      actorType="agent"
+                      actorId={row.agentId}
+                      size={22}
+                      enableHoverCard
+                    />
+                    <span className="cursor-pointer truncate text-sm font-medium">
+                      {agent?.name ?? row.agentId}
+                    </span>
+                  </div>
+                  <div className="relative h-2 overflow-hidden rounded-full bg-muted">
+                    <div
+                      className="h-full rounded-full bg-chart-1 transition-[width] duration-300 ease-out"
+                      style={{ width: `${pct}%` }}
+                    />
+                  </div>
+                  <div
+                    className={`text-right text-xs tabular-nums ${sortBy === "tokens" ? "font-medium text-foreground" : "text-muted-foreground"}`}
+                  >
+                    {formatTokens(row.tokens)}
+                  </div>
+                  <div
+                    className={`text-right tabular-nums ${sortBy === "cost" ? "text-sm font-medium" : "text-xs text-muted-foreground"}`}
+                  >
+                    ${row.cost.toFixed(2)}
+                  </div>
+                  <div
+                    className={`text-right text-xs tabular-nums ${sortBy === "time" ? "font-medium text-foreground" : "text-muted-foreground"}`}
+                  >
+                    {formatDuration(row.seconds, lessThanMinuteLabel)}
+                  </div>
+                  <div
+                    className={`text-right text-xs tabular-nums ${sortBy === "tasks" ? "font-medium text-foreground" : "text-muted-foreground"}`}
+                  >
+                    {row.taskCount}
+                  </div>
+                </div>
+              );
+            })}
+          </div>
+        </>
+      )}
+    </div>
+  );
+}
+
+function DashboardSkeleton() {
+  return (
+    <div className="space-y-5">
+      <Skeleton className="h-28 rounded-lg" />
+      <Skeleton className="h-56 rounded-lg" />
+      <Skeleton className="h-48 rounded-lg" />
+    </div>
+  );
+}
+
+function DashboardEmpty() {
+  const { t } = useT("usage");
+  return (
+    <div className="flex flex-col items-center rounded-lg border border-dashed py-12 text-center">
+      <BarChart3 className="h-6 w-6 text-muted-foreground/40" />
+      <p className="mt-3 text-sm font-medium">{t(($) => $.empty.title)}</p>
+      <p className="mt-1 max-w-md text-xs text-muted-foreground">
+        {t(($) => $.empty.body)}
+      </p>
+    </div>
+  );
+}

packages/views/dashboard/index.ts

@@ -0,0 +1 @@
+export { DashboardPage } from "./components/dashboard-page";

packages/views/dashboard/utils.test.ts

@@ -0,0 +1,213 @@
+import { describe, expect, it } from "vitest";
+import {
+  aggregateAgentTokens,
+  aggregateDailyCost,
+  computeDailyTotals,
+  formatDuration,
+  mergeAgentDashboardRows,
+} from "./utils";
+
+describe("aggregateDailyCost", () => {
+  it("collapses multiple rows per day into one stack and sorts by date asc", () => {
+    const result = aggregateDailyCost([
+      {
+        date: "2026-05-10",
+        model: "claude-sonnet-4-6",
+        input_tokens: 1_000_000,
+        output_tokens: 500_000,
+        cache_read_tokens: 0,
+        cache_write_tokens: 0,
+        task_count: 3,
+      },
+      {
+        date: "2026-05-09",
+        model: "claude-sonnet-4-6",
+        input_tokens: 1_000_000,
+        output_tokens: 0,
+        cache_read_tokens: 0,
+        cache_write_tokens: 0,
+        task_count: 1,
+      },
+    ]);
+
+    // Sort: oldest day first.
+    expect(result.map((r) => r.date)).toEqual(["2026-05-09", "2026-05-10"]);
+    // claude-sonnet-4-6: input $3/M, output $15/M.
+    // 2026-05-09 → 1M input × $3 = $3 input, $0 output, $0 cache.
+    expect(result[0]).toMatchObject({ input: 3, output: 0, cacheWrite: 0, total: 3 });
+    // 2026-05-10 → $3 input + (0.5M × $15) = $7.5 output. Total $10.5.
+    expect(result[1]).toMatchObject({ input: 3, output: 7.5, cacheWrite: 0, total: 10.5 });
+  });
+
+  it("treats unmapped models as zero-cost", () => {
+    const result = aggregateDailyCost([
+      {
+        date: "2026-05-10",
+        model: "made-up-model",
+        input_tokens: 999_999_999,
+        output_tokens: 0,
+        cache_read_tokens: 0,
+        cache_write_tokens: 0,
+        task_count: 0,
+      },
+    ]);
+    expect(result[0]?.total).toBe(0);
+  });
+});
+
+describe("aggregateAgentTokens", () => {
+  it("folds per-(agent, model) rows into per-agent totals and sorts by cost desc", () => {
+    const rows = aggregateAgentTokens([
+      {
+        agent_id: "small-spender",
+        model: "claude-sonnet-4-6",
+        input_tokens: 100_000,
+        output_tokens: 0,
+        cache_read_tokens: 0,
+        cache_write_tokens: 0,
+        task_count: 1,
+      },
+      {
+        agent_id: "big-spender",
+        model: "claude-sonnet-4-6",
+        input_tokens: 5_000_000,
+        output_tokens: 0,
+        cache_read_tokens: 0,
+        cache_write_tokens: 0,
+        task_count: 3,
+      },
+      {
+        agent_id: "big-spender",
+        model: "claude-haiku-4-5",
+        input_tokens: 1_000_000,
+        output_tokens: 0,
+        cache_read_tokens: 0,
+        cache_write_tokens: 0,
+        task_count: 2,
+      },
+    ]);
+
+    expect(rows.map((r) => r.agentId)).toEqual(["big-spender", "small-spender"]);
+    expect(rows[0]?.taskCount).toBe(5);
+    // big-spender across two models — verify cost > small-spender's.
+    expect(rows[0]!.cost).toBeGreaterThan(rows[1]!.cost);
+  });
+});
+
+describe("computeDailyTotals", () => {
+  it("sums tokens across rows and adds estimated cost", () => {
+    const totals = computeDailyTotals([
+      {
+        date: "2026-05-10",
+        model: "claude-sonnet-4-6",
+        input_tokens: 1_000_000,
+        output_tokens: 0,
+        cache_read_tokens: 0,
+        cache_write_tokens: 0,
+        task_count: 2,
+      },
+      {
+        date: "2026-05-09",
+        model: "claude-sonnet-4-6",
+        input_tokens: 2_000_000,
+        output_tokens: 0,
+        cache_read_tokens: 0,
+        cache_write_tokens: 0,
+        task_count: 3,
+      },
+    ]);
+    expect(totals.input).toBe(3_000_000);
+    expect(totals.cost).toBe(9); // 3M × $3/M
+    expect(totals.taskCount).toBe(5);
+  });
+});
+
+describe("mergeAgentDashboardRows", () => {
+  it("uses run-time rollup's per-agent task count, not the token sum", () => {
+    // Token rollup returns two (agent, model) rows for the same task
+    // (the agent ran one task that touched two models). The token-side
+    // aggregator sums per-row task_count and lands at 2; the run-time
+    // rollup correctly reports the underlying distinct count of 1.
+    const tokenRows = [
+      {
+        agentId: "agent-a",
+        tokens: 3_000_000,
+        cost: 12,
+        taskCount: 2, // overcounted because (model-1: 1) + (model-2: 1)
+      },
+    ];
+    const runTimeRows = [
+      {
+        agent_id: "agent-a",
+        total_seconds: 600,
+        task_count: 1, // truth: one task touched both models
+        failed_count: 0,
+      },
+    ];
+    const merged = mergeAgentDashboardRows(tokenRows, runTimeRows);
+    expect(merged).toHaveLength(1);
+    expect(merged[0]!.taskCount).toBe(1);
+    expect(merged[0]!.seconds).toBe(600);
+  });
+
+  it("falls back to token count when no run-time row exists (in-flight task)", () => {
+    // Tokens reported mid-run; task hasn't terminated yet so the run-time
+    // rollup is silent on this agent. Keep the token-side estimate
+    // instead of dropping the agent from the table entirely.
+    const merged = mergeAgentDashboardRows(
+      [{ agentId: "agent-b", tokens: 100, cost: 0.5, taskCount: 1 }],
+      [],
+    );
+    expect(merged[0]!.taskCount).toBe(1);
+    expect(merged[0]!.seconds).toBe(0);
+  });
+
+  it("includes agents that have run-time but no tokens", () => {
+    // Task errored before reporting any usage — run-time row exists but
+    // there's no corresponding token row. Agent must still appear on the
+    // list with zeroed-out token columns.
+    const merged = mergeAgentDashboardRows(
+      [],
+      [{ agent_id: "agent-c", total_seconds: 30, task_count: 1, failed_count: 1 }],
+    );
+    expect(merged).toHaveLength(1);
+    expect(merged[0]!.tokens).toBe(0);
+    expect(merged[0]!.cost).toBe(0);
+    expect(merged[0]!.taskCount).toBe(1);
+  });
+
+  it("sorts by cost desc with run-time as a tiebreaker", () => {
+    const merged = mergeAgentDashboardRows(
+      [
+        { agentId: "low", tokens: 100, cost: 1, taskCount: 1 },
+        { agentId: "high", tokens: 100, cost: 9, taskCount: 1 },
+        { agentId: "zero-cost-long", tokens: 0, cost: 0, taskCount: 0 },
+      ],
+      [
+        { agent_id: "zero-cost-long", total_seconds: 1000, task_count: 5, failed_count: 0 },
+      ],
+    );
+    expect(merged.map((r) => r.agentId)).toEqual(["high", "low", "zero-cost-long"]);
+  });
+});
+
+describe("formatDuration", () => {
+  it("formats seconds-only durations", () => {
+    expect(formatDuration(45, "<1m")).toBe("45s");
+  });
+  it("formats minutes and seconds when under one hour", () => {
+    expect(formatDuration(150, "<1m")).toBe("2m 30s");
+    expect(formatDuration(60, "<1m")).toBe("1m");
+  });
+  it("formats hours and minutes when under one day", () => {
+    expect(formatDuration(3 * 3600 + 17 * 60, "<1m")).toBe("3h 17m");
+    expect(formatDuration(3600, "<1m")).toBe("1h");
+  });
+  it("formats days and hours when more than 24 hours", () => {
+    expect(formatDuration(2 * 86400 + 5 * 3600, "<1m")).toBe("2d 5h");
+  });
+  it("falls back to the supplied label for sub-second durations", () => {
+    expect(formatDuration(0, "<1m")).toBe("<1m");
+    expect(formatDuration(0.4, "<1m")).toBe("<1m");
+  });
+});

packages/views/dashboard/utils.ts

@@ -0,0 +1,238 @@
+import type {
+  DashboardUsageDaily,
+  DashboardUsageByAgent,
+  DashboardAgentRunTime,
+} from "@multica/core/types";
+import { estimateCost, estimateCostBreakdown, type DailyTokenData } from "../runtimes/utils";
+
+// ---------------------------------------------------------------------------
+// Dashboard data aggregations
+//
+// The workspace dashboard returns the same per-(date, model) and
+// per-(agent, model) shapes the runtime page does, so cost math reuses
+// `estimateCost` / `estimateCostBreakdown` from the runtimes utils. What
+// the runtimes view does with `aggregateByDate` (works on RuntimeUsage,
+// which carries a `provider` field) we replicate here with a tighter
+// type — fewer optional fields, less conditional logic on the consumer
+// side.
+// ---------------------------------------------------------------------------
+
+export interface DailyCostStack {
+  date: string;
+  label: string;
+  input: number;
+  output: number;
+  cacheWrite: number;
+  total: number;
+}
+
+function formatDateLabel(d: string): string {
+  // Anchor to local midnight so the formatted label matches the bucket the
+  // server picked (which is already in workspace time). Pasting the raw
+  // date as the body of `new Date()` would interpret it as UTC and shift
+  // by the user's offset.
+  const date = new Date(d + "T00:00:00");
+  return `${date.getMonth() + 1}/${date.getDate()}`;
+}
+
+// Per-(date, model) rows → 1 row per date with cost broken into the three
+// segments the stacked bar chart consumes. Stable sort by date asc so the
+// chart x-axis is left-to-right oldest-to-newest.
+export function aggregateDailyCost(usage: DashboardUsageDaily[]): DailyCostStack[] {
+  const map = new Map<string, { input: number; output: number; cacheWrite: number }>();
+  for (const u of usage) {
+    const b = estimateCostBreakdown(u);
+    const entry = map.get(u.date) ?? { input: 0, output: 0, cacheWrite: 0 };
+    entry.input += b.input;
+    entry.output += b.output;
+    entry.cacheWrite += b.cacheWrite;
+    map.set(u.date, entry);
+  }
+  const round = (n: number) => Math.round(n * 100) / 100;
+  return [...map.entries()]
+    .sort(([a], [b]) => a.localeCompare(b))
+    .map(([date, s]) => {
+      const input = round(s.input);
+      const output = round(s.output);
+      const cacheWrite = round(s.cacheWrite);
+      return {
+        date,
+        label: formatDateLabel(date),
+        input,
+        output,
+        cacheWrite,
+        total: round(input + output + cacheWrite),
+      };
+    });
+}
+
+// Per-(date, model) rows → 1 row per date with raw token counts split
+// across the four chart segments. Independent of pricing — unmapped
+// models still contribute here, even if they're excluded from cost.
+// Mirrors `aggregateByDate(...).dailyTokens` from the runtimes utils so
+// the Tokens chart on the Usage page consumes the same shape as the one
+// on the runtime-detail page.
+export function aggregateDailyTokens(usage: DashboardUsageDaily[]): DailyTokenData[] {
+  const map = new Map<
+    string,
+    { input: number; output: number; cacheRead: number; cacheWrite: number }
+  >();
+  for (const u of usage) {
+    const entry = map.get(u.date) ?? {
+      input: 0,
+      output: 0,
+      cacheRead: 0,
+      cacheWrite: 0,
+    };
+    entry.input += u.input_tokens;
+    entry.output += u.output_tokens;
+    entry.cacheRead += u.cache_read_tokens;
+    entry.cacheWrite += u.cache_write_tokens;
+    map.set(u.date, entry);
+  }
+  return [...map.entries()]
+    .sort(([a], [b]) => a.localeCompare(b))
+    .map(([date, t]) => ({
+      date,
+      label: formatDateLabel(date),
+      input: t.input,
+      output: t.output,
+      cacheRead: t.cacheRead,
+      cacheWrite: t.cacheWrite,
+    }));
+}
+
+export interface DashboardTokenTotals {
+  input: number;
+  output: number;
+  cacheRead: number;
+  cacheWrite: number;
+  cost: number;
+  taskCount: number;
+}
+
+// Whole-window totals for the KPI tiles. taskCount sums DISTINCT task counts
+// per row — these are already collapsed server-side per (date, model), so
+// the value can over-count if the same task has tokens in two days; that's
+// acceptable for a KPI ("rough volume") and the per-agent run-time card
+// gives the precise figure.
+export function computeDailyTotals(usage: DashboardUsageDaily[]): DashboardTokenTotals {
+  return usage.reduce<DashboardTokenTotals>(
+    (acc, u) => ({
+      input: acc.input + u.input_tokens,
+      output: acc.output + u.output_tokens,
+      cacheRead: acc.cacheRead + u.cache_read_tokens,
+      cacheWrite: acc.cacheWrite + u.cache_write_tokens,
+      cost: acc.cost + estimateCost(u),
+      taskCount: acc.taskCount + u.task_count,
+    }),
+    { input: 0, output: 0, cacheRead: 0, cacheWrite: 0, cost: 0, taskCount: 0 },
+  );
+}
+
+export interface AgentCostRow {
+  agentId: string;
+  tokens: number;
+  cost: number;
+  taskCount: number;
+}
+
+// Fold per-(agent, model) rows into one row per agent. Cost is the sum
+// across this agent's models, which is the figure the user cares about.
+// Sort by cost desc so the heaviest spender lands first.
+export function aggregateAgentTokens(rows: DashboardUsageByAgent[]): AgentCostRow[] {
+  const map = new Map<string, AgentCostRow>();
+  for (const r of rows) {
+    const entry = map.get(r.agent_id) ?? {
+      agentId: r.agent_id,
+      tokens: 0,
+      cost: 0,
+      taskCount: 0,
+    };
+    entry.tokens +=
+      r.input_tokens + r.output_tokens + r.cache_read_tokens + r.cache_write_tokens;
+    entry.cost += estimateCost(r);
+    entry.taskCount += r.task_count;
+    map.set(r.agent_id, entry);
+  }
+  return [...map.values()].sort((a, b) => b.cost - a.cost);
+}
+
+export interface AgentDashboardRow {
+  agentId: string;
+  tokens: number;
+  cost: number;
+  seconds: number;
+  taskCount: number;
+}
+
+// Merge per-agent token totals with per-agent run-time totals into one
+// row per agent.
+//
+// taskCount comes from `runTimeRows` when available — that rollup is a
+// true per-agent distinct count (`COUNT(*)` on (agent, terminal-task) in
+// SQL). The token rollup's per-(agent, model) counts double-count a task
+// when it spans multiple models, so we only fall back to it for agents
+// with no terminal run yet (in-flight tasks reported tokens but haven't
+// completed). Sorted by cost desc, then run time desc.
+export function mergeAgentDashboardRows(
+  tokenRows: AgentCostRow[],
+  runTimeRows: DashboardAgentRunTime[],
+): AgentDashboardRow[] {
+  const runTimeByAgent = new Map(
+    runTimeRows.map((r) => [r.agent_id, r] as const),
+  );
+  const merged = new Map<string, AgentDashboardRow>();
+  for (const r of tokenRows) {
+    const rt = runTimeByAgent.get(r.agentId);
+    merged.set(r.agentId, {
+      agentId: r.agentId,
+      tokens: r.tokens,
+      cost: r.cost,
+      seconds: rt?.total_seconds ?? 0,
+      taskCount: rt ? rt.task_count : r.taskCount,
+    });
+  }
+  // Agents with run-time rows but zero tokens still belong on the list
+  // (a task that errored before producing usage). Their token columns
+  // stay at 0.
+  for (const r of runTimeRows) {
+    if (merged.has(r.agent_id)) continue;
+    merged.set(r.agent_id, {
+      agentId: r.agent_id,
+      tokens: 0,
+      cost: 0,
+      seconds: r.total_seconds,
+      taskCount: r.task_count,
+    });
+  }
+  return [...merged.values()].sort((a, b) => {
+    if (b.cost !== a.cost) return b.cost - a.cost;
+    return b.seconds - a.seconds;
+  });
+}
+
+// Compact human duration: "1h 23m" / "12m 30s" / "45s" / "<1m". Used for
+// the dashboard run-time KPI and the per-agent run-time column. Keeps two
+// segments max — three segments adds visual noise without precision the
+// dashboard actually needs.
+export function formatDuration(seconds: number, lessThanMinuteLabel: string): string {
+  if (seconds < 0 || !Number.isFinite(seconds)) return lessThanMinuteLabel;
+  if (seconds < 60) {
+    if (seconds < 1) return lessThanMinuteLabel;
+    return `${Math.round(seconds)}s`;
+  }
+  const totalMinutes = Math.floor(seconds / 60);
+  const hours = Math.floor(totalMinutes / 60);
+  const mins = totalMinutes % 60;
+  if (hours === 0) {
+    const secs = Math.floor(seconds) % 60;
+    return secs > 0 ? `${mins}m ${secs}s` : `${mins}m`;
+  }
+  if (hours >= 24) {
+    const days = Math.floor(hours / 24);
+    const h = hours % 24;
+    return h > 0 ? `${days}d ${h}h` : `${days}d`;
+  }
+  return mins > 0 ? `${hours}h ${mins}m` : `${hours}h`;
+}