tower-studio 0.1.1 → 0.1.3

package/skills/tower/SKILL.md

@@ -0,0 +1,375 @@
+---
+name: tower
+description: AI task orchestration platform — use Tower MCP tools to manage workspaces, projects, tasks, and monitor agent execution
+---
+
+# Tower
+
+Tower is an AI task orchestration platform. This skill teaches you how to use Tower's MCP tools to manage projects and tasks.
+
+## MCP Setup
+
+Before using Tower tools, check if the `tower` MCP server is available. If tools like `list_workspaces` are not found, guide the user to configure MCP:
+
+**Claude Code** — add to `~/.claude/settings.json`:
+```json
+{
+  "mcpServers": {
+    "tower": {
+      "command": "npx",
+      "args": ["tsx", "<project-root>/src/mcp/index.ts"]
+    }
+  }
+}
+```
+
+**Other AI agents** — add the same config to your MCP client settings file.
+
+After configuration, restart the AI session. The Tower tools will then be available.
+
+---
+
+## When to Use
+
+Use Tower tools when the user wants to:
+- View, create, or manage tasks and projects
+- Check task execution status or progress
+- Interact with a running task's terminal
+- Search across workspaces, projects, or tasks
+- Organize tasks with labels or status changes
+- Get daily work summary or pending todo list
+- Check what was done today or what's left to do
+
+---
+
+## Scenarios
+
+### "Show me my workspaces" / "List projects"
+
+1. Call `list_workspaces`
+2. If user asks about a specific workspace, call `list_projects` with the workspaceId
+3. Present results in a clean table format
+
+### "Create a task" / "Add a task for ..."
+
+1. Call `list_workspaces` to find the target workspace
+2. Call `list_projects` with workspaceId to find the target project
+3. Ask user to confirm the project (or infer from context)
+4. **Worktree mode**: default is direct mode (no branch isolation). If user says "use worktree", "-w", or "branch isolation", set `useWorktree: true`. Optionally pass `baseBranch` to specify which branch to check out from (e.g. `baseBranch: "develop"`). If `baseBranch` is omitted, the project's current git branch is auto-detected.
+5. **SubPath detection**: check the project description for directory structure hints (e.g. "monorepo: packages/web, packages/api"). If the task clearly belongs to a subdirectory, set `subPath` (e.g. "packages/web"). If unclear, omit it — it's optional.
+6. **References (any files/images)**: ALL user-provided files — including pasted screenshots, uploaded images, and local file paths — should be passed as `references: ["/path/to/file"]` on `create_task`. The tool copies files into the project asset library automatically.
+   - **Local file paths**: pass directly (e.g. `references: ["/path/to/doc.md", "/path/to/design.png"]`)
+   - **Pasted images with known paths**: if the platform provides file paths for pasted media (e.g. OpenClaw's `{{MediaPaths}}`, Claude Code temp files), pass those paths directly — they are local files
+   - **Base64 only (no file path)**: if you only have base64 data with no local path, upload first via `manage_assets` with `action: "upload"`, `projectId`, `base64`, `mimeType`. Get back `{ id: assetId, path }`. Then pass the returned `path` in `references` — `create_task` will automatically copy and link the asset, no separate `link_task` needed
+   - **`link_task` only for retroactive linking**: use `manage_assets` with `action: "link_task"` only when you need to associate existing assets with an already-created task (e.g. user wants to add references after task creation)
+7. Call `create_task` with projectId, title, and optional description/priority/labelIds/subPath/useWorktree/baseBranch/references
+8. `autoStart` defaults to true — task will be created and immediately started. If user says "don't start", "just create", or "-nostart", set `autoStart: false`
+
+### "Start a task" / "Run this task" / "Execute task ..."
+
+1. Call `start_task_execution` with taskId and an optional prompt (instruction for the AI agent)
+2. If no prompt is given, use the task's title/description as context
+3. The task status changes to IN_PROGRESS automatically
+
+### "What's running?" / "Check task progress"
+
+1. Call `get_task_execution_status` with taskId
+2. If status is running, call `get_task_terminal_output` with taskId (default 50 lines)
+3. Summarize: status + recent output + duration
+
+### "Send a message to the task" / "Tell it to ..."
+
+1. Call `send_task_terminal_input` with taskId and the text (include `\n` for Enter)
+2. Wait briefly, then call `get_task_terminal_output` to see the response
+
+### "Move task to done" / "Cancel this task"
+
+1. Call `move_task` with taskId and the target status (DONE, CANCELLED, etc.)
+
+### "Search for ..." / "Find tasks about ..."
+
+1. Call `search` with the query string
+2. Optionally set category to `task`, `project`, or `repository`
+3. Results include `navigateTo` paths for UI navigation
+
+### "Label this task" / "Tag it as ..."
+
+1. Call `list_labels` with workspaceId to see available labels
+2. Call `set_task_labels` with taskId and the desired labelId array
+3. Note: this replaces all existing labels — include current ones if you want to keep them
+
+### "今天做了什么？" / "Daily summary" / "工作总结"
+
+1. Call `daily_summary` (no params for today, or pass `date: "YYYY-MM-DD"` for a specific day)
+2. Results are grouped by workspace → project, with:
+   - **completed**: tasks moved to DONE today
+   - **inProgress**: tasks with activity today but not yet done, including `progressSummary` (last AI chat excerpt)
+3. Present stats: totalCompleted, totalInProgress
+4. Format as a readable report
+
+### "今天有什么待办？" / "Daily todo" / "还有哪些任务没完成？"
+
+1. Call `daily_todo` (returns all TODO/IN_PROGRESS/IN_REVIEW tasks)
+2. Optional filters:
+   - `workspaceId` — narrow to one workspace
+   - `projectId` — narrow to one project
+   - `status` — e.g. `["IN_PROGRESS"]` for only active tasks
+   - `priority` — e.g. `["CRITICAL", "HIGH"]` for urgent only
+3. Results sorted by priority severity (CRITICAL first), grouped by workspace → project
+4. Each task includes `lastSessionId` for resuming execution
+5. Present stats: total count, breakdown by status and priority
+
+### "Create a project" / "Set up a new project"
+
+1. Call `list_workspaces` to pick the workspace
+2. Call `create_project` with workspaceId, name, and optionally:
+   - `gitUrl` — makes it a GIT project with worktree support
+   - `localPath` — path to the local repository
+
+---
+
+## Task Lifecycle
+
+```
+TODO → IN_PROGRESS → IN_REVIEW → DONE
+                                → CANCELLED
+```
+
+- Tasks start as `TODO`
+- `IN_PROGRESS` means an agent is actively working
+- `IN_REVIEW` means execution completed, awaiting review/merge
+- `DONE` means merged and completed
+- `CANCELLED` means dropped
+
+---
+
+## Task Description Format
+
+The `description` field supports Markdown. **Never copy the user's raw message as-is.** Always restructure it into a clear, actionable format:
+
+```markdown
+## 目标
+<one sentence summary of what to achieve>
+
+## 需求
+- <requirement 1>
+- <requirement 2>
+- ...
+
+## 参考
+- <file paths, API endpoints, design references if any>
+
+## 备注
+- <constraints, edge cases, things to watch out for>
+```
+
+Rules:
+- `title` should be short (under 30 chars), summarizing the task
+- `description` should be structured Markdown that an AI agent can execute from
+- Extract actionable requirements from the user's natural language
+- Omit sections that have no content (e.g. skip 备注 if nothing to note)
+- If user provides file paths, put them in 参考 section AND in `references` parameter
+
+---
+
+## Display Templates
+
+All query results MUST follow these templates. Do NOT invent your own format. When results are empty, output "No {items} found." (e.g. "No tasks found.", "No workspaces found.").
+
+### Priority Markers
+
+Use consistently across all templates: 🔴 CRITICAL · 🟠 HIGH · 🟡 MEDIUM · ⚪ LOW
+
+### Labels Format
+
+Always render labels as comma-separated names (e.g. `bug, frontend`). Omit the column if no task has labels.
+
+---
+
+### Workspaces (`list_workspaces`)
+
+```
+| Workspace | Projects | Description |
+|-----------|----------|-------------|
+| {name}    | {projectCount} | {description ?? "—"} |
+```
+
+### Projects (`list_projects`)
+
+Note: the response does not include workspace name. Use the workspace name from the prior `list_workspaces` call or the user's context.
+
+```
+📂 {workspaceName}
+
+| Project | Type | Tasks | Path |
+|---------|------|-------|------|
+| {name} ({alias}) | {type} | {taskCount} | {localPath ?? "—"} |
+```
+
+### Tasks (`list_tasks`)
+
+```
+📋 {projectName}
+
+| ID | Task | Status | Priority | Labels |
+|----|------|--------|----------|--------|
+| {id (first 8 chars)} | {title} | {status} | {priority} | {labels} |
+```
+
+### Task Creation Confirmation
+
+After `create_task` succeeds:
+
+```
+✅ Task created: **{title}**
+- Project: {projectName}
+- Priority: {priority}
+- Status: {status}
+- Worktree: {yes/no}
+{autoStart ? "⚡ Execution started" : ""}
+```
+
+### Daily Summary (`daily_summary`)
+
+Fields: `stats.totalCompleted`, `stats.totalInProgress`, grouped `workspaces[].projects[].completed[]` and `inProgress[]`.
+
+```
+# 📊 Daily Summary — {date}
+
+**Stats**: ✅ {stats.totalCompleted} completed · 🔄 {stats.totalInProgress} in progress
+
+## {workspace.name}
+
+### {project.name}
+
+**Completed**:
+| Task | Priority | Completed At |
+|------|----------|-------------|
+| ✅ {title} | {priority} | {completedAt (HH:mm)} |
+
+**In Progress**:
+| Task | Status | Priority | Progress |
+|------|--------|----------|----------|
+| 🔄 {title} | {status} | {priority} | {progressSummary ?? "—"} |
+```
+
+If no activity: "No activity recorded for {date}."
+
+### Daily Todo (`daily_todo`)
+
+Fields: `stats.total`, `stats.byPriority.{CRITICAL,HIGH,MEDIUM,LOW}`, `stats.byStatus.{TODO,IN_PROGRESS,IN_REVIEW}`.
+
+```
+# 📝 Pending Tasks
+
+**Stats**: {stats.total} tasks · 🔴 {stats.byPriority.CRITICAL} · 🟠 {stats.byPriority.HIGH} · 🟡 {stats.byPriority.MEDIUM} · ⚪ {stats.byPriority.LOW}
+
+## {workspace.name}
+
+### {project.name}
+
+| # | Task | Status | Priority | Labels |
+|---|------|--------|----------|--------|
+| 1 | {title} {lastSessionId ? "🔁" : ""} | {status} | {priority} | {labels} |
+```
+
+Sorted by priority (CRITICAL first). 🔁 = resumable session.
+
+### Search Results (`search`)
+
+Categories: `task`, `project`, `repository`, `note`, `asset`, `all`. Result count = `results.length`.
+
+```
+🔍 Results for "{query}" ({results.length} found)
+
+| Type | Name | Location | Snippet |
+|------|------|----------|---------|
+| {type} | {title} | {subtitle} | {snippet ?? "—"} |
+```
+
+### Execution Status (`get_task_execution_status`)
+
+```
+⚙️ **{taskTitle}**
+- Execution: {executionStatus} · Terminal: {terminalStatus}
+- Started: {startedAt} {endedAt ? "· Ended: " + endedAt : ""}
+- ID: {executionId}
+- Output (last lines):
+\`\`\`
+{outputSnippet ?? "No output"}
+\`\`\`
+```
+
+### Start Execution Confirmation (`start_task_execution`)
+
+```
+⚡ Execution started
+- Task: {taskId}
+- Execution ID: {executionId}
+- Worktree: {worktreePath ?? "direct mode"}
+```
+
+### Terminal Output (`get_task_terminal_output`)
+
+```
+📺 Terminal — {taskId} ({total} total lines, showing last {lines.length})
+
+\`\`\`
+{lines.join("\n")}
+\`\`\`
+```
+
+### Labels (`list_labels`)
+
+```
+🏷️ Labels for {workspaceName}
+
+| Label | Color | Type |
+|-------|-------|------|
+| {name} | {color} | {isBuiltin ? "Builtin" : "Custom"} |
+```
+
+### Project Identification (`identify_project`)
+
+```
+🔎 Project matches for "{query}"
+
+| Project | Alias | Workspace | Confidence |
+|---------|-------|-----------|------------|
+| {name} | {alias ?? "—"} | {workspaceName} | {(confidence * 100).toFixed(0)}% |
+```
+
+### Notes (`manage_notes` — list/get)
+
+```
+📝 Notes for {projectName}
+
+| Title | Updated | Preview |
+|-------|---------|---------|
+| {title} | {updatedAt (MM-DD HH:mm)} | {content (first 60 chars)}... |
+```
+
+### Assets (`manage_assets` — list)
+
+```
+📎 Assets for {projectName}
+
+| Name | Type | Size | Linked Tasks |
+|------|------|------|-------------|
+| {originalName} | {mimeType} | {size} | {taskCount} |
+```
+
+---
+
+## Important Rules
+
+- **Scope boundary**: You are a **task management operator only**. If the user asks you to write code, explain code, debug, search the web, read/write files, or anything outside Tower task management, reply: "抱歉，我只能帮你管理工作区、项目和任务。编码、调试等操作请在任务终端中完成。" Do NOT attempt to answer out-of-scope questions.
+- **Display format is mandatory**: always use the templates above, never output raw JSON or invent custom formats
+- **Empty results**: always output "No {items} found." — never silently return nothing
+- **SubPath**: for monorepo or multi-folder projects, use `subPath` on task creation to specify the working directory (e.g. "packages/web"). The project description should document the directory structure. If not sure, omit subPath.
+- **Cascade deletes**: deleting a workspace removes all its projects and tasks
+- **Label replacement**: `set_task_labels` and `update_task` with labelIds do a full replace, not merge
+- **Builtin labels**: cannot be deleted (isBuiltin: true)
+- **One terminal per task**: each task can have at most one active PTY session
+- **Search limit**: returns at most 20 results per query
+- **Search categories**: `task`, `project`, `repository`, `note`, `asset`, `all`

package/src/actions/__tests__/agent-actions-username.test.ts

@@ -0,0 +1,30 @@
+import { describe, it, expect } from "vitest";
+
+/**
+ * Mirrors the username injection logic used in startPtyExecution,
+ * resumePtyExecution, and continueLatestPtyExecution in agent-actions.ts.
+ *
+ * This helper is self-contained so tests pass independently of the
+ * full PTY spawn infrastructure.
+ */
+function buildUsernameArgs(username: string | null): string[] {
+  if (!username) return [];
+  return ["--append-system-prompt", `The user's name is ${username}.`];
+}
+
+describe("buildUsernameArgs", () => {
+  it("returns the correct --append-system-prompt args for a valid username", () => {
+    expect(buildUsernameArgs("Alice")).toEqual([
+      "--append-system-prompt",
+      "The user's name is Alice.",
+    ]);
+  });
+
+  it("returns empty array for empty string username", () => {
+    expect(buildUsernameArgs("")).toEqual([]);
+  });
+
+  it("returns empty array for null username", () => {
+    expect(buildUsernameArgs(null)).toEqual([]);
+  });
+});

package/src/actions/__tests__/asset-actions.test.ts

@@ -0,0 +1,251 @@
+import { describe, it, expect, vi, beforeEach } from "vitest";
+
+vi.mock("@/lib/db", () => ({
+  db: {
+    projectAsset: {
+      create: vi.fn(),
+      delete: vi.fn(),
+      findUnique: vi.fn(),
+      findMany: vi.fn(),
+    },
+    project: { findUnique: vi.fn() },
+  },
+}));
+vi.mock("next/cache", () => ({ revalidatePath: vi.fn() }));
+vi.mock("node:fs", async (importOriginal) => {
+  const actual = await importOriginal<typeof import("node:fs")>();
+  return {
+    ...actual,
+    existsSync: vi.fn(() => false),
+    promises: { unlink: vi.fn(), writeFile: vi.fn() },
+    mkdirSync: vi.fn(),
+  };
+});
+vi.mock("@/lib/file-utils", () => ({
+  ensureAssetsDir: vi.fn(() => "/mock/assets/dir"),
+}));
+vi.mock("@/actions/config-actions", () => ({
+  getConfigValue: vi.fn(async () => 52428800), // 50MB default
+}));
+
+import { db } from "@/lib/db";
+import * as fs from "node:fs";
+import { revalidatePath } from "next/cache";
+import { ensureAssetsDir } from "@/lib/file-utils";
+import { getConfigValue } from "@/actions/config-actions";
+import {
+  createAsset,
+  deleteAsset,
+  getProjectAssets,
+  getTaskAssets,
+  getAssetById,
+  uploadAsset,
+} from "@/actions/asset-actions";
+
+const mockDb = db as {
+  projectAsset: {
+    create: ReturnType<typeof vi.fn>;
+    delete: ReturnType<typeof vi.fn>;
+    findUnique: ReturnType<typeof vi.fn>;
+    findMany: ReturnType<typeof vi.fn>;
+  };
+  project: { findUnique: ReturnType<typeof vi.fn> };
+};
+
+function makeFormData(
+  overrides: { file?: object | null; projectId?: string | null; taskId?: string; description?: string } = {}
+): FormData {
+  const fd = {
+    get: vi.fn((key: string) => {
+      if (key === "file") return "file" in overrides ? overrides.file : makeFile("test.txt", 100);
+      if (key === "projectId") return "projectId" in overrides ? overrides.projectId : "proj123";
+      if (key === "taskId") return overrides.taskId ?? null;
+      if (key === "description") return overrides.description ?? null;
+      return null;
+    }),
+  } as unknown as FormData;
+  return fd;
+}
+
+function makeFile(name: string, size: number, type = "text/plain") {
+  return {
+    name,
+    size,
+    type,
+    arrayBuffer: vi.fn(async () => new ArrayBuffer(size)),
+  };
+}
+
+describe("asset-actions", () => {
+  beforeEach(() => {
+    vi.clearAllMocks();
+  });
+
+  // ── createAsset ──────────────────────────────────────────────────────────
+
+  describe("createAsset", () => {
+    it("creates an asset and revalidates path", async () => {
+      const fakeAsset = { id: "a1", filename: "file.txt", path: "/p/file.txt", projectId: "proj123" };
+      mockDb.projectAsset.create.mockResolvedValue(fakeAsset);
+
+      const result = await createAsset({
+        filename: "file.txt",
+        path: "/p/file.txt",
+        projectId: "proj123",
+      });
+
+      expect(mockDb.projectAsset.create).toHaveBeenCalledWith(
+        expect.objectContaining({ data: expect.objectContaining({ filename: "file.txt", projectId: "proj123" }) })
+      );
+      expect(revalidatePath).toHaveBeenCalledWith("/workspaces");
+      expect(result).toBe(fakeAsset);
+    });
+
+    it("throws on empty filename (schema validation)", async () => {
+      await expect(
+        createAsset({ filename: "", path: "/p/file.txt", projectId: "proj123" })
+      ).rejects.toThrow();
+
+      expect(mockDb.projectAsset.create).not.toHaveBeenCalled();
+    });
+  });
+
+  // ── deleteAsset ──────────────────────────────────────────────────────────
+
+  describe("deleteAsset", () => {
+    it("calls unlink when asset has a path, then deletes from DB", async () => {
+      const fakeAsset = { id: "a1", path: "/some/file.txt" };
+      mockDb.projectAsset.findUnique.mockResolvedValue(fakeAsset);
+      (fs.promises.unlink as ReturnType<typeof vi.fn>).mockResolvedValue(undefined);
+      mockDb.projectAsset.delete.mockResolvedValue(fakeAsset);
+
+      await deleteAsset("a1");
+
+      expect(fs.promises.unlink).toHaveBeenCalledWith("/some/file.txt");
+      expect(mockDb.projectAsset.delete).toHaveBeenCalledWith({ where: { id: "a1" } });
+      expect(revalidatePath).toHaveBeenCalledWith("/workspaces");
+    });
+
+    it("skips unlink when asset path is null, still deletes from DB", async () => {
+      const fakeAsset = { id: "a2", path: null };
+      mockDb.projectAsset.findUnique.mockResolvedValue(fakeAsset);
+      mockDb.projectAsset.delete.mockResolvedValue(fakeAsset);
+
+      await deleteAsset("a2");
+
+      expect(fs.promises.unlink).not.toHaveBeenCalled();
+      expect(mockDb.projectAsset.delete).toHaveBeenCalledWith({ where: { id: "a2" } });
+    });
+  });
+
+  // ── getProjectAssets ─────────────────────────────────────────────────────
+
+  describe("getProjectAssets", () => {
+    it("queries with projectId (includes all assets)", async () => {
+      mockDb.projectAsset.findMany.mockResolvedValue([]);
+
+      await getProjectAssets("proj123");
+
+      expect(mockDb.projectAsset.findMany).toHaveBeenCalledWith(
+        expect.objectContaining({ where: { projectId: "proj123" } })
+      );
+    });
+  });
+
+  // ── getTaskAssets ────────────────────────────────────────────────────────
+
+  describe("getTaskAssets", () => {
+    it("queries with taskId filter", async () => {
+      mockDb.projectAsset.findMany.mockResolvedValue([]);
+
+      await getTaskAssets("task456");
+
+      expect(mockDb.projectAsset.findMany).toHaveBeenCalledWith(
+        expect.objectContaining({ where: { taskId: "task456" } })
+      );
+    });
+  });
+
+  // ── getAssetById ─────────────────────────────────────────────────────────
+
+  describe("getAssetById", () => {
+    it("calls findUnique with the given assetId", async () => {
+      const fakeAsset = { id: "a1" };
+      mockDb.projectAsset.findUnique.mockResolvedValue(fakeAsset);
+
+      const result = await getAssetById("a1");
+
+      expect(mockDb.projectAsset.findUnique).toHaveBeenCalledWith({ where: { id: "a1" } });
+      expect(result).toBe(fakeAsset);
+    });
+  });
+
+  // ── uploadAsset ──────────────────────────────────────────────────────────
+
+  describe("uploadAsset", () => {
+    it("happy path: writes file and creates asset", async () => {
+      const fakeProject = { id: "proj123" };
+      const fakeAsset = { id: "a1", filename: "test.txt", projectId: "proj123" };
+      mockDb.project.findUnique.mockResolvedValue(fakeProject);
+      mockDb.projectAsset.create.mockResolvedValue(fakeAsset);
+      (fs.promises.writeFile as ReturnType<typeof vi.fn>).mockResolvedValue(undefined);
+
+      const fd = makeFormData({ file: makeFile("test.txt", 100) });
+      const result = await uploadAsset(fd);
+
+      expect(ensureAssetsDir).toHaveBeenCalledWith("proj123");
+      expect(fs.promises.writeFile).toHaveBeenCalled();
+      expect(mockDb.projectAsset.create).toHaveBeenCalled();
+      expect(result).toBe(fakeAsset);
+    });
+
+    it("throws when file is missing from FormData", async () => {
+      const fd = makeFormData({ file: null });
+
+      await expect(uploadAsset(fd)).rejects.toThrow("Missing file or projectId");
+    });
+
+    it("throws when projectId is missing from FormData", async () => {
+      const fd = makeFormData({ projectId: null });
+
+      await expect(uploadAsset(fd)).rejects.toThrow("Missing file or projectId");
+    });
+
+    it("throws when file is too large", async () => {
+      // getConfigValue returns 100 bytes max
+      (getConfigValue as ReturnType<typeof vi.fn>).mockResolvedValue(100);
+      const largeFile = makeFile("big.bin", 200);
+      const fd = makeFormData({ file: largeFile });
+
+      await expect(uploadAsset(fd)).rejects.toThrow(/File too large/);
+    });
+
+    it("throws when projectId is invalid (not in DB)", async () => {
+      mockDb.project.findUnique.mockResolvedValue(null);
+      const fd = makeFormData({ file: makeFile("test.txt", 10) });
+
+      await expect(uploadAsset(fd)).rejects.toThrow("Invalid projectId");
+    });
+
+    it("sanitizes filename with path traversal attempt", async () => {
+      const fakeProject = { id: "proj123" };
+      const fakeAsset = { id: "a1", filename: "passwd", projectId: "proj123" };
+      mockDb.project.findUnique.mockResolvedValue(fakeProject);
+      mockDb.projectAsset.create.mockResolvedValue(fakeAsset);
+      (fs.promises.writeFile as ReturnType<typeof vi.fn>).mockResolvedValue(undefined);
+
+      // file named with path traversal: path.basename should strip to "passwd"
+      const traversalFile = makeFile("../../../etc/passwd", 10);
+      const fd = makeFormData({ file: traversalFile });
+
+      const result = await uploadAsset(fd);
+
+      // Verify writeFile was called — the asset was created (not thrown)
+      expect(fs.promises.writeFile).toHaveBeenCalled();
+      // Verify the create call used the sanitized filename (basename only)
+      const createCall = mockDb.projectAsset.create.mock.calls[0][0];
+      expect(createCall.data.filename).toBe("passwd");
+      expect(result).toBe(fakeAsset);
+    });
+  });
+});

package/src/actions/__tests__/assistant-actions.test.ts

@@ -0,0 +1,20 @@
+import { describe, it } from "vitest";
+
+describe("assistant-actions", () => {
+  describe("startAssistantSession", () => {
+    it.todo("creates a PTY session keyed by __assistant__ (BE-01)");
+    it.todo("includes --allowedTools mcp__tower__* in CLI args (BE-03)");
+    it.todo("includes --append-system-prompt with configured prompt (BE-02)");
+    it.todo("destroys existing session before creating new one (UX-01)");
+    it.todo("throws if no default CliProfile exists");
+  });
+
+  describe("stopAssistantSession", () => {
+    it.todo("destroys the __assistant__ session (BE-05)");
+  });
+
+  describe("getAssistantSessionStatus", () => {
+    it.todo("returns 'idle' when no session exists");
+    it.todo("returns 'running' when session is active");
+  });
+});