release: merge develop into main

README.md

@@ -6,7 +6,9 @@
 
 **Coder Studio, made for vibe coding.**
 
-Run Claude Code and Codex in one workspace. Keep your terminal, files, Git view, and AI sessions available from any device.
+An agentic workspace for real development. Run, inspect, and supervise coding agents with terminals, files, Git, sessions, and review in one browser workspace.
+
+Built-in support today: Claude Code and Codex. Your code and runtime stay on your machine.
 
 [![npm version](https://img.shields.io/npm/v/@spencer-kit/coder-studio.svg)](https://www.npmjs.com/package/@spencer-kit/coder-studio)
 [![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)
@@ -21,17 +23,17 @@ Run Claude Code and Codex in one workspace. Keep your terminal, files, Git view,
 
 [![Workspace Preview](docs/help/assets/screenshot-desktop-workspace-full.png)](docs/help/assets/screenshot-desktop-workspace-full.png)
 
-<div align="center">Preview the full workspace layout built for vibe coding, supervision, and device switching.</div>
+<div align="center">Preview the full workspace layout built for agent runs, review, supervision, and device switching.</div>
 
 ## Why It Feels Different
 
-- **One browser workspace for vibe coding** — Keep terminal, files, Git, and agent sessions in one place.
+- **One browser workspace for real agent work** — Keep terminals, files, Git, sessions, and review in one place.
 - **Built for device switching** — Start on desktop, continue on tablet, and check progress from your phone.
-- **Objective-driven multi-step orchestration** — Let Supervisor steer long-running AI tasks so you do not have to babysit every turn, reduce repetitive manual prompting, and get more consistent outcomes.
+- **Keep control local** — Your code and runtime stay on your machine.
 
 ## Why Coder Studio?
 
-Vibe coding agents are powerful, but the raw workflow is still fragmented:
+Vibe coding agents are fast, but real development still gets fragmented:
 
 - the agent runs in one terminal
 - files and diffs live in another editor
@@ -46,7 +48,7 @@ Coder Studio turns that scattered workflow into one local browser workspace.
 | Long agent tasks | Watch a terminal or come back later and reconstruct context | Keep sessions, terminal output, files, and Git changes visible in one workspace |
 | Cross-device work | Use SSH, remote desktop, or rebuild context on another machine | Reopen the same local workspace from desktop, tablet, or phone |
 | Reviewing AI changes | Jump between terminal, editor, and Git tools | Inspect files and diffs beside the agent session |
-| Multiple agents | Manage separate terminal windows and histories | Run Claude and Codex sessions side by side in one workspace |
+| Multiple agents | Manage separate terminal windows and histories | Run built-in Claude Code and Codex sessions side by side in one workspace today |
 | Local-first control | Move work into a hosted IDE or cloud VM | Keep the runtime and project files on your own machine |
 
 ## Quick Start
@@ -59,7 +61,7 @@ npm install -g @spencer-kit/coder-studio
 coder-studio open
 ```
 
-Your browser opens automatically. Select your project folder and start working with Claude Code or OpenAI Codex.
+Your browser opens automatically. Select your project folder and start working with Claude Code or OpenAI Codex today.
 
 > **No AI CLI installed yet?** You can still browse files and use the terminal. Install Claude Code or Codex later when needed.
 
@@ -81,7 +83,7 @@ Your browser opens automatically. Select your project folder and start working w
 
 ### AI-Assisted Coding
 
-- Run Claude Code and Codex sessions side by side
+- Run Claude Code and Codex sessions side by side today
 - Keep terminal, editor, Git, and supervisor state in one unified interface
 - Resume active AI work from another device without rebuilding context
 
@@ -113,11 +115,12 @@ The same workspace URL works across all devices — interface adapts automatical
 |---------|-------------|
 | **Cross-Device Workspace** | Reopen the same coding environment from desktop, tablet, or phone without rebuilding context |
 | **Supervisor Loops** | Run objective-driven evaluation and follow-up cycles for long AI tasks with less manual babysitting |
-| **Claude Code + Codex** | Use both agent CLIs inside one workspace instead of splitting your workflow across separate tools |
-| **Unified Terminal, Editor, and Git** | Keep PTY terminals, Monaco editing, diffs, and changed files in one browser UI |
+| **Built-in Agent Providers** | Use Claude Code and Codex inside one workspace today instead of splitting your workflow across separate tools |
+| **Unified Terminal, Files, and Git** | Keep PTY terminals, Monaco editing, diffs, and changed files in one browser UI |
+| **Reviewable AI Work** | Inspect changed files and diffs beside the session before trusting the result |
 | **Responsive Workspace UI** | Use layouts tuned for desktop, tablet, and mobile instead of a desktop-only interface squeezed onto small screens |
 | **Session Continuity** | Resume active sessions and keep AI work visible across device switches |
-| **Local-First Runtime** | Keep code and runtime on your machine instead of relying on a cloud IDE |
+| **Local Runtime Control** | Keep code and runtime on your machine instead of relying on a cloud IDE |
 
 ---
 
@@ -150,7 +153,7 @@ The same workspace URL works across all devices — interface adapts automatical
 
 ## 👥 Who Should Use Coder Studio
 
-- **AI Coding Power Users** — Daily Claude Code / Codex users who want better session management
+- **Developers Running Coding Agents** — Want terminals, files, Git, sessions, and review in one place
 - **Multi-Device Developers** — Switch between office, home, and mobile devices frequently
 - **Developers Running Long AI Tasks** — Want Supervisor to keep multi-step work moving without constant babysitting
 - **Privacy-Conscious Developers** — Want code to stay on local machine, not cloud IDE

README.zh-CN.md

@@ -6,7 +6,9 @@
 
 **Coder Studio，生来就是 vibe coding。**
 
-在同一个工作台里运行 Claude Code 和 Codex，让终端、文件、Git 视图和 AI 会话跟着你在不同设备间延续。
+面向真实开发的 agentic workspace。用一个浏览器工作区运行、检查和监督 coding agent，把终端、文件、Git、会话和代码审查放在一起。
+
+当前内置支持：Claude Code 和 Codex。你的代码和运行时保留在自己的机器上。
 
 [![npm version](https://img.shields.io/npm/v/@spencer-kit/coder-studio.svg)](https://www.npmjs.com/package/@spencer-kit/coder-studio)
 [![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)
@@ -21,17 +23,17 @@
 
 [![工作区预览](docs/help/assets/screenshot-desktop-workspace-full.png)](docs/help/assets/screenshot-desktop-workspace-full.png)
 
-<div align="center">预览这个为 vibe coding、Supervisor 监督和跨设备切换而设计的完整工作区布局。</div>
+<div align="center">预览这个为 Agent 运行、改动审查、Supervisor 监督和跨设备切换而设计的完整工作区布局。</div>
 
 ## 为什么它不一样
 
-- **一个浏览器里完成 AI 编程工作流** — 把终端、文件、Git 和 AI 会话放到同一个工作台。
+- **一个浏览器里完成真实 Agent 工作流** — 把终端、文件、Git、会话和代码审查放到同一个工作台。
 - **真正为设备切换而设计** — 在桌面端开始，在平板继续，用手机随时查看 Agent 进度。
-- **目标驱动的多轮调度** — 让 Supervisor 接管长任务推进，你不必全程盯守每一轮输出，减少机械重复的人工催促，并获得更稳定的执行效果。
+- **保留本地控制权** — 你的代码和运行时都留在自己的机器上。
 
 ## 为什么选择 Coder Studio？
 
-vibe coding agent 已经很强，但原始工作流仍然是割裂的：
+vibe coding agent 很快，但真实开发里的工作流仍然是割裂的：
 
 - Agent 跑在一个终端里
 - 文件和 diff 在另一个编辑器里
@@ -46,7 +48,7 @@ Coder Studio 把这些分散的环节收进同一个本地浏览器工作台。
 | 长时间 Agent 任务 | 盯着终端，或者回来后重新拼上下文 | 会话、终端输出、文件和 Git 变更都在同一个工作区里 |
 | 跨设备继续 | SSH、远程桌面，或在另一台机器重新配置 | 桌面、平板、手机重新打开同一个本地工作区 |
 | 审阅 AI 改动 | 在终端、编辑器、Git 工具之间切换 | 在 Agent 会话旁边直接查看文件和 diff |
-| 多 Agent 并行 | 多个终端窗口和历史记录分散管理 | Claude 和 Codex 会话在同一个工作区里并行管理 |
+| 多 Agent 并行 | 多个终端窗口和历史记录分散管理 | 今天先把内置的 Claude Code 和 Codex 会话并行放进同一个工作区 |
 | 本地优先 | 把环境迁到云 IDE 或远程 VM | 运行时和项目文件留在自己的机器上 |
 
 ## 快速开始
@@ -81,7 +83,7 @@ coder-studio open
 
 ### AI 辅助编程
 
-- 并行运行 Claude Code 和 Codex 会话
+- 今天并行运行 Claude Code 和 Codex 会话
 - 终端、编辑器、Git 和 Supervisor 状态统一在一个界面
 - 切换设备后继续当前 AI 工作，不必重新建立上下文
 
@@ -113,11 +115,12 @@ coder-studio open
 |------|------|
 | **跨设备工作区** | 在桌面、平板和手机之间重新打开同一个编码环境，不必重新建立上下文 |
 | **Supervisor 监督循环** | 围绕目标运行评估与续推循环，减少长任务中的人工盯守 |
-| **Claude Code + Codex** | 在同一个工作区里使用两套 Agent CLI，而不是把工作流拆散到多个工具中 |
-| **终端、编辑器和 Git 一体化** | 在同一个浏览器界面里完成 PTY 终端、Monaco 编辑、diff 和变更查看 |
+| **内置 Agent Provider** | 今天先在同一个工作区里使用 Claude Code 和 Codex，而不是把工作流拆散到多个工具中 |
+| **终端、文件和 Git 一体化** | 在同一个浏览器界面里完成 PTY 终端、Monaco 编辑、diff 和变更查看 |
+| **可审查的 AI 改动** | 先在 Agent 会话旁检查文件和 diff，再决定是否信任结果 |
 | **响应式工作区界面** | 提供面向桌面、平板和手机的布局，而不是把桌面界面硬塞进小屏幕 |
 | **会话连续性** | 切换设备后继续当前活跃会话，让 AI 工作保持可见 |
-| **本地优先运行时** | 代码和运行时都留在你的机器上，不依赖云 IDE |
+| **本地运行时控制** | 代码和运行时都留在你的机器上，不依赖云 IDE |
 
 ---
 
@@ -150,7 +153,7 @@ coder-studio open
 
 ## 👥 谁适合使用
 
-- **AI 编程深度用户** — 每天使用 Claude Code / Codex，想要更好的会话管理
+- **运行 coding agent 的开发者** — 希望把终端、文件、Git、会话和代码审查放到同一个地方
 - **多设备开发者** — 频繁在办公室、家和移动设备之间切换
 - **运行长任务的开发者** — 希望由 Supervisor 持续推进多轮任务，而不是全程人工盯守
 - **注重隐私的开发者** — 希望代码留在本地机器，不依赖云 IDE

docs/superpowers/plans/2026-05-17-agentic-workspace-phase-3-workspace-intelligence.md

@@ -6,7 +6,7 @@
 
 **Goal:** Add a project understanding layer that summarizes Git, package manager, framework, commands, docs, and agent instruction state for the active workspace.
 
-**Architecture:** Server-side workspace inspection produces a typed `WorkspaceIntelligenceSummary`. Web UI consumes it through a command and renders a compact setup/context panel inside the workspace.
+**Architecture:** Server-side workspace inspection produces a typed `WorkspaceIntelligenceSummary`. The command stays server-facing and is reused by later agent-instruction and context flows; no dedicated workspace UI panel is kept in this phase.
 
 **Tech Stack:** TypeScript, Node filesystem APIs, Zod, Vitest, React Testing Library.
 
@@ -18,7 +18,6 @@ Includes:
 
 - Workspace inspection module.
 - `workspace.intelligence` command.
-- Summary panel in the workspace UI.
 - Detection for Git, package managers, package scripts, common frameworks, README/docs, and `AGENTS.md`.
 
 Excludes:
@@ -36,10 +35,6 @@ Excludes:
 - Create: `packages/server/src/__tests__/workspace/intelligence.test.ts`
 - Modify: `packages/server/src/commands/workspace.ts`
 - Create: `packages/server/src/__tests__/workspace-intelligence-command.test.ts`
-- Create: `packages/web/src/features/workspace-intelligence/actions/use-workspace-intelligence.ts`
-- Create: `packages/web/src/features/workspace-intelligence/components/workspace-intelligence-panel.tsx`
-- Create: `packages/web/src/features/workspace-intelligence/components/workspace-intelligence-panel.test.tsx`
-- Modify: `packages/web/src/features/workspace/views/desktop/workspace-desktop-view.tsx`
 
 ## Data Model
 
@@ -96,24 +91,22 @@ export interface WorkspaceIntelligenceSummary {
 - [ ] Detect docs via `README.md` and top-level `docs/`.
 - [ ] Detect `AGENTS.md`.
 - [ ] Register `workspace.intelligence`.
-- [ ] Add a desktop panel showing project type, commands, Git state, docs, and instruction state.
-- [ ] Keep the panel action-oriented: show recommended commands and `AGENTS.md` state instead of a passive dashboard.
+- [ ] Keep the command server-side and reuse it for later agent-instruction and context features.
 
 ## Acceptance Criteria
 
 - Opening a workspace can produce a stable typed summary.
 - Summary works for non-Git folders.
 - Summary works when `package.json` is missing.
-- UI makes clear whether `AGENTS.md` exists.
+- UI surfaces for `AGENTS.md` are deferred to the agent-instructions phase.
 - No provider-specific assumptions are required.
 
 ## Verification
 
 ```bash
 pnpm exec vitest run \
   packages/server/src/__tests__/workspace/intelligence.test.ts \
-  packages/server/src/__tests__/workspace-intelligence-command.test.ts \
-  packages/web/src/features/workspace-intelligence/components/workspace-intelligence-panel.test.tsx
+  packages/server/src/__tests__/workspace-intelligence-command.test.ts
 ```
 
 Expected: all tests pass.
@@ -124,4 +117,3 @@ Expected: all tests pass.
 git add packages/core packages/server packages/web/src/features/workspace-intelligence packages/web/src/features/workspace/views/desktop/workspace-desktop-view.tsx
 git commit -m "feat: add workspace intelligence summary"
 ```
-

docs/superpowers/plans/2026-05-17-agentic-workspace-phase-4-agent-instructions.md

@@ -6,7 +6,7 @@
 
 **Goal:** Let users create, inspect, and edit universal project-level agent instructions through `AGENTS.md`.
 
-**Architecture:** Build on Workspace Intelligence. The server generates deterministic `AGENTS.md` content from project facts; the web UI exposes create/edit actions without tying the feature to any single provider.
+**Architecture:** Build on Workspace Intelligence. The server generates deterministic `AGENTS.md` content from project facts and exposes read/write/health commands. No separate workspace intelligence panel or dedicated frontend editor is part of this phase.
 
 **Tech Stack:** TypeScript, Node filesystem APIs, Zod, Vitest, React Testing Library.
 
@@ -17,8 +17,7 @@
 Includes:
 
 - `AGENTS.md` generation from workspace intelligence.
-- Server commands to read, generate, and write instructions.
-- UI panel for instruction state and editing.
+- Server commands to read, generate, write, and inspect instructions.
 - Provider-specific notes section.
 - Instruction health checks.
 
@@ -39,9 +38,6 @@ Excludes:
 - Create: `packages/server/src/commands/agent-instructions.ts`
 - Modify: `packages/server/src/commands/index.ts`
 - Create: `packages/server/src/__tests__/agent-instructions-command.test.ts`
-- Create: `packages/web/src/features/agent-instructions/actions/use-agent-instructions.ts`
-- Create: `packages/web/src/features/agent-instructions/components/agent-instructions-panel.tsx`
-- Create: `packages/web/src/features/agent-instructions/components/agent-instructions-panel.test.tsx`
 
 ## Commands
 
@@ -102,12 +98,7 @@ Omit command lines that are unknown rather than inserting placeholders.
   - has safety rules
 - [ ] Implement server commands using safe workspace path resolution.
 - [ ] Add command tests for missing workspace, missing file, generated content, write roundtrip, and health output.
-- [ ] Add UI panel with actions:
-  - create from project context
-  - open existing instructions
-  - save changes
-  - show health issues
-- [ ] Add tests for create, edit, and health states.
+- [ ] Add command-only workspace tests for create, edit, and health states.
 
 ## Acceptance Criteria
 
@@ -122,16 +113,14 @@ Omit command lines that are unknown rather than inserting placeholders.
 pnpm exec vitest run \
   packages/server/src/__tests__/agent-instructions/generator.test.ts \
   packages/server/src/__tests__/agent-instructions/health.test.ts \
-  packages/server/src/__tests__/agent-instructions-command.test.ts \
-  packages/web/src/features/agent-instructions/components/agent-instructions-panel.test.tsx
+  packages/server/src/__tests__/agent-instructions-command.test.ts
 ```
 
 Expected: all tests pass.
 
 ## Suggested Commit
 
 ```bash
-git add packages/core packages/server packages/web/src/features/agent-instructions
+git add packages/core packages/server docs/superpowers/plans/2026-05-17-agentic-workspace-phase-4-agent-instructions.md
 git commit -m "feat: add universal agent instructions"
 ```
-