Feature Request: Native Terminal AI Agent Support (ACP Integration & Third-Party CLI Orchestration)

### Feature summary

Add native support for external terminal/CLI AI agents (e.g., Claude Code, Codex CLI, Grok CLI, Agy CLI) and support for the Agent Communication Protocol (ACP). This will allow developers to run, manage, and orchestrate third-party CLI agents directly inside the GitHub native app without compromising GitHub's core security control plane or forcing users onto alternative tools.

### What problem are you trying to solve?

Developers are rapidly adopting specialized CLI-based AI coding agents to automate workflows across terminal windows. Currently, restricting the ecosystem or keeping terminal agent integration limited exclusively to GitHub Copilot creates friction for users who rely on multi-agent setups.

Alternative developer tools and terminal wrappers (such as Herdr, Orca, Atrium, Synara, and Muxy) are gaining rapid traction and star counts because they offer first-class terminal agent orchestration. When IDE alternatives (like Cursor Origin and next-gen agentic clients) release full ecosystem capabilities, developers may migrate away from official GitHub clients if those native apps cannot orchestrate their preferred CLI tools natively.

### Proposed solution

1. **Native Terminal Agent Orchestration:** Add built-in integration support for popular AI CLI tools (e.g., Claude CLI, Codex CLI, Grok CLI, Agy CLI) into the app interface.
2. **Standardized Protocol Support (ACP):** Implement support for the Agent Communication Protocol (ACP) to provide a standardized interface for third-party CLI agents.
3. **Secure Control Plane Architecture:** Build a secure execution boundary within the GitHub App. This allows third-party terminal agents to execute locally while ensuring GitHub's core authentication, repository state, and control plane remain completely isolated and secure.

### Workflow impact

* **Improved Developer Velocity:** Developers can trigger, monitor, and switch between specialized AI terminal agents directly within the app without juggling external terminal windows or third-party wrappers.
* **Ecosystem Flexibility:** Developers gain the freedom to choose the best agent for specific tasks (e.g., debugging, refactoring, code generation) while staying inside the official GitHub app environment.
* **Streamlined UI:** Agent outputs, process statuses, and execution logs are unified within a single workspace UI.

### Installation context

* Native desktop runtime environment (GitHub Desktop / GitHub App across macOS, Windows, Linux).
* Requires user permission controls for granting individual CLI tools access to terminal execution contexts inside the app.

### Additional context

Developer preference is shifting heavily toward customizable, CLI-first agent orchestration platforms. The growing popularity of projects in this domain highlights a strong market demand:

* **Herdr:** [https://github.com/ogulcancelik/herdr](https://github.com/ogulcancelik/herdr)
* **Orca:** [https://github.com/stablyai/orca](https://github.com/stablyai/orca)
* **Atrium:** [https://getatrium.dev/](https://getatrium.dev/)
* **Synara:** [https://www.trysynara.com/](https://www.trysynara.com/)
* **Muxy:** [https://github.com/muxy-app/muxy](https://github.com/muxy-app/muxy)

By integrating native support for third-party CLI agents and ACP, GitHub can protect its competitive moat, secure its control plane against untrusted extensions, and drive massive adoption among power developers.

该 issue 未指出任何仓库文件、测试或入口点。首先将所请求的 ACP 支持、受支持的 CLI 代理、桌面平台和权限边界缩小到一个具体范围；通过明确的编排、状态、日志记录和安全要求来定义完成标准。

由索引模型根据 Issue 内容生成。