The Copilot CLI extension system is powerful, but there's a meaningful gap th...

The Copilot CLI extension system is powerful, but there's a meaningful gap that limits what developers can build: there are no lifecycle hooks for detecting when a turn stalls mid-response, when the agent goes idle unexpectedly, or when a session is opened. This means reliability-focused extensions, such as a watchdog that detects when the AI stops working mid-task and prompts the user with a "Continue" or restart option, simply cannot be built today. Adding event hooks for turn-stall detection, unexpected idle states, and app startup would unlock a whole category of resilience and recovery extensions that would meaningfully improve the user experience, especially during long autonomous runs where a silent stall currently leaves the user with no indication that work has stopped. These hooks would let developers build safety nets that keep sessions productive without any manual intervention, which feels aligned with the goal of an autonomous coding assistant. I would love to see this considered, and I'm happy to help validate or test any prototype implementation.

---
| Field | Value |
| --- | --- |
| App version | 1.0.19 |
| OS | Windows 10.0.26200 |
| Theme | ABOLKOG |
| Path | /chat |
| Tenure | Day 3 (Week 1) |

首先审查 Copilot CLI 扩展系统及其现有的 lifecycle 能力。定义如何向扩展暴露 turn 卡住、意外的 idle 状态和会话打开事件，然后验证 watchdog 或 recovery 扩展能否按预期响应每个事件。

由索引模型根据 Issue 内容生成。