When tools are unavailble the session just gets stuck in a loop of saying it ...

When tools are unavailble the session just gets stuck in a loop of saying it has done the work without actually doing anything. It should instead inform the user that something is broken and to try again later
(or even better the agent could periodically check if it can proceed similar to what it does with agent merge)

---
| Field | Value |
| --- | --- |
| App version | 1.1.3 |
| OS | Windows 10.0.26200 |
| Theme | GitHub |
| Path | /chat |
| Tenure | Day 6 (Week 1) |

Start by reproducing the issue in the /chat path with unavailable tools on app version 1.1.3 and observe how the session handles tool failures. Trace the session's tool-availability and completion handling; done means it stops claiming success and informs the user that something is broken or provides a retry path.

索引モデルが issue の本文から書いたものです。