[General] Pull requests created by this app tend to be somewhat impenetrable for human ...

**Mood:** 😔
**Category:** General

Pull requests created by this app tend to be somewhat impenetrable for human reviewers. They contain terse and detailed descriptions of how the code changed, but they're missing the purpose and goals behind the change, even when these are clear from the prompts provided by the user. When Copilot does incorporate information from prompts into PR descriptions, it tends to give too much weight to correction/steering prompts compared to the original motivating prompts, causing confusion.

---
| Field | Value |
| --- | --- |
| App version | 1.0.12 |
| OS | macOS 26.5.1 |
| Theme | GitHub |
| Path | /chat |
| Tenure | Week 12 |

Start from the /chat flow and inspect how the app assembles pull request descriptions from user prompts. Compare the generated descriptions with the motivating and correction prompts; done means the result clearly explains the purpose and goals without letting steering prompts obscure the original intent.

索引モデルが issue の本文から書いたものです。