Chat markdown rendering of numbered lists is misleading

[Skn0tt]

• Copilot Chat Extension Version: 0.37.2026012101
• VS Code Version: 1.109.0-insider
• OS Version: latest macOS
• Feature (e.g. agent/edit/ask mode): chat markdown rendering
• Selected model (e.g. GPT 4.1, Claude 3.7 Sonnet): all
• Logs:

Steps to Reproduce:

1. type the following chat message in the chat sidebar:

2. do this differently
4. yes, but change on thing

2. see how it gets rendered:

2. do this differently
3. yes, but change on thing

It appears as if my agent received a different message than I typed, even though under the hood it received the right thing.
This is Markdown's rule about sequential list numbers at work, and it's a papercut in UX.

A possible fix would be to add a customizing marked.js rendering in chatContentMarkdownRenderer.ts. I'd be happy to contribute a fix.

[cy1225671022-wq]

11

[cy1225671022-wq]

1

[roblourens]

I can see that it would be annoying, looks like you are responding to numbered questions. Not sure how to customize the rendering though.

[Skn0tt]

I've experienced this mostly in Plan mode, where i'm responding to the questions, yes. But recently i've been seing more questionnaires, so maybe it's fine.