You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Thinking/reasoning text is unreadable on dark backgrounds (hardcoded dim color) #3866
The "Thinking…" reasoning text displayed during model inference is nearly invisible on dark terminal backgrounds. The text appears to use a hardcoded dark gray foreground that has insufficient contrast against dark backgrounds.
This started with a recent update (noticed on 1.0.64).
Environment
OS: Windows 11 + WSL2
Terminal: Windows Terminal (dark color scheme)
Multiplexer: tmux
Copilot CLI version: 1.0.64
Model: Claude Opus 4.6 (extended thinking enabled)
Steps to reproduce
Use a dark terminal color scheme (default Windows Terminal dark, or similar)
Run copilot in tmux
Ask a question that triggers extended thinking/reasoning
Observe the "Thinking…" block text is nearly invisible
Expected behavior
Thinking/reasoning text should have sufficient contrast on both dark and light backgrounds. Ideally it would:
Use a semantic ANSI color (e.g., brightBlack / color 8) rather than a hardcoded RGB
Or respect the terminal's dim attribute rendering
Or be configurable via a theme/config setting
Workaround
Bumping brightBlack in the Windows Terminal color scheme to #808080 or lighter partially helps, but this affects all dim text globally.
I did some deep diving with strace and found out that this issue is highly relevant to OSC 4.
Here is my test matrix:
✅ Working properly:
Arch Linux + Kitty + tmux (3.6b)
WSL + Windows Terminal (without tmux)
❌ Reproduces the issue (Bad):
WSL + Windows Terminal + tmux (3.5a) — Completely no OSC 4 forwarding.
WSL + Windows Terminal + tmux (3.6a/3.6b) — OSC 4 is forwarded, but the sequence is broken. The OSC 4 response is out of order and incomplete.
WSL + Alacritty + tmux (3.6a/3.6b) — (Not fully analyzed yet. I included this as a control group to rule out Windows Terminal specific issues).
Conclusion:
Based on these tests, the problem seems to be related to how tmux interacts with certain components in WSL/Windows (possibly ConPTY). At this moment, it's hard to say definitively whether the root cause lies within tmux or Windows.
Unfortunately, I don't have the time to do further debugging on this right now, but I hope these findings help point in the right direction!
Additional Note & Suggestion for Copilot CLI
My suggestion for the copilot-cli would be to avoid relying strictly on OSC 4. Simply querying the default foreground/background color is usually enough to infer whether the terminal is using a light or dark theme, which would allow the CLI to apply an appropriate color scheme.
For comparison, both codex-cli and claude-code are able to render colors perfectly fine even when OSC 4 is unavailable. Their auto modes can successfully detect the terminal's light/dark mode and apply the correct configuration automatically. Falling back to a similar approach in copilot-cli would make it much more robust across different terminal environments.
Current Workaround
You can explicitly configure pane-colours in your tmux config. This prevents tmux from forwarding OSC 4 to the outer terminal and instead returns the values directly to the app. I've verified this works.
Honestly, I‘m really not for this workaround, because it”s not KISS(repeating the palette from terminal to tmux). Just in any case you really want to make copilot-cli happy :)
This appears to be resolved as of GitHub Copilot CLI v1.0.95. Please update to that release or a newer stable version. We're closing this as completed. If it still happens, please comment with your version and reproduction steps.
Activity
X1aomu commentedon Jun 23, 2026
Debugging
I did some deep diving with
straceand found out that this issue is highly relevant to OSC 4.Here is my test matrix:
✅ Working properly:
❌ Reproduces the issue (Bad):
Conclusion:
Based on these tests, the problem seems to be related to how tmux interacts with certain components in WSL/Windows (possibly ConPTY). At this moment, it's hard to say definitively whether the root cause lies within tmux or Windows.
Unfortunately, I don't have the time to do further debugging on this right now, but I hope these findings help point in the right direction!
Additional Note & Suggestion for Copilot CLI
My suggestion for the copilot-cli would be to avoid relying strictly on OSC 4. Simply querying the default foreground/background color is usually enough to infer whether the terminal is using a light or dark theme, which would allow the CLI to apply an appropriate color scheme.
For comparison, both
codex-cliandclaude-codeare able to render colors perfectly fine even when OSC 4 is unavailable. Theirautomodes can successfully detect the terminal's light/dark mode and apply the correct configuration automatically. Falling back to a similar approach incopilot-cliwould make it much more robust across different terminal environments.Current Workaround
You can explicitly configure pane-colours in your tmux config. This prevents tmux from forwarding OSC 4 to the outer terminal and instead returns the values directly to the app. I've verified this works.
Honestly, I‘m really not for this workaround, because it”s not KISS(repeating the palette from terminal to tmux). Just in any case you really want to make copilot-cli happy :)
Here is the example tmux conf:
copilot-cli-bot commentedon Oct 10, 2026
This appears to be resolved as of GitHub Copilot CLI v1.0.95. Please update to that release or a newer stable version. We're closing this as completed. If it still happens, please comment with your version and reproduction steps.