[BUG] Agents UI: status-rail text overlaps agent output pane at left edge

Status Fixed / completed
Reported on v2.1.139
Maintainer reply None cached
Activity 7 comments · opened May 12, 2026 · closed May 14, 2026

Preflight Checklist

  • [x] I have searched existing issues and this hasn't been reported yet
  • [x] This is a single bug report (please file separate reports for different bugs)
  • [x] I am using the latest version of Claude Code

What's Wrong?

In the new agents feature (released ~2026-05-11), the agent output view appears to use a two-pane layout — a narrow status/summary rail on the left and the full agent output on the right. The two panes overlap at the left edge: truncated text from the rail renders on top of, or in the same column as, the first few characters of the agent's full output.

The fragments that bleed through look like the first 1–4 characters of lines from the agent's status summary (e.g. "St", "Th", "Ph", "Ho", "Ke", "Wh1.",
"Wh2.", "1.", "e"), and they obscure the first few characters of the actual output rendered in the main pane.

This only happens in the agents view. Normal Claude Code conversation output renders cleanly. See attached screenshot — Python pseudocode in an agent output pane, with rail fragments overlaying the leftmost column.

What Should Happen?

The status rail and the agent output pane should occupy distinct, non- overlapping regions. The first character of every line in the agent output
pane should be fully readable, with no overlay from the status rail.

Error Messages/Logs

Steps to Reproduce

  1. Spawn one or more agents (e.g. via the Agent tool in a Claude Code session, or any subagent invocation that produces streaming output).
  2. Open the agents view to inspect an agent's output.
  3. While the agent is streaming, or after it finishes, scroll through its output pane.
  4. Observe truncated 1–4 character fragments overlaid on the leftmost column of the main output. The fragments correspond to the starts of lines from the agent's status summary on the left rail.

Easier to reproduce when the agent output contains code blocks or content whose indentation differs from the status summary's text.

Claude Model

Opus

Is this a regression?

No, this never worked

Last Working Version

_No response_

Claude Code Version

2.1.139

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

PowerShell

Additional Information

  • Feature: new agents view, first seen on 2026-05-11.
  • Terminal: Windows Terminal hosting PowerShell 7.6.0.
  • Only the agents view is affected — main Claude Code output renders cleanly.
  • Screenshot attached shows Python pseudocode in an agent output pane with rail fragments ("$V", "St", "Th", "de", "Ph", "Ho", "Ke", "Wh1.", "Wh2.",

"1.", "e") overlaying the leftmost column.

View original on GitHub ↗

7 Comments

saichandunannapanenicopilot · 3 months ago

<img width="199" height="376" alt="Image" src="https://github.com/user-attachments/assets/e42ff633-e963-4ba4-aa32-05ca0c530f4a" />

CaptFaraday · 3 months ago

Reproducing on Windows 11 / Windows Terminal / Claude Code v2.1.139, attached to a backgrounded session via agent view.

Single stray characters bleed through at the left edge of the output pane during normal transcript scrolling — e.g. an isolated R appearing at column 0 ahead of a wrapped line, an Ad floating alone above a recap block, single letters (re, c, m) appearing one or two characters into otherwise-aligned text. The bleed-through characters don't correspond to any text actually in the message; they look like status-rail glyphs from adjacent rows painting underneath the transcript column.

Not tied to a specific tool call or action — happens during normal output. Theme dark-daltonized, tui: "default" in user settings (though agent view forces alt-screen regardless).

saichandunannapanenicopilot · 3 months ago

Another repro on the same version (2.1.139, Windows Terminal / PowerShell). This time the bleed is corrupting prose, not just indented code:

<img width="335" height="36" alt="Image" src="https://github.com/user-attachments/assets/69fe15e9-8228-4281-85c4-97938095f573" />

Rendered: Not "instead of," bct "dlongside":
Actual: Not "instead of," but "alongside":

The rail fragments (c, dl) are landing inside words at the left edge and replacing the real characters, not just overlaying them visually. Confirms this isn't specific to code blocks or unusual indentation.

kanetik · 3 months ago

Seeing the same, and TBH it's making the agents view kinda useless to me. The one saving grace is you can hit < then > to go out and right back in and all is well until you scroll

saichandunannapanenicopilot · 3 months ago

Thanks for the fix @bogini . I am still seeing multiple rendering issues.

Here is an example:

<img width="1176" height="755" alt="Image" src="https://github.com/user-attachments/assets/355c7727-411d-4400-8205-d9c0838c769d" />

<img width="1269" height="742" alt="Image" src="https://github.com/user-attachments/assets/08dee986-eb06-4f94-8ed7-b5a2b76ec1a9" />

CaptFaraday · 3 months ago

Yes I don't know why they close the issue when the same issue is still there in the update

github-actions[bot] · 1 month ago

This issue has been automatically locked since it was closed and has not had any activity for 7 days. If you're experiencing a similar issue, please file a new issue and reference this one if it's relevant.