Flaky test: test_capture_pane_flags[join_wrapped_numbers] matches marker in command echo

Bug

test_capture_pane_flags[join_wrapped_numbers] fails intermittently because the completion marker is detected in the shell command echo line before the command actually executes.

Root cause

The command_complete() check uses:

def command_complete() -> bool:
    output = "\n".join(pane.capture_pane())
    return marker in output

This matches the marker string inside the prompt echo ($ printf ... echo "__DONE_join_wrapped_numbers__") — a false positive. The shell echoes the full command before executing it, so the marker appears in the pane buffer before seq produces any output.

Expected behavior

The marker should only be detected when it appears as command output (its own line), not as part of the typed command echo.

Fix

Change the check to skip lines that start with the shell prompt ($):

def command_complete() -> bool:
    lines = pane.capture_pane()
    return any(marker in line and not line.lstrip().startswith("$") for line in lines)

Reproduction

The failure is timing-dependent. On fast systems seq completes before the next capture, masking the bug. On slower systems or CI, the race is more visible.

Start with test_capture_pane_flags[join_wrapped_numbers] and the command_complete() check shown in the issue. Reproduce the intermittent failure, ensure the marker is detected only from command output rather than the shell command echo, and run the named test to confirm it passes reliably.