Problem

Four tests/test_tmux_cli_controller.py tests fail on main as well as on
feat/workflow-callback:

• TestCreatePane::test_invalid_pane_id_returns_none
• TestCreatePane::test_valid_pane_id_returned
• TestExecute::test_execute_successful_command
• TestExecute::test_execute_failed_command

Findings

The create-pane tests use a single _run_tmux_command return value, but
create_pane() now lists existing panes and verifies the returned pane. The mocks
no longer model that command sequence. A mocked short value such as %123 or
invalid reaches list_panes(), whose parser assumes at least five pipe-separated
fields and raises IndexError.

The execute tests hard-code markers such as __TMUX_EXEC_START_12345__, while
generate_execution_markers() now creates PID-and-nanosecond-specific markers.
The production parser correctly looks for the generated markers, so the stale
fixtures time out or return unparsed output.

There is also a real robustness gap independent of the stale mocks:
list_panes() should reject malformed or incomplete tmux records instead of
raising IndexError.

Proposed work

• Update create-pane mocks with command-aware side effects covering list, split,
  and verification calls.
• Patch marker generation in execute tests or derive captured output from the
  generated markers.
• Make list_panes() safely skip or report malformed records.
• Add focused malformed-output regression cases.
• Run the complete tmux controller and execution-helper suites.

Acceptance criteria

• All four named tests pass without weakening production verification.
• Malformed tmux output cannot crash list_panes().
• Tests still verify successful output parsing, nonzero exit codes, invalid pane
  IDs, and verified pane creation.
• The full repository suite introduces no new failures.