Codex left WezTerm Lua config broken after edit instead of validating available config parser

What version of Codex is running?

codex-cli 0.114.0

What happened?

During a local Windows config-editing task, Codex modified a WezTerm Lua config and completed the task without running an available validation command. On the next WezTerm startup, the config failed to load with a Lua runtime error:

runtime error: [string "C:\\Users\\<user>\\.wezterm.lua"]:120: wrong number of arguments to 'insert'
stack traceback:
        [C]: in function 'table.insert'
        [string "C:\\Users\\<user>\\.wezterm.lua"]:120: in local 'ssh_command'
        [string "C:\\Users\\<user>\\.wezterm.lua"]:126: in main chunk

The generated helper was assembling SSH command arguments for a WezTerm profile:

local function ssh_command(...)
  local args = {}

  for _, value in ipairs(ssh_prefix) do
    table.insert(args, value)
  end

  for i = 1, select('#', ...) do
    local value = select(i, ...)
    table.insert(args, value)
  end

  return args
end

local ssh_target = ssh_command('-t', ssh_host, remote_shell_command)

A follow-up fix replaced the varargs/table.insert path with explicit list input and args[#args + 1] = value, then validated successfully with:

wezterm --config-file C:\Users\<user>\.wezterm.lua show-keys

Expected behavior

For known application config edits, especially where a fast local parser/validation command is available, Codex should run that validation before completing the task or otherwise surface the risk. In this case, wezterm show-keys would have caught the startup-breaking config immediately.

Actual behavior

Codex left a startup-breaking WezTerm config in place and reported the task as done. The user only discovered the failure by launching WezTerm afterward.

Environment

• OS: Windows / PowerShell
• Codex CLI: 0.114.0
• WezTerm: 20240203-110809-5046fc22

Hostnames, usernames, and private IPs have been redacted from this report.