Codex CLI 0.145.0: spawn_agent rejects gpt-5.6-luna even though /model lists it

What variant of Codex are you using?

Codex CLI / TUI 0.145.0 (session_meta.cli_version = "0.145.0"), OpenAI provider, Ubuntu 24.04.4 under WSL2.

What happened?

The top-level Select Model and Effort picker lists all three GPT-5.6 variants, including:

• gpt-5.6-sol
• gpt-5.6-terra
• gpt-5.6-luna

Native multi-agent spawning accepts Sol and Terra model/effort overrides, but rejects Luna as unknown.

Reproduction

From one live Codex CLI 0.145.0 parent thread, call native spawn_agent with a non-full-history fork:

{
  "task_name": "luna_medium_probe",
  "fork_turns": "none",
  "model": "gpt-5.6-luna",
  "reasoning_effort": "medium",
  "message": "Return exactly LUNA_MEDIUM_NATIVE_OK"
}

Actual tool result:

Unknown model `gpt-5.6-luna` for spawn_agent. Available models: gpt-5.6-sol, gpt-5.6-terra

Positive controls from the same parent thread succeeded and returned terminal child receipts:

{"fork_turns":"none","model":"gpt-5.6-sol","reasoning_effort":"max"}
{"fork_turns":"none","model":"gpt-5.6-terra","reasoning_effort":"low"}

Their children returned SOL_MAX_NATIVE_OK and TERRA_LOW_NATIVE_OK, respectively.

Expected behavior

Because the active top-level model catalog exposes gpt-5.6-luna, and 0.145.0 advertises configurable sub-agent models/reasoning levels, spawn_agent should accept Luna when it is supported by the current account/backend, or explain why the sub-agent catalog is intentionally narrower than the top-level catalog.

Additional information

This is specifically about native spawn_agent model selection, not codex exec --model ... or top-level session selection.