Sol Ultra spawning Codex CLI tasks (`codex exec`) because “the normal delegation UI does not provide those guarantees as precisely”

What version of the Codex App are you using (From “About Codex” dialog)?

26.707.62119

What subscription do you have?

Pro

What platform is your computer?

Darwin 27.0.0 arm64 arm

What issue are you seeing?

Sol Ultra in Chatdex is spawning Codex CLI tasks (codex exec) because, in its own words, “the normal delegation UI does not provide [the required] guarantees as precisely”

It’s intentional, and only for independent reviews—not implementation. codex exec lets the orchestrator enforce and prove the reviewer’s exact runtime: - Model: gpt-5.6-terra - Reasoning: max - Sandbox: read-only - Working directory: /Users/s/proj/<dirname> - Output: a temporary review report plus persisted runtime evidence That matters because the plan requires an independent reviewer—not the implementing task reviewing its own work—and we wanted auditable proof that it used the correct model, worktree, and permissions. The normal delegation UI does not provide those guarantees as precisely. [emphasis added]

---

Perhaps the worktree and permission configs are correct[?], but I've often seen sessions say similar things like this when asked to spawn agents with specific effort/model/etc. (sometimes followed by functional “surprise” that they were able to do it)

This seems to suggest Codex's “normal delegation UI” should be more explicit with the models about these capabilities to avoid unexpected behavior

What steps can reproduce the bug?

Feedback ID: 019f4841-f0ad-7531-aa47-b460dfbe2088

What is the expected behavior?

The model not being uncertain about its capabilities when spawning new sessions.

Additional information

_No response_