Codex App: show effective provider, model, reasoning effort, and service tier for V2 subagents

What variant of Codex are you using?

App

What feature would you like to see?

When the App uses the MultiAgentV2 collaboration surface, show each active and completed subagent's effective runtime configuration in the Subagents UI:

• model provider
• model
• reasoning effort
• service tier

For example:

Worker · openai · gpt-5.6-luna · max · priority

These should be the effective values after role configuration, explicit spawn overrides, inheritance, policy, and fallback resolution—not merely the values requested by the parent.

The information should remain available after completion and after a task is replayed or resumed. If a value is genuinely unavailable, the UI should represent that explicitly instead of silently presenting the parent's configuration as the child's.

This likely requires the V2 app-server collaboration metadata (or another supported V2 child-thread query) to expose all four resolved fields consistently so the App does not have to infer them from the parent or scrape internal rollout state.

Suggested acceptance criteria:

• Active and completed V2 subagents show effective provider, model, reasoning effort, and service tier.
• The values reflect the spawned child's resolved configuration.
• The same values survive live updates, completion, replay, and resume.
• Null/default/inherited values have documented semantics and are rendered without misleading substitutions.
• The App has coverage for explicitly overridden, inherited, fallback, and unavailable values.

Additional information

I searched open and closed issues and PRs before filing. The closest reports are complementary but do not cover this complete App-visible tuple:

• #32283 requests model and reasoning-effort labels in the App's Subagents panel, but not provider or service tier.
• #17312 requests provider/model visibility, but not reasoning effort or service tier and is not scoped to V2.
• #32504 requests provider/model/effort in the canonical MultiAgentV2 activity contract, but not service tier and is primarily the underlying machine-readable contract.
• #35187 requests per-subagent Normal/Fast selection, not display of the resolved configuration.
• #32031 covers V2 spawn-schema discoverability for model/effort/service tier, not effective GUI metadata and not provider.

Showing the complete resolved tuple together is important for cross-provider routing verification, diagnosing inheritance or fallback, and understanding unexpected Fast/Priority usage without opening child storage or inspecting JSONL state.

Start by reading the MultiAgentV2 app-server collaboration metadata and the Subagents UI entry point to trace how child-thread configuration is exposed and rendered. Check the existing App coverage around live updates, completion, replay, and resume; done means active and completed subagents show the resolved provider, model, reasoning effort, and service tier, including explicit handling for unavailable or inherited values.