feat(delegate): support per-task model and reasoning_effort in delegate_task

[neaucode-bot]

delegate_task now accepts top-level and per-task model and reasoning_effort. Per-task values resolve in order: per-task -> delegation.reasoning_effort config -> parent inheritance. _dispatch_delegate_task forwards both so they reach _build_child_agent, and DELEGATE_TASK_SCHEMA exposes the new fields on both the top-level and per-task shapes.

Covers schema, dispatch, and child-build paths with tests.

Made with Cursor

[alt-glitch]

This was generated by AI during triage.

Duplicate of #50465 — both add per-task model and reasoning_effort to delegate_task (same run_agent.py dispatch + tools/delegate_tool.py schema paths). #50465 is the earlier, more comprehensive open PR (also adds per-task credential resolution for batch fan-out); this is the narrower subset in the same saturated per-call-model-override cluster. Marking duplicate so reviewers consolidate on the superset PR.

[teknium1]

Thanks for the focused implementation and tests.

This automated hermes-sweeper review is closing this under the standing delegation-model-routing policy: delegate_task does not support model-facing per-call or per-task model/provider/reasoning routing. The supported control is the profile-level delegation.provider, delegation.model, and delegation.reasoning_effort configuration.

• Current main states this contract directly in tools/delegate_tool.py:3296.
• The configuration path is documented in website/docs/user-guide/configuration.md:1936-1944.
• This is the same feature cluster as feat(delegation): support per-call child model overrides #50465, where a maintainer explicitly rejected the approach (We do not want this, July 5, 2026).

If a different delegation behavior is needed, please re-scope it around the existing delegation.* configuration mechanism rather than adding per-call routing fields to the tool schema.

---

Closed as not-planned per standing maintainer policy (delegation-model-routing). This is a design-direction decision, not a code-quality judgment — see the Contribution Rubric in AGENTS.md for what the project is looking for. If you believe this policy was misapplied to your change, comment here and a maintainer will take a look.