Routines: model selector missing from create and edit forms — runs silently fall back to default model

Description

The routines UI (claude.ai/code/routines, also reproduced in the desktop app) no longer offers a model selector anywhere — neither on the New routine creation form nor in the Edit routine modal. The docs still describe one:

"The prompt input includes a model selector. Claude uses the selected model on every run." — https://code.claude.com/docs/en/routines.md

With the selector gone, every routine run silently uses the platform default model. Routines that were originally created with an explicitly selected model (Opus 4.8, extra-high effort) are now firing on Sonnet 5 at max effort with no way to change it back.

Steps to reproduce

1. Open claude.ai/code/routines (or Routines in the desktop app).
2. Click New routine — the creation form has Name, Instructions (with mic/dictation control), repo + environment picker, trigger, and Connectors / Behavior / Notifications tabs. No model selector.
3. Open an existing routine → Edit — same set of controls, no model selector. The Behavior tab contains only the "Auto-fix pull requests" toggle.
4. Let a routine fire — the run uses the default model regardless of what the routine was originally configured with.

Expected

A model (and ideally effort) selector on the routine create/edit form, per the docs — or, failing that, routines honoring the model previously stored in their configuration instead of resetting to the default.

Environment

• claude.ai routines UI and macOS desktop app, observed 2026-07-19
• Routine runs in a cloud environment against a GitHub repo via the Claude_Code_Remote connector

Notes

Possibly related: #48976 (desktop scheduled tasks ignoring model settings, closed not-planned) — but this report is about the cloud routines UI itself losing the selector control.