[DOCS] Auto mode classifier model selection for external sessions is undocumented

Documentation Type

Missing documentation (feature not documented)

Documentation Location

https://code.claude.com/docs/en/permission-modes

Section/Topic

The Cost and latency section and the surrounding explanation of how the auto mode permission classifier is selected and initialized. The auto mode configuration page should cross-reference the same classifier-model behavior.

Current Documentation

The permission modes page currently says:

The classifier runs on a server-configured model that is independent of your /model selection, so switching models does not change classifier availability. Classifier calls count toward your token usage. Each check sends a portion of the transcript plus the pending action, adding a round-trip before execution.

The page also documents that auto mode supports specific models and providers, but it does not explain how the classifier model is chosen for an external session or when that choice is validated.

What's Wrong or Missing?

Claude Code v2.1.210 improved auto mode so the permission classifier defaults to Sonnet 5 for external sessions, validates that choice on the session's first request, and pins the classifier model for the rest of the session. None of those user-visible rules are documented.

The current wording that the classifier uses a server-configured model does not tell users what “external sessions” means, why the first request can determine whether auto mode is usable, whether later requests can switch classifier models, or how the pinned choice relates to the model selected with /model. Users running auto mode through a third-party provider, gateway, or other external session surface also cannot tell whether Sonnet 5 must be available there or what happens when the first-request validation fails.

Suggested Improvement

Expand the auto mode documentation with a short subsection describing classifier model selection and session lifetime. Explain:

1. Which session types are considered external.
2. That the classifier defaults to Sonnet 5 for those sessions as of Claude Code v2.1.210.
3. That Claude Code validates the classifier model on the session's first request and pins the validated choice for the session.
4. That the classifier model is independent of the conversation model selected with /model.
5. What error or fallback behavior users should expect when the provider or gateway cannot serve the classifier model.

Include the v2.1.210 minimum-version marker and cross-link this explanation from the auto mode configuration page so users can distinguish classifier availability from the main model's availability.

Impact

Medium - Makes feature difficult to understand

Additional Context

Affected Pages:

| Page | Line(s) | Context |
|------|---------|---------|
| https://code.claude.com/docs/en/permission-modes | 200-207 | Auto mode model and provider requirements |
| https://code.claude.com/docs/en/permission-modes | 319-361 | Classifier fallback, model independence, cost, and latency |
| https://code.claude.com/docs/en/auto-mode-config | 5-17, 74-78 | Auto mode classifier configuration and provider context |

Total scope: 2 pages affected

The behavior change is documented in the Claude Code v2.1.210 release notes. The existing pages explain classifier availability, cost, and the distinction from the /model selection, but not the external-session default, first-request validation, or session-level pinning.