Suggested test: copilot/auto-model-alias

## Motivation

Link to the gh-aw PR: github/gh-aw#48655 — feat: add `auto` as a proper builtin model alias for Copilot

`auto` was previously missing from the builtin model alias map despite being a known Copilot meta-alias (§9.2). Without it, the AWF proxy had no policy knowledge of `auto`, risking mis-routing. This PR adds `auto` to the alias registry so it is treated like other builtin aliases.

## Proposed test

- **Workflow file**: `test-copilot-auto-model-alias.md`
- **Trigger**: `workflow_dispatch`
- **Engine**: copilot
- **Safe output**: `create-issue` (or `noop`)
- **Variant**: standard

## Minimal test prompt sketch

Declare `model: auto` in the workflow frontmatter. Instruct the agent to create a simple issue (or emit a noop) confirming it ran successfully. The test validates that a workflow using `model: auto` compiles, dispatches, and completes without a `missing_model_pricing` / `unknown_model_ai_credits` error — the regression class fixed in #48655.

## New fixtures or secrets needed

None — `githubnext/gh-aw-test` is sufficient.

## Notes

Open suggestion #9065 (`copilot/evals-model-alias`) is about a different scenario (model alias in evals metrics). Open suggestion #6208 (`test-copilot-model-alias-small`) targets a small/cheap model alias. Neither covers the `auto` meta-alias specifically. This test is the minimal regression guard for the #48655 fix.

> Generated by [🔍 Suggest New E2E Tests](https://github.com/githubnext/gh-aw-test/actions/runs/30439497125) · sonnet46 · 21.2 AIC · ⌖ 4.31 AIC · ⊞ 6.1K · [◷](https://github.com/search?q=repo%3Agithubnext%2Fgh-aw-test+is%3Aissue+%22gh-aw-workflow-call-id%3A+githubnext%2Fgh-aw-test%2Fsuggest-new-e2e-tests%22&type=issues)

Create test-copilot-auto-model-alias.md as the workflow fixture. Use workflow_dispatch, the copilot engine, standard variant, and model: auto in the frontmatter; configure a safe create-issue or noop output. Run it with githubnext/gh-aw-test and confirm it compiles, dispatches, and completes without missing_model_pricing or unknown_model_ai_credits errors.

Written by the indexing model from the issue text.