fix(cli): honor target_model when resolving custom providers

resolve_runtime_provider() documents target_model as the explicit model
override for mid-session switches and auxiliary slots, but the custom
provider path (_resolve_named_custom_runtime) never received it and
silently substituted the provider's configured default_model instead.

This made auxiliary slots such as auxiliary.background_review silently
run the provider's default model rather than the configured one — e.g. an
ocx-proxy slot configured for gemini-flash actually executed
cursor/claude-sonnet-5, hitting upstream rate limits.

Pass target_model through to the custom runtime resolver and prefer it
over the provider's default model in both the pooled and non-pooled
credential paths.

hermes_cli/runtime_provider.py

@@ -1167,7 +1167,8 @@ def _resolve_named_custom_runtime(
     if pool_result:
         # Propagate the model name even when using pooled credentials —
         # the pool doesn't know about the custom_providers model field.
-        model_name = custom_provider.get("model")
+        # An explicit ``target_model`` wins (same rule as the non-pool path).
+        model_name = target_model or custom_provider.get("model")
         if model_name:
             pool_result["model"] = model_name
         if isinstance(custom_provider.get("max_output_tokens"), int):
@@ -1228,7 +1229,13 @@ def _resolve_named_custom_runtime(
     }
     # Propagate the model name so callers can override self.model when the
     # provider name differs from the actual model string the API expects.
-    if custom_provider.get("model"):
+    # An explicit ``target_model`` wins over the provider's configured
+    # default (regression: auxiliary slots / background-review resolve a
+    # concrete model for a custom provider and must not silently fall back
+    # to ``default_model``).
+    if target_model:
+        result["model"] = target_model
+    elif custom_provider.get("model"):
         result["model"] = custom_provider["model"]
     if isinstance(custom_provider.get("max_output_tokens"), int):
         result["max_output_tokens"] = custom_provider["max_output_tokens"]

tests/hermes_cli/test_runtime_provider_resolution.py

@@ -1677,3 +1677,73 @@ def test_resolve_runtime_provider_opencode_free_missing_env_still_resolves(monke
     assert resolved["provider"] == "opencode-free"
     assert resolved["api_key"] == "opencode-zen-free-keyless"
     assert resolved["base_url"] == "https://opencode.ai/zen/v1"
+
+
+def test_custom_provider_explicit_target_model_wins(monkeypatch):
+    """An explicit target_model must not be silently replaced by the custom
+    provider's configured default model (regression: auxiliary slots such as
+    background-review resolve a concrete model and got default_model instead)."""
+    monkeypatch.setattr(
+        rp,
+        "_get_named_custom_provider",
+        lambda p: {
+            "name": "myproxy",
+            "base_url": "http://127.0.0.1:10100/v1",
+            "api_key": "no-key-required",
+            "model": "default-model",
+        },
+    )
+
+    resolved = rp.resolve_runtime_provider(requested="myproxy", target_model="myproxy/gemini-flash")
+
+    assert resolved is not None
+    assert resolved["provider"] == "custom"
+    assert resolved["model"] == "myproxy/gemini-flash"
+    assert resolved["base_url"] == "http://127.0.0.1:10100/v1"
+
+
+def test_custom_provider_without_target_model_keeps_default(monkeypatch):
+    """No target_model -> the provider's configured model is preserved."""
+    monkeypatch.setattr(
+        rp,
+        "_get_named_custom_provider",
+        lambda p: {
+            "name": "myproxy",
+            "base_url": "http://127.0.0.1:10100/v1",
+            "api_key": "no-key-required",
+            "model": "default-model",
+        },
+    )
+
+    resolved = rp.resolve_runtime_provider(requested="myproxy")
+
+    assert resolved is not None
+    assert resolved["model"] == "default-model"
+
+
+def test_custom_provider_pool_target_model_wins(monkeypatch):
+    """Pooled-credentials path also honors target_model over the default."""
+    monkeypatch.setattr(
+        rp,
+        "_try_resolve_from_custom_pool",
+        lambda *a, **k: {
+            "provider": "custom",
+            "api_key": "pooled-key",
+            "base_url": "http://127.0.0.1:10100/v1",
+        },
+    )
+    monkeypatch.setattr(
+        rp,
+        "_get_named_custom_provider",
+        lambda p: {
+            "name": "myproxy",
+            "base_url": "http://127.0.0.1:10100/v1",
+            "api_key": "no-key-required",
+            "model": "default-model",
+        },
+    )
+
+    resolved = rp.resolve_runtime_provider(requested="myproxy", target_model="myproxy/gemini-flash")
+
+    assert resolved is not None
+    assert resolved["model"] == "myproxy/gemini-flash"