Pass the capability matrix's models as a pytest option

NISSE_CAPABILITY_MODELS was neither a secret nor configuration: which
models to measure is a question for one run, so it is now an argument,
--capability-models=a,b, declared in tests/conftest.py where pytest
reads options at startup. With no models the matrix skips as before, so
CI is unchanged. The last non-secret NISSE_* variable other than
NISSE_CONFIG and NISSE_ENV is gone.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

tests/integration/test_model_capability.py

@@ -9,11 +9,15 @@ What it measures is the thing that actually breaks: whether a model can hold
 the tool-call contract our loops depend on. llama3.2 could not — it returned
 `{` as an entire plan — and no amount of prompt work fixed it.
 
-Opt in with NISSE_CAPABILITY_MODELS, comma separated. Skipped otherwise,
-because it spends real tokens across several models.
-"""
+Opt in by naming the models to measure, for one run:
 
-import os
+    uv run pytest -m integration tests/integration/test_model_capability.py \
+        --capability-models=local-fast,local-coder
+
+Skipped otherwise, because it spends real tokens across every model listed.
+Which models to measure is a question for a particular run, so it is an
+argument, not configuration.
+"""
 
 import pytest
 
@@ -31,15 +35,28 @@ from nisse_dev.llm import AnthropicPassthroughClient, LiteLLMClient, LLMError
 from nisse_dev.planner import PLANNER_TOOLS
 from nisse_dev.tools import EXECUTOR_TOOLS
 
-SENTINEL = "NISSE_CAPABILITY_MODELS"
-MODELS = [m.strip() for m in os.environ.get(SENTINEL, "").split(",") if m.strip()]
+pytestmark = pytest.mark.integration
 
-pytestmark = [
-    pytest.mark.integration,
-    pytest.mark.skipif(
-        not MODELS, reason=f"set {SENTINEL} to a comma-separated model list"
-    ),
-]
+
+def pytest_generate_tests(metafunc: pytest.Metafunc) -> None:
+    if "model" not in metafunc.fixturenames:
+        return
+    raw = metafunc.config.getoption("--capability-models")
+    models = [m.strip() for m in raw.split(",") if m.strip()]
+    if models:
+        metafunc.parametrize("model", models)
+    else:
+        metafunc.parametrize(
+            "model",
+            [
+                pytest.param(
+                    None,
+                    marks=pytest.mark.skip(
+                        reason="no models to measure; pass --capability-models"
+                    ),
+                )
+            ],
+        )
 
 
 @pytest.fixture
@@ -60,7 +77,6 @@ def user(text: str) -> list[StoredEvent]:
     return [StoredEvent(id=1, ts="t", event=UserMessage(text=text, source="probe"))]
 
 
-@pytest.mark.parametrize("model", MODELS)
 def test_model_answers_at_all(model: str, catalog: ModelCatalog, routes) -> None:
     normalized, passthrough = routes
     client = select_client(model, normalized, passthrough, catalog)
@@ -71,7 +87,6 @@ def test_model_answers_at_all(model: str, catalog: ModelCatalog, routes) -> None
     assert texts, f"{model} produced no text: {produced!r}"
 
 
-@pytest.mark.parametrize("model", MODELS)
 def test_model_holds_the_planner_contract(
     model: str, catalog: ModelCatalog, routes
 ) -> None:
@@ -100,7 +115,6 @@ def test_model_holds_the_planner_contract(
         )
 
 
-@pytest.mark.parametrize("model", MODELS)
 def test_model_round_trips_a_tool_result(
     model: str, catalog: ModelCatalog, routes
 ) -> None:
@@ -130,7 +144,6 @@ def test_model_round_trips_a_tool_result(
     assert follow_up, f"{model} produced nothing after its tool result"
 
 
-@pytest.mark.parametrize("model", MODELS)
 def test_model_holds_the_prompt_level_json_contract(
     model: str, catalog: ModelCatalog, routes
 ) -> None:
@@ -154,7 +167,6 @@ def test_model_holds_the_prompt_level_json_contract(
     )
 
 
-@pytest.mark.parametrize("model", MODELS)
 def test_route_selection_matches_the_model(
     model: str, catalog: ModelCatalog, routes
 ) -> None: