feat(skills): install deepeval skill from confident-ai/deepeval

Adds the deepeval eval-loop skill via `npx skills add` with skills-lock.json
for reproducible reinstalls. Symlinked to Claude Code via .claude/skills/.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016z2ZFYHQCex8yZAMVMTZzZ

.agents/skills/deepeval/templates/metrics.py

@@ -0,0 +1,38 @@
+from deepeval.metrics import (
+    AnswerRelevancyMetric,
+    ContextualRelevancyMetric,
+    StepEfficiencyMetric,
+    TaskCompletionMetric,
+)
+
+
+# Keep metrics in one module so eval files stay focused on app execution.
+# Reuse existing project metrics and thresholds before adding new ones.
+SINGLE_TURN_TRACE_METRICS = [
+    TaskCompletionMetric(),
+    StepEfficiencyMetric(),
+]
+
+SINGLE_TURN_NO_TRACING_METRICS = [
+    AnswerRelevancyMetric(),
+]
+
+MULTI_TURN_METRICS = []
+
+# Component-level metrics are span-specific. Do not create one shared
+# COMPONENT_METRICS list for the whole app. Name each list after the exact
+# component/span it evaluates, then attach it with either:
+# - next_agent_span / next_llm_span / next_tool_span / next_retriever_span
+# - @observe(metrics=[...]) when the integration or manual instrumentation
+#   creates the component span directly.
+RETRIEVER_SPAN_METRICS = [
+    ContextualRelevancyMetric(),
+]
+
+GENERATOR_LLM_SPAN_METRICS = [
+    AnswerRelevancyMetric(),
+]
+
+TOOL_SPAN_METRICS = []
+
+PLANNER_AGENT_SPAN_METRICS = []

.agents/skills/deepeval/templates/test_multi_turn_e2e.py

@@ -0,0 +1,28 @@
+from importlib import import_module
+
+import pytest
+
+from deepeval import assert_test
+from deepeval.dataset import EvaluationDataset
+from deepeval.simulator import ConversationSimulator
+
+from metrics import MULTI_TURN_METRICS
+
+MAX_TURNS = 10
+ai_app = import_module("ai_app")
+
+
+simulator = ConversationSimulator(model_callback=ai_app.chatbot_callback)
+dataset = EvaluationDataset()
+dataset.add_goldens_from_json_file(file_path="tests/evals/.dataset.json")
+
+
+@pytest.mark.parametrize(
+    "test_case",
+    simulator.simulate(
+        conversational_goldens=dataset.goldens,
+        max_user_simulations=MAX_TURNS,
+    ),
+)
+def test_multi_turn(test_case):
+    assert_test(test_case=test_case, metrics=MULTI_TURN_METRICS)

.agents/skills/deepeval/templates/test_single_turn_no_tracing.py

@@ -0,0 +1,32 @@
+from importlib import import_module
+
+import pytest
+
+from deepeval import assert_test
+from deepeval.dataset import EvaluationDataset, Golden
+from deepeval.test_case import LLMTestCase
+
+from metrics import SINGLE_TURN_NO_TRACING_METRICS
+
+
+ai_app = import_module("ai_app")
+
+
+dataset = EvaluationDataset()
+dataset.add_goldens_from_json_file(file_path="tests/evals/.dataset.json")
+
+
+@pytest.mark.parametrize("golden", dataset.goldens)
+def test_single_turn_no_tracing(golden: Golden):
+    actual_output = ai_app.run_ai_app(golden.input)
+    test_case = LLMTestCase(
+        input=golden.input,
+        actual_output=actual_output,
+        expected_output=getattr(golden, "expected_output", None),
+        context=getattr(golden, "context", None),
+        retrieval_context=getattr(golden, "retrieval_context", None),
+    )
+    assert_test(
+        test_case=test_case,
+        metrics=SINGLE_TURN_NO_TRACING_METRICS,
+    )

.agents/skills/deepeval/templates/test_single_turn_tracing.py

@@ -0,0 +1,21 @@
+from importlib import import_module
+
+import pytest
+
+from deepeval import assert_test
+from deepeval.dataset import EvaluationDataset, Golden
+
+from metrics import SINGLE_TURN_TRACE_METRICS
+
+
+ai_app = import_module("ai_app")
+
+
+dataset = EvaluationDataset()
+dataset.add_goldens_from_json_file(file_path="tests/evals/.dataset.json")
+
+
+@pytest.mark.parametrize("golden", dataset.goldens)
+def test_single_turn_tracing(golden: Golden):
+    ai_app.run_traced_ai_app(golden.input)
+    assert_test(golden=golden, metrics=SINGLE_TURN_TRACE_METRICS)