feat: Phase 23 — Differentiated functionalities

Trace-driven learning pipeline:
- TrainingDataMiner: extract SFT/routing/agent pairs from traces
- LoRATrainer: fine-tune local models from trace-derived data
- AgentConfigEvolver: rewrite agent configs from trace analysis
- LearningOrchestrator: coordinate mine→train→evolve cycle, wired into SystemBuilder

Eval framework (15 real IPW benchmarks):
- Datasets: SuperGPQA, GPQA, MMLU-Pro, MATH-500, Natural Reasoning, HLE,
  SimpleQA, WildChat, IPW, GAIA, FRAMES, SWE-bench, SWEfficiency,
  TerminalBench, TerminalBench Native
- Scorers: MCQ extraction, LLM-judge, exact match, structural validation
- CLI: jarvis eval list|run|compare|report

Composable abstractions:
- Recipe system: TOML composition of all 5 pillars (3 built-in recipes)
- Agent templates: 15 pre-configured TOML manifests with system prompts
- Bundled skills: 20 ready-to-use TOML skill manifests
- Operator recipes: researcher (4h), correspondent (5min), sentinel (2h)

102 files changed, ~11,500 lines added. 3241 tests pass (44 skipped).

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

tests/cli/test_eval_cmd.py

@@ -0,0 +1,39 @@
+"""Tests for the ``jarvis eval`` CLI commands."""
+
+from __future__ import annotations
+
+from click.testing import CliRunner
+
+from openjarvis.cli import cli
+
+
+class TestEvalCLI:
+    """Tests for the eval command group."""
+
+    def test_eval_group_exists(self):
+        """``jarvis eval --help`` shows run/compare/report/list subcommands."""
+        result = CliRunner().invoke(cli, ["eval", "--help"])
+        assert result.exit_code == 0
+        assert "run" in result.output
+        assert "compare" in result.output
+        assert "report" in result.output
+        assert "list" in result.output
+
+    def test_eval_list_benchmarks(self):
+        """``jarvis eval list`` exits 0 and outputs benchmark names."""
+        result = CliRunner().invoke(cli, ["eval", "list"])
+        assert result.exit_code == 0
+        assert "supergpqa" in result.output
+        assert "gaia" in result.output
+        assert "frames" in result.output
+        assert "wildchat" in result.output
+        # Should also show backends
+        assert "jarvis-direct" in result.output
+        assert "jarvis-agent" in result.output
+
+    def test_eval_run_missing_args(self):
+        """``jarvis eval run`` without required args fails gracefully."""
+        result = CliRunner().invoke(cli, ["eval", "run"])
+        # Should fail because neither --config nor --benchmark/--model given
+        assert result.exit_code != 0
+        assert "config" in result.output.lower() or "benchmark" in result.output.lower()