feat(learning): add learning runtime with pattern mining

Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)

Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>

backend/app/agents/learning/bridge.py

@@ -0,0 +1,45 @@
+from __future__ import annotations
+
+from app.agents.schemas.learning import LearningDecision, LearningSignal
+
+
+def route_learning_signal(signal: LearningSignal) -> str:
+    if signal.signal_type == "preference":
+        return "memory"
+    if signal.signal_type in {"workflow", "decomposition", "tool_success"}:
+        return "skill"
+    if signal.signal_type == "correction":
+        return "audit"
+    return "memory"
+
+
+def build_learning_bridge_summary(signals: list[LearningSignal]) -> dict[str, object]:
+    memory_count = 0
+    skill_count = 0
+    audit_count = 0
+
+    for signal in signals:
+        route = route_learning_signal(signal)
+        if route == "memory":
+            memory_count += 1
+        elif route == "skill":
+            skill_count += 1
+        else:
+            audit_count += 1
+
+    return {
+        "memory_signal_count": memory_count,
+        "skill_signal_count": skill_count,
+        "audit_signal_count": audit_count,
+    }
+
+
+def update_learning_decision_with_bridge(
+    decision: LearningDecision,
+    signals: list[LearningSignal],
+) -> LearningDecision:
+    bridge_summary = build_learning_bridge_summary(signals)
+    metadata = dict(decision.metadata or {})
+    metadata["bridge"] = bridge_summary
+    decision.metadata = metadata
+    return decision