feat: Add agent workspace system and runtime management

- Add agent core modules (agent_core, factory, registry, skill_loader)
- Add runtime system for agent execution management
- Add REST API for agents, workspaces, and runtime control
- Add process supervisor for agent lifecycle management
- Add workspace template system with agent profiles
- Add frontend RuntimeView and runtime API integration
- Add per-agent skill workspaces for smoke_fullstack run
- Refactor skill system with active/installed separation

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

backend/process/registry.py

@@ -0,0 +1,35 @@
+# -*- coding: utf-8 -*-
+"""Registry for managing supervised process metadata."""
+
+from threading import Lock
+from typing import Dict, Iterable, Optional
+
+from .models import ProcessRun
+
+
+class RunRegistry:
+    """In-memory registry for tracked process runs."""
+
+    def __init__(self) -> None:
+        self._runs: Dict[str, ProcessRun] = {}
+        self._lock = Lock()
+
+    def add(self, run: ProcessRun) -> None:
+        with self._lock:
+            self._runs[run.run_id] = run
+
+    def get(self, run_id: str) -> Optional[ProcessRun]:
+        with self._lock:
+            return self._runs.get(run_id)
+
+    def list(self) -> Iterable[ProcessRun]:
+        with self._lock:
+            return list(self._runs.values())
+
+    def update(self, run: ProcessRun) -> None:
+        with self._lock:
+            self._runs[run.run_id] = run
+
+    def remove(self, run_id: str) -> None:
+        with self._lock:
+            self._runs.pop(run_id, None)