manager: same lifecycle as agents; auto-spawn on hive-c0re start

hive-c0re/src/coordinator.rs

@@ -118,6 +118,21 @@ impl Coordinator {
         Self::manager_dir().join("mcp.sock")
     }
 
+    /// Ensure a runtime dir + (for sub-agents) per-agent socket exists. For
+    /// the manager, `manager_server::start` owns the socket — just return
+    /// the dir. For sub-agents this is `register_agent` (creates a fresh
+    /// listener bound to `socket_path(name)`). Source directory of the
+    /// `/run/hive/mcp.sock` bind that ends up in `set_nspawn_flags`.
+    pub fn ensure_runtime(&self, name: &str) -> Result<PathBuf> {
+        if name == crate::lifecycle::MANAGER_NAME {
+            let dir = Self::manager_dir();
+            std::fs::create_dir_all(&dir)
+                .with_context(|| format!("create manager dir {}", dir.display()))?;
+            return Ok(dir);
+        }
+        self.register_agent(name)
+    }
+
     /// Per-agent state root (parent of `config/`, future `prompts/`, etc.).
     pub fn agent_state_root(name: &str) -> PathBuf {
         PathBuf::from(format!("{AGENT_STATE_ROOT}/{name}"))