fix(#977): add MANAGER_CONTAINER=h-root, migrate root container name

docs/agent-hierarchy.md

@@ -104,13 +104,15 @@ which axis the post-milestone version reads each special-case along:
 
 ### A — naming + bootstrap
 
-- `MANAGER_AGENT = "root"` (broker recipient name) and
-  `MANAGER_NAME = "root"` (container name). Both now `"root"`;
-  renamed from `"manager"`/`"hm1nd"` via the migration script in
-  `migrate.rs` (idempotent, marker-guarded).
+- `MANAGER_AGENT = "root"` (broker recipient name),
+  `MANAGER_NAME = "root"` (logical name, state-dir key), and
+  `MANAGER_CONTAINER = "h-root"` (nixos-container name). The `h-`
+  prefix makes the manager consistent with all sub-agents and lets
+  `lifecycle::list()` use a single `starts_with("h-")` filter.
+  Existing `root` containers are renamed to `h-root` by migration
+  phase 5 in `migrate.rs` (idempotent, marker-guarded).
 - `auto_update::ensure_manager` runs at hive-c0re boot and spawns
-  `root` if missing. Becomes "ensure the root agent exists" once any
-  agent can be at the root. **Topology**: root has no parent, so
+  `h-root` if missing. **Topology**: root has no parent, so
   hive-c0re itself owns its lifecycle (no parent to delegate to).
 
 ### B — wire-protocol privileges
@@ -236,8 +238,9 @@ repos they're working on.
 meta-flake's per-agent flake.nix wrapper). For the manager unit
 that means a hardcoded `HIVE_LABEL` env value:
 
-- `HIVE_LABEL = "root"` — container name; matches what `meta.rs`
-  injects at deploy time.
+- `HIVE_LABEL = "root"` — logical agent name; matches what `meta.rs`
+  injects at deploy time. (The nixos-container is `h-root`, but the
+  harness identifies itself by logical name.)
 
 Real deploys never read these — `meta::render_flake` overrides them
 via the generated wrapper. They exist so the manager