docs(graph): clarify runtime guard wording and nested paths for agent inputs

Graph-Agents.md

@@ -309,7 +309,9 @@ resolved strictly (a missing key fails the node before the child starts, with
 - A value that is exactly one reference (`"{{inner_width}}"`) hands the
   parent's JSON value through unchanged: numbers stay numbers, lists stay
   lists, objects stay objects. Scripts in the child see the real type;
-  the child's own templates stringify it the usual way.
+  the child's own templates stringify it the usual way. Nested paths work
+  the same way (`"{{limits.width}}"` hands through `limits.width` with its
+  type intact).
 - Anything else (`"Repo: {{repo}}"`, `"{{a}}-{{b}}"`) renders to a string.
 - **`null` is a value, not an absence.** A lone `"{{k}}"` whose parent value
   is `null` overlays the child's default with `null`. Because `state_updates`
@@ -318,7 +320,11 @@ resolved strictly (a missing key fails the node before the child starts, with
   `x: null`, not its own default.
 
 **Two things are rejected**, both at load time by the validator and again at
-runtime for graphs running with `validate_before_run: false`:
+runtime for graphs running with `validate_before_run: false`. The validator
+messages below name the node; the runtime guards name the agent instead
+(`agent '<name>' has no graph.yaml; …` / `agent '<name>': `inputs.initial_prompt`
+is reserved …`) and arrive wrapped in the usual `at node '<id>': Agent '<name>'
+failed:` context, so the node is still recoverable from the error chain.
 
 - `inputs:` on a node whose `agent` has a `config.yaml` but no `graph.yaml`
   (`agent node '<id>': `inputs:` requires agent '<name>' to be a graph agent