interrupt to get the same treatment too!

docs/core_state_mapping_and_merge.md

@@ -79,6 +79,12 @@ dump only canonical `input` and `output` bindings. `wf_authoring` exposes the
 same canonical binding lists and keeps `in_map`, `input_values`, and `out_map`
 only as deprecated Python-builder sugar that compiles into canonical bindings.
 
+Interrupts follow the same pattern with different field names: `request` uses
+canonical input bindings to build the public interrupt payload, and `resume`
+uses canonical output bindings to write resume payload fields back into state.
+Legacy interrupt `request_map` and `out_map` inputs are accepted only as
+parse compatibility and should not be written by new plans.
+
 ## Explicitness Rules
 
 ### Node-local writes must not overlap

docs/workflow_artifacts.md

@@ -150,16 +150,18 @@ metadata at runtime without becoming a declared schema field.
 Current core interrupt semantics are node-level. An `InterruptNode` has:
 
 - `kind`
-- `request_map`, mapping state/input/context paths to public interrupt payload
-  fields
-- `out_map`, mapping resume payload fields back into state
+- `request`, a canonical input-binding list mapping state/input/context paths
+  or literal values to public interrupt payload fields
+- `resume`, a canonical output-binding list mapping resume payload fields back
+  into workflow state
 - declared resume `outcomes`
 
 That means an artifact can document interrupt boundaries by scanning its
 declarative plan for interrupt nodes and deriving their request/resume payload
-schemas from the maps and workflow state/input schemas. It should not need to
-store unrelated child graph internals just to describe the public interrupt
-points.
+schemas from the bindings and workflow state/input schemas. Legacy
+`request_map` and `out_map` inputs are parse-only compatibility shapes; saved
+artifacts should write `request` and `resume`. They should not need to store
+unrelated child graph internals just to describe the public interrupt points.
 
 ## Composition Rule
 

docs/workflow_drafts.md

@@ -296,17 +296,29 @@ Declares an interrupting step.
 {
   "interrupt": {
     "kind": "input",
-    "request": {
-      "state.question": "question"
-    },
-    "resume": {
-      "answer": "state.answer"
-    },
+    "request": [
+      {
+        "target": {"root": "local", "parts": ["question"]},
+        "path": {"root": "state", "parts": ["question"]}
+      }
+    ],
+    "resume": [
+      {
+        "source": {"root": "local", "parts": ["answer"]},
+        "target": {"root": "state", "parts": ["answer"]}
+      }
+    ],
     "outcomes": ["resumed", "cancelled"]
   }
 }
 ```
 
+Draft interrupts use the same binding shapes as core interrupt nodes:
+`request` builds the public interrupt payload, while `resume` maps the payload
+provided on resume back into workflow state. Older map-shaped `request` and
+`resume` values are accepted only as parse compatibility and dump back to the
+canonical list shape.
+
 Saved interrupting artifacts are still limited in the current execution
 surface. If a deployment reports `interrupting_artifact_unsupported`, that is a
 known platform limitation rather than a draft bug.