diff --git a/docs/asynchronous-outbound-messaging-design.md b/docs/asynchronous-outbound-messaging-design.md
index d2b53cfe..ccdcea9d 100644
--- a/docs/asynchronous-outbound-messaging-design.md
+++ b/docs/asynchronous-outbound-messaging-design.md
@@ -24,6 +24,8 @@ protocols.
 
 The implementation must satisfy the following core requirements:
 
+<!-- markdownlint-disable MD013 -->
+
 | ID | Requirement                                                                                                                                            |
 | G1 | Any async task must be able to push frames to a live connection.                                                                                       |
 | G2 | Ordering-safety: Pushed frames must interleave correctly with normal request/response traffic and respect any per-message sequencing rules.            |
@@ -31,6 +33,8 @@ The implementation must satisfy the following core requirements:
 | G4 | Generic—independent of any particular protocol; usable by both servers and clients built on wireframe.                                                 |
 | G5 | Preserve the simple “return a reply” path for code that does not need pushes, ensuring backward compatibility and low friction for existing users.     |
 
+<!-- markdownlint-enable MD013 -->
+
 ## 3. Core Architecture: The Connection Actor
 
 The foundation of this design is the **actor-per-connection** model, where each
@@ -153,8 +157,8 @@ impl<F: FrameLike> PushHandle<F> {
         &self,
         frame: F,
         priority: PushPriority,
-        policy: PushPolicy
-) -> Result<(), PushError>;
+        policy: PushPolicy,
+    ) -> Result<(), PushError>;
 }
 ```
 
@@ -182,41 +186,46 @@ classDiagram
         high_prio_tx: mpsc::Sender<F>
         low_prio_tx: mpsc::Sender<F>
     }
-    class PushHandle {
+    class PushHandle~F~ {
         +push_high_priority(frame: F): Result<(), PushError>
         +push_low_priority(frame: F): Result<(), PushError>
         +try_push(frame: F, priority: PushPriority, policy: PushPolicy): Result<(), PushError>
     }
-    class PushQueues {
+    class PushQueues~F~ {
         +high_priority_rx: mpsc::Receiver<F>
         +low_priority_rx: mpsc::Receiver<F>
-        +bounded(high_capacity: usize, low_capacity: usize): (PushQueues, PushHandle)
+        +bounded(high_capacity: usize, low_capacity: usize): (PushQueues~F~, PushHandle~F~)
         +recv(): Option<(PushPriority, F)>
     }
 
-    PushHandleInner <.. PushHandle : contains
-    PushQueues o-- PushHandle : bounded()
+    PushHandleInner <.. PushHandle~F~ : contains
+    PushQueues~F~ o-- PushHandle~F~ : bounded(high_capacity, low_capacity)
     PushHandle --> PushPriority
     PushHandle --> PushPolicy
     PushHandle --> PushError
-    PushQueues --> PushHandle
-    PushQueues --> FrameLike
-    PushHandle --> FrameLike
 ```
 
+The diagram uses `~F~` to represent the `<F>` generic parameter because Mermaid
+treats angle brackets as HTML.
+
 ```mermaid
 flowchart TD
     Producer[Producer]
-    PushHandle[PushHandle]
+    Handle[PushHandle~F~]
     HighQueue[High Priority Queue]
     LowQueue[Low Priority Queue]
     Policy[PushPolicy]
     Error[PushError or Drop]
 
-    Producer -->|push_high_priority / push_low_priority / try_push| PushHandle
-    PushHandle -->|High| HighQueue
-    PushHandle -->|Low| LowQueue
-    PushHandle -->|If queue full| Policy
+    Producer -->|push_high_priority| Handle
+    Handle -->|priority: High| HighQueue
+    Producer -->|push_low_priority| Handle
+    Handle -->|priority: Low| LowQueue
+
+    Producer -->|try_push| Policy
+    Policy -->|Queue available| Handle
+    Handle -->|priority: High| HighQueue
+    Handle -->|priority: Low| LowQueue
     Policy -->|ReturnErrorIfFull| Error
     Policy -->|DropIfFull| Error
     Policy -->|WarnAndDropIfFull| Error
@@ -342,6 +351,8 @@ features of the 1.0 release.
 
 ## 7. Measurable Objectives & Success Criteria
 
+<!-- markdownlint-disable MD013 -->
+
 | Category        | Objective                                                                                                           | Success Metric                                                                                                                                                                              |
 | API Correctness | The PushHandle, SessionRegistry, and WireframeProtocol trait are implemented exactly as specified in this document. | 100% of the public API surface is present and correctly typed.                                                                                                                              |
 | Functionality   | Pushed frames are delivered reliably and in the correct order of priority.                                          | A test with concurrent high-priority, low-priority, and streaming producers must show that all frames are delivered and that the final written sequence respects the strict priority order. |
@@ -349,3 +360,5 @@ features of the 1.0 release.
 | Resilience      | The SessionRegistry must not leak memory when connections are terminated.                                           | A long-running test that creates and destroys thousands of connections must show no corresponding growth in the SessionRegistry's size or the process's overall memory footprint.           |
 | Performance     | The overhead of the push mechanism should be minimal for connections that do not use it.                            | A benchmark of a simple request-response workload with the push feature enabled (but unused) should show < 2% performance degradation compared to a build without the feature.              |
 | Performance     | The latency for a high-priority push under no contention should be negligible.                                      | The time from push_high_priority().await returning to the frame being written to the socket buffer should be < 10µs.                                                                        |
+
+<!-- markdownlint-enable MD013 -->