Grade verified task attachments and preserve interrupted report outcomes

Capture bounded text attachments through the task API, verify size and hash, and include their saved content in completion grading and transcripts. Cover further proposal headings without treating interrupted journeys as passes. Keep artifact errors separate from behavior errors and regenerate the native question example.

Co-Authored-By: Paperclip <noreply@paperclip.ing>

server/src/services/native-runtime/runner-api-reference.ts

@@ -245,7 +245,18 @@ export const runnerApiReference: Record<string, { section: string; description?:
                   }
                 ]
               }
-            ]
+            ],
+            "questionSet": {
+              "schema": "paperclip.question_set.v1",
+              "questions": [
+                {
+                  "id": "responsibility",
+                  "prompt": "What should the new agent be responsible for?",
+                  "required": true,
+                  "answerMode": "text"
+                }
+              ]
+            }
           }
         }
       },