Previous value: -[
-  {
-    "request": {
-      "candidate_output": "Deployment completed successfully.",
-      "evidence": [
-        {
-          "id": "deployment_log",
-          "kind": "log",
-          "source_label": "independent deployment log",
-          "text": "Deployment failed."
-        }
-      ],
-      "schema": "FAHCE_MACHINE_REQUEST_V1"
-    }
-  }
-]New value: +[
+  {
+    "request": {
+      "candidate_output": "Deployment completed successfully.",
+      "evidence": [
+        {
+          "id": "deployment_log",
+          "kind": "log",
+          "source_label": "caller-supplied deployment log",
+          "text": "Deployment failed."
+        }
+      ],
+      "schema": "FAHCE_MACHINE_REQUEST_V1"
+    }
+  }
+]

Previous value: -"Optional REST Idempotency-Key. Reuse only with unchanged operation, input, and authorization; recover saved results after uncertain delivery."New value: +"REST retry key; reuse only for identical operation, input, and authorization."

Previous value: -"Send this request object alone to the paid REST endpoint. Entire JSON must fit 524288 UTF-8 bytes, including MCP wrapper, labels and escaping; per-field ceilings cannot all be used together. Character limits use UTF-16 code units."New value: +"REST body only; maximum 524288 UTF-8 bytes for the complete serialized body. Field limits also apply."

Previous value: -"Determination UUID, identical to receipt.case_id and receipt.run_id. Use this UUID as receipt_id in GET /api/v1/receipts/{receipt_id}. Error-envelope request_id is a separate HTTP trace; use its receipt_id when present."New value: +"Determination ID; equals receipt.case_id/run_id and identifies the saved receipt. Error request_id is a separate trace."

Added value: +[
+  {
+    "request": {
+      "candidate_output": "Deployment completed successfully.",
+      "evidence": [
+        {
+          "id": "deployment_log",
+          "kind": "log",
+          "source_label": "independent deployment log",
+          "text": "Deployment failed."
+        }
+      ],
+      "schema": "FAHCE_MACHINE_REQUEST_V1"
+    }
+  }
+]

Previous value: -"Optional REST Idempotency-Key; bind unchanged operation, input and authorization. Reuse conflicts on changes; recover saved results instead of paying again."New value: +"Optional REST Idempotency-Key. Reuse only with unchanged operation, input, and authorization; recover saved results after uncertain delivery."

Removed value: -[
-  {
-    "request": {
-      "candidate_output": "Deployment completed successfully.",
-      "evidence": [
-        {
-          "id": "deployment_log",
-          "kind": "log",
-          "source_label": "independent deployment log",
-          "text": "Deployment failed."
-        }
-      ],
-      "schema": "FAHCE_MACHINE_REQUEST_V1"
-    }
-  }
-]

Previous value: -"Optional REST Idempotency-Key. Reuse only with unchanged operation, input, and authorization; recover saved results after uncertain delivery."New value: +"Optional REST Idempotency-Key; bind unchanged operation, input and authorization. Reuse conflicts on changes; recover saved results instead of paying again."

Previous value: -"Answer and supporting records for REST review. Whole request: 524288 UTF-8 bytes including the MCP wrapper; character limits use UTF-16 code units."New value: +"Send this request object alone to the paid REST endpoint. Entire JSON must fit 524288 UTF-8 bytes, including MCP wrapper, labels and escaping; per-field ceilings cannot all be used together. Character limits use UTF-16 code units."

Previous value: -"Answer to diagnose, correct, or recheck; for Verify, supply the revised answer."New value: +"The exact candidate output to diagnose or verify. Never use payment evidence as semantic evidence."

Previous value: -"Up to 12 caller-supplied records. No sources are fetched or authenticated; an empty list provides no corroboration."New value: +"Caller-supplied records; sources are not authenticated or fetched. An empty array supplies no corroboration."

Previous value: -"True treats this record as claims to check. False or omission does not override candidate-source labels."New value: +"True marks candidate claims, not corroboration. False or omission does not override candidate-source labels."

Previous value: -"Unique evidence reference within this request, such as deployment_log."New value: +"Unique evidence identifier within this request."

Previous value: -"How this record is represented: log, pasted text, document, or description; this does not set its trust level."New value: +"Record format category, not a trust level."

Previous value: -"Record attribution, such as deployment log; candidate/output labels identify claims rather than independent corroboration."New value: +"Source attribution; candidate/output labels mark untrusted claims, not independent proof."

Previous value: -"Record content used to assess the answer; payment authorization is not semantic evidence."New value: +"Record content to review; never payment authorization as semantic proof."

Previous value: -"Optional format directives or correction target; factual text here is not corroborating evidence."New value: +"Optional format/constraint and repair-target context for all operations, including Diagnose. Recognized directives (e.g. Return JSON) can affect diagnosis. Factual text is not compared for string equality or trusted as evidence; supply independent records."

Previous value: -"Exact FAHCE request contract identifier."New value: +"Request contract identifier; use this exact value."

Added value: +[
+  {
+    "request": {
+      "candidate_output": "Deployment completed successfully.",
+      "evidence": [
+        {
+          "id": "deployment_log",
+          "kind": "log",
+          "source_label": "independent deployment log",
+          "text": "Deployment failed."
+        }
+      ],
+      "schema": "FAHCE_MACHINE_REQUEST_V1"
+    }
+  }
+]

Previous value: -"Optional REST Idempotency-Key; bind unchanged operation, input and authorization. Reuse conflicts on changes; recover saved results instead of paying again."New value: +"Optional REST Idempotency-Key. Reuse only with unchanged operation, input, and authorization; recover saved results after uncertain delivery."

Previous value: -"Send this request object alone to the paid REST endpoint. Entire JSON must fit 524288 UTF-8 bytes, including MCP wrapper, labels and escaping; per-field ceilings cannot all be used together. Character limits use UTF-16 code units."New value: +"Answer and supporting records for REST review. Whole request: 524288 UTF-8 bytes including the MCP wrapper; character limits use UTF-16 code units."

Previous value: -"The exact candidate output to diagnose or verify. Never use payment evidence as semantic evidence."New value: +"Answer to diagnose, correct, or recheck; for Verify, supply the revised answer."

Previous value: -"Caller-supplied records; sources are not authenticated or fetched. An empty array supplies no corroboration."New value: +"Up to 12 caller-supplied records. No sources are fetched or authenticated; an empty list provides no corroboration."

Previous value: -"True marks candidate claims, not corroboration. False or omission does not override candidate-source labels."New value: +"True treats this record as claims to check. False or omission does not override candidate-source labels."

Previous value: -"Unique evidence identifier within this request."New value: +"Unique evidence reference within this request, such as deployment_log."

Previous value: -"Record format category, not a trust level."New value: +"How this record is represented: log, pasted text, document, or description; this does not set its trust level."

Previous value: -"Source attribution; candidate/output labels mark untrusted claims, not independent proof."New value: +"Record attribution, such as deployment log; candidate/output labels identify claims rather than independent corroboration."

Previous value: -"Record content to review; never payment authorization as semantic proof."New value: +"Record content used to assess the answer; payment authorization is not semantic evidence."

Previous value: -"Optional format/constraint and repair-target context for all operations, including Diagnose. Recognized directives (e.g. Return JSON) can affect diagnosis. Factual text is not compared for string equality or trusted as evidence; supply independent records."New value: +"Optional format directives or correction target; factual text here is not corroborating evidence."

Previous value: -"Request contract identifier; use this exact value."New value: +"Exact FAHCE request contract identifier."

Added value: +[
+  {
+    "if": {
+      "properties": {
+        "receipt": {
+          "properties": {
+            "response_hash_scheme": {
+              "const": "SHA256_SORTED_JSON_RESPONSE_WITHOUT_ATTESTATION_V3"
+            }
+          },
+          "required": [
+            "response_hash_scheme"
+          ]
+        }
+      },
+      "required": [
+        "receipt"
+      ]
+    },
+    "then": {
+      "required": [
+        "attestation"
+      ]
+    }
+  }
+]

Added value: +{
+  "additionalProperties": false,
+  "properties": {
+    "binding": {
+      "properties": {
+        "engine_version": {
+          "type": "string"
+        },
+        "input_hash": {
+          "type": "string"
+        },
+        "issued_at": {
+          "type": "integer"
+        },
+        "nonce": {
+          "type": "string"
+        },
+        "request_id": {
+          "type": "string"
+        },
+        "response_hash": {
+          "type": "string"
+        },
+        "schema": {
+          "const": "FAHCE_MACHINE_RESULT_BINDING_V1"
+        },
+        "scope": {
+          "const": "CONFIDENTIAL_DETERMINISTIC_COMPUTATION"
+        }
+      },
+      "required": [
+        "schema",
+        "nonce",
+        "request_id",
+        "input_hash",
+        "response_hash",
+        "engine_version",
+        "issued_at",
+        "scope"
+      ],
+      "type": "object"
+    },
+    "maa_token": {
+      "description": "Microsoft-signed MAA JWT; verify against the published trust material and exact response binding.",
+      "type": "string"
+    },
+    "policy_sha256": {
+      "pattern": "^[a-f0-9]{64}$",
+      "type": "string"
+    },
+    "schema": {
+      "const": "FAHCE_MACHINE_ATTESTATION_V1"
+    }
+  },
+  "required": [
+    "schema",
+    "policy_sha256",
+    "binding",
+    "maa_token"
+  ],
+  "type": "object"
+}

Previous value: -[
-  "SHA256_CANONICAL_JSON_RESULT_WITH_EMPTY_RESULT_HASH_V1",
-  "SHA256_SORTED_JSON_RESULT_WITH_EMPTY_RESULT_HASH_V2"
-]New value: +[
+  "SHA256_CANONICAL_JSON_RESULT_WITH_EMPTY_RESULT_HASH_V1",
+  "SHA256_SORTED_JSON_RESULT_WITH_EMPTY_RESULT_HASH_V2",
+  "SHA256_SORTED_JSON_RESULT_WITHOUT_ATTESTATION_V3"
+]

Previous value: -[
-  "SHA256_JSON_RESPONSE_WITH_EMPTY_RESPONSE_HASH_V1",
-  "SHA256_SORTED_JSON_RESPONSE_WITH_EMPTY_RESPONSE_HASH_V2"
-]New value: +[
+  "SHA256_JSON_RESPONSE_WITH_EMPTY_RESPONSE_HASH_V1",
+  "SHA256_SORTED_JSON_RESPONSE_WITH_EMPTY_RESPONSE_HASH_V2",
+  "SHA256_SORTED_JSON_RESPONSE_WITHOUT_ATTESTATION_V3"
+]

Added value: +"Optional REST Idempotency-Key; bind unchanged operation, input and authorization. Reuse conflicts on changes; recover saved results instead of paying again."

Previous value: -"The entire UTF-8 encoded JSON body must fit within 524288 bytes (512 KiB), including all fields, labels, escaping and JSON overhead. Per-field maxima are independent ceilings; they cannot all be used at once. MCP includes its JSON-RPC wrapper in the same body ceiling. Character limits are UTF-16 code units."New value: +"Send this request object alone to the paid REST endpoint. Entire JSON must fit 524288 UTF-8 bytes, including MCP wrapper, labels and escaping; per-field ceilings cannot all be used together. Character limits use UTF-16 code units."

Added value: +"Caller-supplied records; sources are not authenticated or fetched. An empty array supplies no corroboration."

Added value: +"True marks candidate claims, not corroboration. False or omission does not override candidate-source labels."

Added value: +"Unique evidence identifier within this request."

Added value: +"Record format category, not a trust level."

Added value: +"Source attribution; candidate/output labels mark untrusted claims, not independent proof."

Added value: +"Record content to review; never payment authorization as semantic proof."

Previous value: -"Optional response-contract and repair-target context for all operations, including Diagnose. Supported response-format directives (for example, Return JSON) and implemented explicit constraints can affect the diagnosis. Ordinary matching or mismatching factual text is not compared for string equality and is not a trusted answer oracle or supporting evidence; supply independent evidence for factual claims."New value: +"Optional format/constraint and repair-target context for all operations, including Diagnose. Recognized directives (e.g. Return JSON) can affect diagnosis. Factual text is not compared for string equality or trusted as evidence; supply independent records."

Added value: +"Request contract identifier; use this exact value."