@armoriq/policy-ir 0.1.0

package/vectors/schema-hash-vectors.json

@@ -0,0 +1,497 @@
+{
+  "_comment": [
+    "Golden vectors for the tool-inventory canonicalizer (Tool Inventory PRD T4, armoriq/conmap-auto#968).",
+    "Three implementations must agree byte-for-byte: conmap src/tool-inventory/canonical.ts,",
+    "armoriq-sdk-customer (python) and armoriq-sdk-customer-ts. conmap is the source of truth;",
+    "each SDK keeps a byte-identical copy of this file plus its sha256 in a sibling .sha256, and a",
+    "test in each repo asserts the file still hashes to it -- so changing a vector forces all three",
+    "copies to move together.",
+    "canonicalJson = sorted keys, compact separators, ensure_ascii escaping. This is NOT RFC 8785;",
+    "see the header of canonical.ts for why, and vector 5 for the number forms left outside the",
+    "agreed set."
+  ],
+  "algorithm": {
+    "canonicalJson": "sorted keys, no whitespace, ASCII-escaped (json.dumps sort_keys=True separators=(',',':') ensure_ascii=True)",
+    "schemaHash": "'sha256:' + sha256_hex(canonicalJson(strip description/title only where they are annotation keywords (never as names under properties/patternProperties/$defs/definitions/dependentSchemas or strings in required, never inside enum/const/default/examples values), resolve local #/ $ref)); null when inputSchema is null",
+    "manifestHash": "'sha256:' + sha256_hex(canonicalJson({name, description|null, annotations|{}, inputSchema|null}))"
+  },
+  "vectors": [
+    {
+      "id": 1,
+      "name": "key-order-irrelevant",
+      "note": "Object keys are sorted, so two spellings that differ only in key order produce identical bytes.",
+      "inputSchema": {
+        "type": "object",
+        "required": [
+          "b"
+        ],
+        "properties": {
+          "b": {
+            "type": "string"
+          },
+          "a": {
+            "type": "number"
+          }
+        }
+      },
+      "schemaHash": "sha256:597b8b01f9b5b640dd6856455c6156c631bb3e36768be54f79f21c89392c064d",
+      "equivalentSpellings": [
+        {
+          "properties": {
+            "a": {
+              "type": "number"
+            },
+            "b": {
+              "type": "string"
+            }
+          },
+          "required": [
+            "b"
+          ],
+          "type": "object"
+        }
+      ]
+    },
+    {
+      "id": 2,
+      "name": "nested-description-and-title-stripped",
+      "note": "description and title are removed wherever they are schema annotation keywords, so prose-only edits never move schemaHash. Equals vector 1.",
+      "inputSchema": {
+        "title": "Thing",
+        "description": "top level prose",
+        "type": "object",
+        "required": [
+          "b"
+        ],
+        "properties": {
+          "a": {
+            "type": "number",
+            "description": "the a"
+          },
+          "b": {
+            "title": "B",
+            "type": "string",
+            "description": "the b"
+          }
+        }
+      },
+      "schemaHash": "sha256:597b8b01f9b5b640dd6856455c6156c631bb3e36768be54f79f21c89392c064d",
+      "equalsVector": 1
+    },
+    {
+      "id": 3,
+      "name": "local-ref-resolved",
+      "note": "A local #/$defs pointer is inlined before hashing, so spelling a shape out and pointing at it hash the same.",
+      "inputSchema": {
+        "type": "object",
+        "$defs": {
+          "Id": {
+            "type": "string",
+            "pattern": "^[a-z]+$"
+          }
+        },
+        "properties": {
+          "id": {
+            "$ref": "#/$defs/Id"
+          }
+        }
+      },
+      "schemaHash": "sha256:ed8f92e2b3119b897b4c3fb2322e4a94aefb4e345b2520c95ef96459ff532073"
+    },
+    {
+      "id": 4,
+      "name": "unicode-including-astral-plane",
+      "note": "Non-ASCII escapes as \\uXXXX; an astral-plane character (U+1F600, U+1F680) escapes as its UTF-16 surrogate pair, matching Python ensure_ascii and JS charCodeAt iteration. All keys here are ASCII on purpose: JS sorts keys by UTF-16 code unit and Python by code point, which differ for astral-plane KEYS, so those are outside the agreed set.",
+      "inputSchema": {
+        "type": "string",
+        "enum": [
+          "caf\u00e9",
+          "\u65e5\u672c\u8a9e",
+          "\ud83d\ude00"
+        ],
+        "const": "na\u00efve \ud83d\ude80"
+      },
+      "schemaHash": "sha256:bd1abce15a0b8bcd6e5c0571a22858e1e14d889ed9ffc68fc2c06f4c5c68be40"
+    },
+    {
+      "id": 5,
+      "name": "number-forms",
+      "note": "Numbers are emitted by the host JSON writer, not RFC 8785. Integers and decimals are distinct canonical bytes.",
+      "knownDivergence": "A float-valued integer diverges across languages (Python json.dumps(1.0) -> \"1.0\", JSON.stringify(1.0) -> \"1\"), as does -0.0 (Python \"-0.0\", JS \"0\"). Neither appears in a real JSON Schema, and neither is in this vector: agreeing on them requires a shared number formatter, which is the RFC 8785 migration, not a patch here.",
+      "inputSchema": {
+        "type": "number",
+        "multipleOf": 1,
+        "minimum": 0,
+        "maximum": 10,
+        "default": 2.5,
+        "exclusiveMinimum": -1
+      },
+      "schemaHash": "sha256:bcafbb074da1b4c0217c2438646e132ab73398dc1570efd186017abdf21f5029"
+    },
+    {
+      "id": 6,
+      "name": "empty-object-schema",
+      "note": "An empty schema {} is a real, hashable schema. It is NOT the same fact as 'no schema' (vector 7).",
+      "inputSchema": {},
+      "schemaHash": "sha256:44136fa355b3678a1146ad16f7e8649e94fb4fc21fe77e8310c060f61caaff8a"
+    },
+    {
+      "id": 7,
+      "name": "null-schema-has-no-hash",
+      "note": "No schema at all: schemaHash is null, never the hash of {} and never the hash of the string 'null'. FR-7 warns on exactly this row.",
+      "inputSchema": null,
+      "schemaHash": null,
+      "manifestDoc": {
+        "name": "unschemad_tool",
+        "description": "declared with a name only"
+      },
+      "manifestHash": "sha256:61f5054dde7e535542c2e04a739dc78d1341504f798c7642ab9ad29f1b456bee"
+    },
+    {
+      "id": 8,
+      "name": "array-order-is-significant",
+      "note": "Canonicalization sorts object keys and never sequences: reordering prefixItems is a different schema.",
+      "inputSchema": {
+        "type": "array",
+        "prefixItems": [
+          {
+            "type": "string"
+          },
+          {
+            "type": "number"
+          },
+          {
+            "type": "boolean"
+          }
+        ],
+        "items": {
+          "anyOf": [
+            {
+              "type": "null"
+            },
+            {
+              "type": "string"
+            }
+          ]
+        }
+      },
+      "schemaHash": "sha256:9dd8bbfb2bc524842d1203e4f914a5743cceaa228fcf328182f4bc9260fb747e"
+    },
+    {
+      "id": 9,
+      "name": "nested-required-arrays",
+      "note": "required lists nested inside properties survive verbatim, order included, at every depth.",
+      "inputSchema": {
+        "type": "object",
+        "required": [
+          "outer"
+        ],
+        "properties": {
+          "outer": {
+            "type": "object",
+            "required": [
+              "z",
+              "a"
+            ],
+            "properties": {
+              "z": {
+                "type": "string"
+              },
+              "a": {
+                "type": "object",
+                "required": [
+                  "q"
+                ],
+                "properties": {
+                  "q": {
+                    "type": "integer"
+                  }
+                }
+              }
+            }
+          }
+        }
+      },
+      "schemaHash": "sha256:864f94090821218ef480510c415eaac315715d2f8ee3be1cbc7517594b51ab13"
+    },
+    {
+      "id": 10,
+      "name": "booleans-nulls-and-escaped-strings",
+      "note": "true/false/null literals, and a string carrying a quote, a backslash, a tab, a newline and a control character.",
+      "inputSchema": {
+        "type": "object",
+        "additionalProperties": false,
+        "deprecated": true,
+        "properties": {
+          "q": {
+            "type": "string",
+            "default": "he said \"hi\"\tone\\two\nthree\u0001",
+            "const": null
+          }
+        }
+      },
+      "schemaHash": "sha256:f5cdf2e3025ce09cb931b6cc918484a14de97a057e9667874593f0a3bc50877c"
+    },
+    {
+      "id": 11,
+      "name": "argument-named-description-is-kept",
+      "note": "`description` and `title` are stripped only as schema annotation keywords, never as argument names under `properties`, `required`, `patternProperties`, `$defs`, `definitions` or `dependentSchemas`. create_ticket(title, description, priority) must NOT hash like create_ticket(priority). The annotation on `priority` is stripped; the two argument names are kept. `equivalentSpellings` differs only in annotation prose and must hash identically; `distinctSpelling` drops the two arguments and must NOT.",
+      "inputSchema": {
+        "type": "object",
+        "properties": {
+          "title": {
+            "type": "string",
+            "title": "Title",
+            "description": "Short summary"
+          },
+          "description": {
+            "type": "string",
+            "description": "Long form body"
+          },
+          "priority": {
+            "type": "integer",
+            "description": "P0-P3"
+          }
+        },
+        "required": [
+          "title",
+          "description"
+        ]
+      },
+      "schemaHash": "sha256:996c1f968a21b24a3f171a432a3de33257901beb2d172e3dfa3d345b7e3e6a3c",
+      "equivalentSpellings": [
+        {
+          "type": "object",
+          "properties": {
+            "title": {
+              "type": "string"
+            },
+            "description": {
+              "type": "string",
+              "title": "Body"
+            },
+            "priority": {
+              "type": "integer"
+            }
+          },
+          "required": [
+            "title",
+            "description"
+          ]
+        }
+      ],
+      "distinctSpelling": {
+        "type": "object",
+        "properties": {
+          "priority": {
+            "type": "integer"
+          }
+        }
+      }
+    },
+    {
+      "id": 12,
+      "name": "real-mcp-input-schema",
+      "note": "A verbatim MCP tools/list inputSchema (GitHub MCP create_issue), prose included, with the manifestHash of the full tool entry.",
+      "inputSchema": {
+        "type": "object",
+        "properties": {
+          "owner": {
+            "type": "string",
+            "description": "Repository owner"
+          },
+          "repo": {
+            "type": "string",
+            "description": "Repository name"
+          },
+          "title": {
+            "type": "string",
+            "description": "Issue title"
+          },
+          "body": {
+            "type": "string",
+            "description": "Issue body content"
+          },
+          "assignees": {
+            "type": "array",
+            "items": {
+              "type": "string"
+            },
+            "description": "Usernames to assign"
+          },
+          "labels": {
+            "type": "array",
+            "items": {
+              "type": "string"
+            },
+            "description": "Labels to apply"
+          },
+          "milestone": {
+            "type": "number",
+            "description": "Milestone number"
+          }
+        },
+        "required": [
+          "owner",
+          "repo",
+          "title"
+        ],
+        "additionalProperties": false,
+        "$schema": "http://json-schema.org/draft-07/schema#"
+      },
+      "schemaHash": "sha256:b06130f66f945f4ca2bc4474340c7f934c5e7b10168978fb8f507ecc146c091b",
+      "manifestDoc": {
+        "name": "create_issue",
+        "description": "Create a new issue in a GitHub repository",
+        "annotations": {
+          "readOnlyHint": false,
+          "destructiveHint": false,
+          "idempotentHint": false
+        },
+        "inputSchema": {
+          "type": "object",
+          "properties": {
+            "owner": {
+              "type": "string",
+              "description": "Repository owner"
+            },
+            "repo": {
+              "type": "string",
+              "description": "Repository name"
+            },
+            "title": {
+              "type": "string",
+              "description": "Issue title"
+            },
+            "body": {
+              "type": "string",
+              "description": "Issue body content"
+            },
+            "assignees": {
+              "type": "array",
+              "items": {
+                "type": "string"
+              },
+              "description": "Usernames to assign"
+            },
+            "labels": {
+              "type": "array",
+              "items": {
+                "type": "string"
+              },
+              "description": "Labels to apply"
+            },
+            "milestone": {
+              "type": "number",
+              "description": "Milestone number"
+            }
+          },
+          "required": [
+            "owner",
+            "repo",
+            "title"
+          ],
+          "additionalProperties": false,
+          "$schema": "http://json-schema.org/draft-07/schema#"
+        }
+      },
+      "manifestHash": "sha256:a0bc53dbf453b8c5860924781ddc912d1b118a64adb3ee4b5089bf599b936113"
+    },
+    {
+      "id": 13,
+      "name": "data-literals-are-not-schemas",
+      "note": "The values of `enum`, `const`, `default` and `examples` are data, not schema nodes: a `title` or `description` key inside them is part of the value and is kept verbatim. `equivalentSpellings` reorders keys inside the literals only; `distinctSpelling` removes those keys from the literals and must NOT hash equal.",
+      "inputSchema": {
+        "type": "object",
+        "properties": {
+          "payload": {
+            "type": "object",
+            "default": {
+              "title": "Untitled",
+              "description": "",
+              "priority": 2
+            },
+            "examples": [
+              {
+                "title": "First",
+                "description": "Body"
+              }
+            ]
+          },
+          "kind": {
+            "type": "object",
+            "enum": [
+              {
+                "title": "bug"
+              },
+              {
+                "title": "task"
+              }
+            ],
+            "const": {
+              "description": "fixed"
+            }
+          }
+        }
+      },
+      "schemaHash": "sha256:cae41d7dfcc5d51cf9ce1756903979b1fc5754fa8e95a69d9e5bec5df0266dc8",
+      "equivalentSpellings": [
+        {
+          "properties": {
+            "kind": {
+              "const": {
+                "description": "fixed"
+              },
+              "enum": [
+                {
+                  "title": "bug"
+                },
+                {
+                  "title": "task"
+                }
+              ],
+              "type": "object"
+            },
+            "payload": {
+              "examples": [
+                {
+                  "description": "Body",
+                  "title": "First"
+                }
+              ],
+              "default": {
+                "priority": 2,
+                "description": "",
+                "title": "Untitled"
+              },
+              "type": "object"
+            }
+          },
+          "type": "object"
+        }
+      ],
+      "distinctSpelling": {
+        "type": "object",
+        "properties": {
+          "payload": {
+            "type": "object",
+            "default": {
+              "priority": 2
+            },
+            "examples": [
+              {}
+            ]
+          },
+          "kind": {
+            "type": "object",
+            "enum": [
+              {},
+              {}
+            ],
+            "const": {}
+          }
+        }
+      }
+    }
+  ]
+}