Agent Registry

2026/11/19 - Agent Registry - 3 updated api methods

Changes  BatchGetDiscoverableRegistryRecord returns a short-lived download URL and SHA-256 checksum for approved skill packages. SearchDiscoverableRegistryRecords adds searchType (KEYWORD, SEMANTIC, HYBRID) and pagination. Adds bearer token authentication for registries that use JWT authorization.

BatchGetDiscoverableRegistryRecord (updated) Link ¶
Changes (response)
{'registryRecords': {'artifact': {'downloadUrl': 'string', 'sha256': 'string'},
                     'status': {'UPDATE_PENDING_AUTH', 'CREATE_PENDING_AUTH'}}}

Retrieves multiple discoverable registry records by ID from a single registry. Records that cannot be retrieved are reported individually in the errors list rather than failing the entire request. For each skill record that has a skill package, the operation also returns artifact. It contains a short-lived download URL and a SHA-256 checksum for the package of the record's latest approved revision. If the registry can't create a download URL for a package, the whole request fails.

See also: AWS API Documentation

Request Syntax

client.batch_get_discoverable_registry_record(
    entries=[
        {
            'registryId': 'string',
            'recordIds': [
                'string',
            ]
        },
    ]
)
type entries:

list

param entries:

[REQUIRED]

The registry-scoped groups of record IDs to retrieve. Currently, you can specify exactly one entry.

  • (dict) --

    Binds one registry to the record IDs requested from it.

    • registryId (string) -- [REQUIRED]

      The identifier of the registry to retrieve the records from. You can provide either the full Amazon Resource Name (ARN) or the registry ID.

    • recordIds (list) -- [REQUIRED]

      The record IDs to retrieve from the registry. You can specify 1 through 100 record IDs.

      • (string) --

        Record identifier that accepts either ARN or ID format

rtype:

dict

returns:

Response Syntax

{
    'registryRecords': [
        {
            'registryArn': 'string',
            'recordArn': 'string',
            'recordId': 'string',
            'name': 'string',
            'description': 'string',
            'displayName': 'string',
            'recordType': 'MCP'|'AGENT'|'CUSTOM'|'SKILL'|'GATEWAY',
            'descriptors': {
                'mcpServer': {
                    'data': 'string',
                    'dataSchemaVersion': 'string',
                    'additionalData': {
                        'tools': {
                            'data': 'string',
                            'dataSchemaVersion': 'string'
                        }
                    },
                    'source': {
                        'fromUrl': {
                            'url': 'string'
                        }
                    }
                },
                'a2aAgentCard': {
                    'data': 'string',
                    'dataSchemaVersion': 'string',
                    'source': {
                        'fromUrl': {
                            'url': 'string'
                        }
                    }
                },
                'agentSkillsDefinition': {
                    'data': 'string',
                    'dataSchemaVersion': 'string',
                    'additionalData': {
                        'skillMd': {
                            'data': 'string',
                            'dataSchemaVersion': 'string',
                            'source': {
                                'fromUrl': {
                                    'url': 'string'
                                }
                            }
                        }
                    }
                },
                'custom': {
                    'data': 'string'
                },
                'http': {
                    'source': {
                        'fromUrl': {
                            'url': 'string'
                        }
                    }
                },
                'agui': {
                    'source': {
                        'fromUrl': {
                            'url': 'string'
                        }
                    }
                }
            },
            'recordVersion': 'string',
            'status': 'DRAFT'|'PENDING_APPROVAL'|'APPROVED'|'REJECTED'|'DEPRECATED'|'CREATING'|'UPDATING'|'CREATE_FAILED'|'UPDATE_FAILED'|'CREATE_PENDING_AUTH'|'UPDATE_PENDING_AUTH',
            'createdAt': datetime(2015, 1, 1),
            'updatedAt': datetime(2015, 1, 1),
            'customMetadata': {...}|[...]|123|123.4|'string'|True|None,
            'artifact': {
                'sha256': 'string',
                'downloadUrl': 'string'
            }
        },
    ],
    'errors': [
        {
            'registryId': 'string',
            'recordId': 'string',
            'errorCode': 'RESOURCE_NOT_FOUND'|'ACCESS_DENIED'|'INTERNAL_ERROR',
            'message': 'string'
        },
    ]
}

Response Structure

  • (dict) --

    • registryRecords (list) --

      The records that were successfully retrieved. Each record correlates to the request by its recordId.

      • (dict) --

        Summary information about a registry record, including its descriptors.

        • registryArn (string) --

          The Amazon Resource Name (ARN) of the parent registry that owns the record.

        • recordArn (string) --

          The Amazon Resource Name (ARN) of the registry record.

        • recordId (string) --

          The unique identifier of the registry record.

        • name (string) --

          The name of the registry record. Names are unique within a registry.

        • description (string) --

          A human-readable description of the registry record. Use this field to explain the record's purpose or content to consumers discovering it in the registry.

        • displayName (string) --

          The human-readable display name of the registry record.

        • recordType (string) --

          The type of the registry record. MCP is a Model Context Protocol server record, AGENT is an Agent-to-Agent (A2A) agent card record, SKILL is an agent skills definition record, and CUSTOM is a record with a custom descriptor.

        • descriptors (dict) --

          The protocol-specific descriptors that describe how to connect to and use the record.

          • mcpServer (dict) --

            The MCP server descriptor, populated when the record type is MCP.

            • data (string) --

              The MCP server descriptor content, serialized as descriptor payload data.

            • dataSchemaVersion (string) --

              The schema version of the descriptor payload.

            • additionalData (dict) --

              Additional data associated with the MCP server descriptor, such as tool definitions.

              • tools (dict) --

                The MCP tools descriptor that defines the tools exposed by the MCP server.

                • data (string) --

                  The MCP tools descriptor content, serialized as descriptor payload data.

                • dataSchemaVersion (string) --

                  The schema version of the descriptor payload.

            • source (dict) --

              The source location from which the MCP (Model Context Protocol) server descriptor content was retrieved.

              • fromUrl (dict) --

                The URL-based descriptor source, populated when descriptor content is synchronized from a URL.

                • url (string) --

                  The URL from which the descriptor content is retrieved. For an agent skill that synchronizes its SKILL.md from GitHub, this is the GitHub folder URL. The data plane doesn't return Amazon S3 ( s3://) source URLs.

          • a2aAgentCard (dict) --

            The A2A agent card descriptor, populated when the record type is AGENT.

            • data (string) --

              The A2A agent card content, serialized as descriptor payload data.

            • dataSchemaVersion (string) --

              The schema version of the descriptor payload.

            • source (dict) --

              The source location from which the A2A (Agent-to-Agent) agent card descriptor content was retrieved.

              • fromUrl (dict) --

                The URL-based descriptor source, populated when descriptor content is synchronized from a URL.

                • url (string) --

                  The URL from which the descriptor content is retrieved. For an agent skill that synchronizes its SKILL.md from GitHub, this is the GitHub folder URL. The data plane doesn't return Amazon S3 ( s3://) source URLs.

          • agentSkillsDefinition (dict) --

            The agent skills definition descriptor, populated when the record type is SKILL.

            • data (string) --

              The agent skills definition content, serialized as descriptor payload data.

            • dataSchemaVersion (string) --

              The schema version of the descriptor payload.

            • additionalData (dict) --

              Additional data for the agent skills definition, such as the skills markdown descriptor.

              • skillMd (dict) --

                The agent skills markdown descriptor associated with the agent skills definition.

                • data (string) --

                  The agent skills markdown content, serialized as descriptor payload data.

                • dataSchemaVersion (string) --

                  The schema version of the descriptor payload.

                • source (dict) --

                  The source location from which the agent skills markdown content was retrieved.

                  • fromUrl (dict) --

                    The URL-based descriptor source, populated when descriptor content is synchronized from a URL.

                    • url (string) --

                      The URL from which the descriptor content is retrieved. For an agent skill that synchronizes its SKILL.md from GitHub, this is the GitHub folder URL. The data plane doesn't return Amazon S3 ( s3://) source URLs.

          • custom (dict) --

            The custom descriptor, populated when the record type is CUSTOM.

            • data (string) --

              The custom descriptor content, serialized as descriptor payload data.

          • http (dict) --

            The HTTP descriptor, populated when the record exposes an HTTP endpoint.

            • source (dict) --

              The source location of the HTTP endpoint.

              • fromUrl (dict) --

                The URL-based descriptor source, populated when descriptor content is synchronized from a URL.

                • url (string) --

                  The URL from which the descriptor content is retrieved. For an agent skill that synchronizes its SKILL.md from GitHub, this is the GitHub folder URL. The data plane doesn't return Amazon S3 ( s3://) source URLs.

          • agui (dict) --

            The AG-UI descriptor, populated when the record exposes an AG-UI protocol endpoint.

            • source (dict) --

              The source location of the AG-UI protocol endpoint.

              • fromUrl (dict) --

                The URL-based descriptor source, populated when descriptor content is synchronized from a URL.

                • url (string) --

                  The URL from which the descriptor content is retrieved. For an agent skill that synchronizes its SKILL.md from GitHub, this is the GitHub folder URL. The data plane doesn't return Amazon S3 ( s3://) source URLs.

        • recordVersion (string) --

          The version identifier of the registry record.

        • status (string) --

          The lifecycle status of the registry record. A record is DRAFT before it is submitted, PENDING_APPROVAL while awaiting curator review, and APPROVED once it is approved and discoverable. REJECTED and DEPRECATED records are not discoverable. The CREATING, UPDATING, CREATE_FAILED, and UPDATE_FAILED values reflect the state of an in-progress or failed asynchronous change.

        • createdAt (datetime) --

          The timestamp when the registry record was created.

        • updatedAt (datetime) --

          The timestamp when the registry record was last updated.

        • customMetadata (:ref:`document<document>`) --

          The custom metadata attached to this registry record. Values are strings (maximum 128 characters) or booleans. This field is only present if the registry has a custom metadata schema configured.

        • artifact (dict) --

          The skill package of the record's latest approved revision, as the package's SHA-256 checksum and a download URL that is valid for up to 5 minutes. Only BatchGetDiscoverableRegistryRecord returns this member, not SearchDiscoverableRegistryRecords. Absent when the record has no skill package.

          • sha256 (string) --

            The SHA-256 checksum of the package, as 64 lowercase hexadecimal characters. The download response carries no signature that you can check, so compare the checksum of the file that you download with this value before you use the package.

          • downloadUrl (string) --

            A presigned Amazon S3 URL that downloads the package. The URL is valid for up to 5 minutes, and the response doesn't include its expiry time. To get a new URL, call the operation again. Anyone who has the URL can download the package until it expires, so don't log, store, or share it.

    • errors (list) --

      The per-record errors for records that could not be retrieved. This list is empty when all requested records were returned.

      • (dict) --

        Describes why a requested record could not be retrieved.

        • registryId (string) --

          The identifier of the registry the record was requested from, echoed from the request.

        • recordId (string) --

          The identifier of the record that could not be retrieved, echoed from the request in the same format that you supplied (ARN or record ID).

        • errorCode (string) --

          The machine-readable reason that the record could not be retrieved.

        • message (string) --

          An optional human-readable detail about the error. Do not parse this value programmatically.

ListDiscoverableRegistryRecords (updated) Link ¶
Changes (response)
{'registryRecords': {'status': {'UPDATE_PENDING_AUTH', 'CREATE_PENDING_AUTH'}}}

Lists the discoverable registry records in a registry. You can optionally filter and paginate the results.

See also: AWS API Documentation

Request Syntax

client.list_discoverable_registry_records(
    registryId='string',
    maxResults=123,
    nextToken='string',
    filters=[
        {
            'name': 'recordType'|'descriptorType',
            'values': [
                'string',
            ]
        },
    ]
)
type registryId:

string

param registryId:

[REQUIRED]

The identifier of the registry whose discoverable records are listed. You can provide either the full Amazon Resource Name (ARN) or the registry ID.

type maxResults:

integer

param maxResults:

The maximum number of records to return in a single page. Valid values are 1 through 100.

type nextToken:

string

param nextToken:

The pagination token returned by a previous request. Use this value to retrieve the next page of results.

type filters:

list

param filters:

The filters to apply to the discoverable registry record list.

  • (dict) --

    A single filter applied to a ListDiscoverableRegistryRecords request.

    • name (string) -- [REQUIRED]

      The attribute to filter on.

    • values (list) -- [REQUIRED]

      The values to match for the attribute.

      • (string) --

        A single filter value. Constrained by length only; the accepted set of values is filter-specific and validated server-side so enum-style values such as READY or AWS_IAM are accepted.

rtype:

dict

returns:

Response Syntax

{
    'registryRecords': [
        {
            'registryArn': 'string',
            'recordArn': 'string',
            'recordId': 'string',
            'name': 'string',
            'description': 'string',
            'displayName': 'string',
            'recordType': 'MCP'|'AGENT'|'CUSTOM'|'SKILL'|'GATEWAY',
            'recordVersion': 'string',
            'status': 'DRAFT'|'PENDING_APPROVAL'|'APPROVED'|'REJECTED'|'DEPRECATED'|'CREATING'|'UPDATING'|'CREATE_FAILED'|'UPDATE_FAILED'|'CREATE_PENDING_AUTH'|'UPDATE_PENDING_AUTH',
            'createdAt': datetime(2015, 1, 1),
            'updatedAt': datetime(2015, 1, 1),
            'descriptorTypes': [
                'string',
            ]
        },
    ],
    'nextToken': 'string'
}

Response Structure

  • (dict) --

    • registryRecords (list) --

      The page of discoverable registry record summaries.

      • (dict) --

        Summary information about a discoverable registry record returned by ListDiscoverableRegistryRecords. This summary does not include descriptors.

        • registryArn (string) --

          The Amazon Resource Name (ARN) of the parent registry that owns the record.

        • recordArn (string) --

          The Amazon Resource Name (ARN) of the registry record.

        • recordId (string) --

          The unique identifier of the registry record.

        • name (string) --

          The name of the registry record. Names are unique within a registry.

        • description (string) --

          A human-readable description of the registry record. Use this field to explain the record's purpose or content to consumers discovering it in the registry.

        • displayName (string) --

          The human-readable display name of the registry record.

        • recordType (string) --

          The type of the registry record. MCP is a Model Context Protocol server record, AGENT is an Agent-to-Agent (A2A) agent card record, SKILL is an agent skills definition record, and CUSTOM is a record with a custom descriptor.

        • recordVersion (string) --

          The version identifier of the registry record.

        • status (string) --

          The lifecycle status of the registry record. A record is DRAFT before it is submitted, PENDING_APPROVAL while awaiting curator review, and APPROVED once it is approved and discoverable. REJECTED and DEPRECATED records are not discoverable. The CREATING, UPDATING, CREATE_FAILED, and UPDATE_FAILED values reflect the state of an in-progress or failed asynchronous change.

        • createdAt (datetime) --

          The timestamp when the registry record was created.

        • updatedAt (datetime) --

          The timestamp when the registry record was last updated.

        • descriptorTypes (list) --

          The descriptor types that are present on this registry record. Each value corresponds to a descriptor entry key on the approved record.

          • (string) --

    • nextToken (string) --

      The pagination token to pass to a subsequent request to retrieve the next page of results. This field is absent when there are no more results.

SearchDiscoverableRegistryRecords (updated) Link ¶
Changes (request, response)
Request
{'nextToken': 'string', 'searchType': 'HYBRID | SEMANTIC | KEYWORD'}
Response
{'nextToken': 'string',
 'registryRecords': {'artifact': {'downloadUrl': 'string', 'sha256': 'string'},
                     'status': {'UPDATE_PENDING_AUTH', 'CREATE_PENDING_AUTH'}},
 'searchType': 'HYBRID | SEMANTIC | KEYWORD'}

Searches the discoverable registry records in a registry using a natural language query. Returns metadata for the matching records ordered by relevance.

Returns at most 1000 results across all pages when searchType is KEYWORD, and at most 20 results when it is HYBRID or SEMANTIC. To retrieve more records, refine the query or filters, or use ListDiscoverableRegistryRecords to list every record.

See also: AWS API Documentation

Request Syntax

client.search_discoverable_registry_records(
    searchQuery='string',
    registryIds=[
        'string',
    ],
    maxResults=123,
    filters={...}|[...]|123|123.4|'string'|True|None,
    searchType='HYBRID'|'SEMANTIC'|'KEYWORD',
    nextToken='string'
)
type searchQuery:

string

param searchQuery:

[REQUIRED]

The natural language query to search for matching registry records. If you specify an empty string, the operation returns the records that match the filters, most recently updated first, up to the result limit for the searchType.

type registryIds:

list

param registryIds:

[REQUIRED]

The registry identifiers to search within. Currently, you must specify exactly one registry identifier. You can provide either the full Amazon Web Services Resource Name (ARN) or the registry ID.

  • (string) --

    Registry identifier that accepts either ARN or ID format

type maxResults:

integer

param maxResults:

The maximum number of results to return in one page. Valid values are 1 through 50 when searchType is KEYWORD, and 1 through 20 when it is HYBRID or SEMANTIC. The default value is 10.

type filters:

:ref:`document<document>`

param filters:

An optional structured JSON metadata filter that narrows the search results. Supports the field-level operators $eq, $ne, and $in, and the logical operators $and and $or on filterable fields.

Specifies additional filtering on custom metadata fields using the customMetadata.{key} prefix. For example, to filter by a custom metadata field: {"customMetadata.environment": {"$eq": "production"}}. For a Boolean field, you can also use a native JSON boolean value, for example: {"customMetadata.requiresApproval": {"$eq": true}}.

type searchType:

string

param searchType:

Specifies how the service ranks registry records against the search query. If you omit this parameter, the service uses HYBRID in Amazon Web Services Regions that support semantic search and KEYWORD in Regions that do not. The response reports the search type that the service used. Valid values are the following:

  • KEYWORD – Ranks records by keyword match between the query terms and the record name, description, descriptors, and custom metadata text values. Available in every Amazon Web Services Region.

  • SEMANTIC – Ranks records by semantic similarity to the query and returns the most similar records, up to maxResults, even when none of them closely matches the query.

  • HYBRID – Combines keyword ranking and semantic ranking.

HYBRID and SEMANTIC are available in a limited number of Regions. If you specify either one in a Region that does not support semantic search, the request fails with a ValidationException that has the reason UnsupportedInRegion.

type nextToken:

string

param nextToken:

The pagination token from a previous response. Use it to retrieve the next page of results, and send the same searchQuery, registryIds, filters, and searchType that you used for the first page. If any of them differ, the request fails with a ValidationException. A pagination token expires 15 minutes after the service returns the first page.

rtype:

dict

returns:

Response Syntax

{
    'registryRecords': [
        {
            'registryArn': 'string',
            'recordArn': 'string',
            'recordId': 'string',
            'name': 'string',
            'description': 'string',
            'displayName': 'string',
            'recordType': 'MCP'|'AGENT'|'CUSTOM'|'SKILL'|'GATEWAY',
            'descriptors': {
                'mcpServer': {
                    'data': 'string',
                    'dataSchemaVersion': 'string',
                    'additionalData': {
                        'tools': {
                            'data': 'string',
                            'dataSchemaVersion': 'string'
                        }
                    },
                    'source': {
                        'fromUrl': {
                            'url': 'string'
                        }
                    }
                },
                'a2aAgentCard': {
                    'data': 'string',
                    'dataSchemaVersion': 'string',
                    'source': {
                        'fromUrl': {
                            'url': 'string'
                        }
                    }
                },
                'agentSkillsDefinition': {
                    'data': 'string',
                    'dataSchemaVersion': 'string',
                    'additionalData': {
                        'skillMd': {
                            'data': 'string',
                            'dataSchemaVersion': 'string',
                            'source': {
                                'fromUrl': {
                                    'url': 'string'
                                }
                            }
                        }
                    }
                },
                'custom': {
                    'data': 'string'
                },
                'http': {
                    'source': {
                        'fromUrl': {
                            'url': 'string'
                        }
                    }
                },
                'agui': {
                    'source': {
                        'fromUrl': {
                            'url': 'string'
                        }
                    }
                }
            },
            'recordVersion': 'string',
            'status': 'DRAFT'|'PENDING_APPROVAL'|'APPROVED'|'REJECTED'|'DEPRECATED'|'CREATING'|'UPDATING'|'CREATE_FAILED'|'UPDATE_FAILED'|'CREATE_PENDING_AUTH'|'UPDATE_PENDING_AUTH',
            'createdAt': datetime(2015, 1, 1),
            'updatedAt': datetime(2015, 1, 1),
            'customMetadata': {...}|[...]|123|123.4|'string'|True|None,
            'artifact': {
                'sha256': 'string',
                'downloadUrl': 'string'
            }
        },
    ],
    'searchType': 'HYBRID'|'SEMANTIC'|'KEYWORD',
    'nextToken': 'string'
}

Response Structure

  • (dict) --

    • registryRecords (list) --

      The registry records that match the search query, ordered by relevance. A page can contain fewer than maxResults records, or none, even when the response includes a nextToken. To retrieve every result, continue until the response has no nextToken.

      • (dict) --

        Summary information about a registry record, including its descriptors.

        • registryArn (string) --

          The Amazon Resource Name (ARN) of the parent registry that owns the record.

        • recordArn (string) --

          The Amazon Resource Name (ARN) of the registry record.

        • recordId (string) --

          The unique identifier of the registry record.

        • name (string) --

          The name of the registry record. Names are unique within a registry.

        • description (string) --

          A human-readable description of the registry record. Use this field to explain the record's purpose or content to consumers discovering it in the registry.

        • displayName (string) --

          The human-readable display name of the registry record.

        • recordType (string) --

          The type of the registry record. MCP is a Model Context Protocol server record, AGENT is an Agent-to-Agent (A2A) agent card record, SKILL is an agent skills definition record, and CUSTOM is a record with a custom descriptor.

        • descriptors (dict) --

          The protocol-specific descriptors that describe how to connect to and use the record.

          • mcpServer (dict) --

            The MCP server descriptor, populated when the record type is MCP.

            • data (string) --

              The MCP server descriptor content, serialized as descriptor payload data.

            • dataSchemaVersion (string) --

              The schema version of the descriptor payload.

            • additionalData (dict) --

              Additional data associated with the MCP server descriptor, such as tool definitions.

              • tools (dict) --

                The MCP tools descriptor that defines the tools exposed by the MCP server.

                • data (string) --

                  The MCP tools descriptor content, serialized as descriptor payload data.

                • dataSchemaVersion (string) --

                  The schema version of the descriptor payload.

            • source (dict) --

              The source location from which the MCP (Model Context Protocol) server descriptor content was retrieved.

              • fromUrl (dict) --

                The URL-based descriptor source, populated when descriptor content is synchronized from a URL.

                • url (string) --

                  The URL from which the descriptor content is retrieved. For an agent skill that synchronizes its SKILL.md from GitHub, this is the GitHub folder URL. The data plane doesn't return Amazon S3 ( s3://) source URLs.

          • a2aAgentCard (dict) --

            The A2A agent card descriptor, populated when the record type is AGENT.

            • data (string) --

              The A2A agent card content, serialized as descriptor payload data.

            • dataSchemaVersion (string) --

              The schema version of the descriptor payload.

            • source (dict) --

              The source location from which the A2A (Agent-to-Agent) agent card descriptor content was retrieved.

              • fromUrl (dict) --

                The URL-based descriptor source, populated when descriptor content is synchronized from a URL.

                • url (string) --

                  The URL from which the descriptor content is retrieved. For an agent skill that synchronizes its SKILL.md from GitHub, this is the GitHub folder URL. The data plane doesn't return Amazon S3 ( s3://) source URLs.

          • agentSkillsDefinition (dict) --

            The agent skills definition descriptor, populated when the record type is SKILL.

            • data (string) --

              The agent skills definition content, serialized as descriptor payload data.

            • dataSchemaVersion (string) --

              The schema version of the descriptor payload.

            • additionalData (dict) --

              Additional data for the agent skills definition, such as the skills markdown descriptor.

              • skillMd (dict) --

                The agent skills markdown descriptor associated with the agent skills definition.

                • data (string) --

                  The agent skills markdown content, serialized as descriptor payload data.

                • dataSchemaVersion (string) --

                  The schema version of the descriptor payload.

                • source (dict) --

                  The source location from which the agent skills markdown content was retrieved.

                  • fromUrl (dict) --

                    The URL-based descriptor source, populated when descriptor content is synchronized from a URL.

                    • url (string) --

                      The URL from which the descriptor content is retrieved. For an agent skill that synchronizes its SKILL.md from GitHub, this is the GitHub folder URL. The data plane doesn't return Amazon S3 ( s3://) source URLs.

          • custom (dict) --

            The custom descriptor, populated when the record type is CUSTOM.

            • data (string) --

              The custom descriptor content, serialized as descriptor payload data.

          • http (dict) --

            The HTTP descriptor, populated when the record exposes an HTTP endpoint.

            • source (dict) --

              The source location of the HTTP endpoint.

              • fromUrl (dict) --

                The URL-based descriptor source, populated when descriptor content is synchronized from a URL.

                • url (string) --

                  The URL from which the descriptor content is retrieved. For an agent skill that synchronizes its SKILL.md from GitHub, this is the GitHub folder URL. The data plane doesn't return Amazon S3 ( s3://) source URLs.

          • agui (dict) --

            The AG-UI descriptor, populated when the record exposes an AG-UI protocol endpoint.

            • source (dict) --

              The source location of the AG-UI protocol endpoint.

              • fromUrl (dict) --

                The URL-based descriptor source, populated when descriptor content is synchronized from a URL.

                • url (string) --

                  The URL from which the descriptor content is retrieved. For an agent skill that synchronizes its SKILL.md from GitHub, this is the GitHub folder URL. The data plane doesn't return Amazon S3 ( s3://) source URLs.

        • recordVersion (string) --

          The version identifier of the registry record.

        • status (string) --

          The lifecycle status of the registry record. A record is DRAFT before it is submitted, PENDING_APPROVAL while awaiting curator review, and APPROVED once it is approved and discoverable. REJECTED and DEPRECATED records are not discoverable. The CREATING, UPDATING, CREATE_FAILED, and UPDATE_FAILED values reflect the state of an in-progress or failed asynchronous change.

        • createdAt (datetime) --

          The timestamp when the registry record was created.

        • updatedAt (datetime) --

          The timestamp when the registry record was last updated.

        • customMetadata (:ref:`document<document>`) --

          The custom metadata attached to this registry record. Values are strings (maximum 128 characters) or booleans. This field is only present if the registry has a custom metadata schema configured.

        • artifact (dict) --

          The skill package of the record's latest approved revision, as the package's SHA-256 checksum and a download URL that is valid for up to 5 minutes. Only BatchGetDiscoverableRegistryRecord returns this member, not SearchDiscoverableRegistryRecords. Absent when the record has no skill package.

          • sha256 (string) --

            The SHA-256 checksum of the package, as 64 lowercase hexadecimal characters. The download response carries no signature that you can check, so compare the checksum of the file that you download with this value before you use the package.

          • downloadUrl (string) --

            A presigned Amazon S3 URL that downloads the package. The URL is valid for up to 5 minutes, and the response doesn't include its expiry time. To get a new URL, call the operation again. Anyone who has the URL can download the package until it expires, so don't log, store, or share it.

    • searchType (string) --

      Specifies the search type that the service used to rank the returned records. If you omitted searchType in the request, this is the default for the Amazon Web Services Region: HYBRID where semantic search is supported, and KEYWORD elsewhere.

    • nextToken (string) --

      The pagination token to pass to a subsequent request to retrieve the next page of results. This field is absent when there are no more results, or when the search has returned the maximum number of results for its search type.