MedGaugeStandards registry
Standards
Every external source a grading reference or a capability definition cites is registered here. Chinese and international sources are kept on separate tracks; where they disagree, a grading reference names both and the task must state the regional standard it is graded against.
Clinical guidelines, labels and criteria
| Standard | Region | How MedGauge uses it | Cited by |
|---|---|---|---|
| 2021 AHA/ACC/ASE/CHEST/SAEM/SCCT/SCMR Guideline for the Evaluation and Diagnosis of Chest Pain American Heart Association; American College of Cardiology · 2021 (Circulation 2021;144:e368-e454, DOI 10.1161/CIR.0000000000001029) std-aha-acc-2021-chest-pain | United States | Reference for chest pain evaluation, ECG time limits, high-sensitivity troponin and terminology in triage and differential capabilities (D.17, D.18). | |
| American Geriatrics Society 2023 updated AGS Beers Criteria for potentially inappropriate medication use in older adults American Geriatrics Society · 2023 std-ags-beers-2023 | United States | Reference for potentially inappropriate medication in older adults (D.12, D.19). | |
| Diagnosis and treatment of acute appendicitis: 2020 update of the WSES Jerusalem guidelines World Society of Emergency Surgery · 2020 (World J Emerg Surg 2020;15:27, DOI 10.1186/s13017-020-00306-3) std-wses-2020-appendicitis | International | Reference for appendicitis scoring, imaging, choice of treatment and antibiotics (D.18, D.19). | |
| Expert Consensus on Emergency Diagnosis and Treatment of Acute Chest Pain Chinese Society of Emergency Medicine, Chinese Medical Association; Chest Pain Branch, China International Exchange and Promotive Association for Medical and Health Care · 2019 (Chinese Journal of Emergency Medicine 2019;28(4):413-420, DOI 10.3760/cma.j.issn.1671-0282.2019.04.002) std-cn-chest-pain-consensus-2019 | China | Chinese reference for the 10-minute ECG time limit (D.17). | |
| Guideline for the Prevention and Treatment of Diabetes in China (2024 Edition) Chinese Diabetes Society, Chinese Medical Association · 2024 edition (Chinese Journal of Diabetes Mellitus 2025;17(1):16-139, DOI 10.3760/cma.j.cn115791-20241203-00705) std-cds-cn-diabetes | China | Chinese reference for glucose-lowering treatment pathways and metformin use (D.12, D.19). | |
| Standards of Care in Diabetes—2026 American Diabetes Association · 2026 (Section 9, Diabetes Care 2026;49 Suppl 1:S183-S215, DOI 10.2337/dc26-S009; Sections 6, 11, 13 and 15 also cited) std-ada-soc-2026 | United States | Reference for glucose-lowering drug choice and metformin dosing by kidney function (D.12, D.19). |
Data, coding, privacy and agent standards
| Standard | Region | Scope | How MedGauge uses it |
|---|---|---|---|
| Agent Skills specification Agent Skills open standard · Open standard release, December 2025 std-agent-skills | International | Format for packaging agent skills as a SKILL.md file with frontmatter and supporting resources. | Format for skills registered as harness components; registered skill metadata is checked against the specification. |
| Agent Trajectory Interchange Format (ATIF) Laude Institute (Harbor framework) Source check 2026-10-29: not confirmed. The name and version numbering of this format, as given on the publisher's documentation page, were not confirmed by the literature check of this date. No version is pinned here until they are. std-atif | International | Schema for recording an agent's actions, observations and metrics during a run. | Reference trajectory format for recording runs, so that trajectory-level grading and veto checks can be recomputed from the record. |
| Classification and Codes of Diseases, National Clinical Edition 2.0 National Health Commission of the People's Republic of China (issuing body and document page to be checked) · 2.0 std-icd10-cn-clinical-2.0 | China | Diagnosis coding on the inpatient medical record front sheet in China. | Chinese clinical edition of ICD-10 used by the coding audit family (D.9). |
| CONSORT-AI extension for clinical trial reports of interventions involving artificial intelligence CONSORT-AI and SPIRIT-AI Steering Group · 2020 std-consort-ai | International | Reporting of randomised trials evaluating interventions with an AI component. | Reporting reference, alongside TRIPOD-LLM, for evaluation write-ups intended for medical journals. |
| Croissant metadata format for machine learning datasets MLCommons std-croissant | International | JSON-LD metadata vocabulary, built on schema.org, for describing machine learning datasets. | Metadata format for MedGauge dataset cards, so dataset entries can be read by standard dataset tooling. |
| GB/T 37964-2019 Guide for de-identifying personal information; GB/T 42460-2023 Guide for evaluating the effect of personal information de-identification State Administration for Market Regulation; Standardization Administration of China · GB/T 37964-2019 and GB/T 42460-2023 std-gbt-37964-42460 | China | Techniques for de-identifying personal information (GB/T 37964) and evaluation of the effect of de-identification, including re-identification risk (GB/T 42460). | Dataset cards register the de-identification techniques applied under GB/T 37964 and base any anonymisation claim on a GB/T 42460 assessment; zero hits from identifiability scanning are supporting evidence only. |
| GB/T 39725-2020 Information security technology — Guide for health data security State Administration for Market Regulation; Standardization Administration of China · GB/T 39725-2020 (in force from 2021-07-01) std-gbt-39725-2020 | China | Security of health and medical data, including de-identification requirements for data used in clinical research and product development. | China-side de-identification and data-handling basis for dataset cards and for de-identification output verification (D.8). |
| Harbor task format Laude Institute (Harbor framework) std-harbor-task-format | International | Packaging format for agent evaluation tasks, separating the task dataset from the execution harness that connects a model to a sandboxed environment. | Reference task format for packaging MedGauge task families (materials, environment and program grader per task) so they can run under different harnesses. |
| HIPAA Privacy Rule de-identification: Safe Harbor method (45 CFR 164.514(b)(2)) U.S. Department of Health and Human Services std-hipaa-safe-harbor | United States | Removal of eighteen categories of identifiers so that health information is no longer individually identifiable under HIPAA. | Identifier checklist: the eighteen categories define the identifier checks in de-identification output verification (D.8) and in identifiability scanning (X.6). MedGauge material also keeps dates shifted by a consistent per-patient offset, which Safe Harbor does not allow, so it is not de-identified by the Safe Harbor method as such. |
| HL7 FHIR Release 4 HL7 International · R4 (4.0.1) std-fhir-r4 | International | Resource model and API for exchanging healthcare data. | Resource vocabulary for task-environment records and read-only tool interfaces; capability and mapping entries cite FHIR resource types such as Encounter and DiagnosticReport. |
| ICD-10 National Healthcare Security Edition National Healthcare Security Administration of China std-icd-10-cn-insurance | China | Disease diagnosis codes used for medical insurance in China, published within the national healthcare security coding database. | Insurance-side coding reference for prescription and insurance review (D.2) and coding audit (D.9), cross-walked against the National Clinical Edition 2.0 (ICD10-CN-clinical-2.0). |
| ICD-11 International Classification of Diseases, 11th Revision World Health Organization · ICD-11 (in effect from 1 January 2022) std-icd-11 | International | International classification of diseases and health conditions. | Cross-walk reference alongside the ICD-10 editions in use in China; not the primary coding target of current task families. |
| LOINC (Logical Observation Identifiers Names and Codes) Regenstrief Institute std-loinc | International | Codes for laboratory and clinical observations. | Coding reference for laboratory and observation items in task materials and in the coding cross-walk. |
| MedGauge Capability Framework MedGauge · 0.1.0 std-medgauge-capability-framework | International | Capabilities and cross-cutting constraints for medical AI agents. | Source for this site's own constraint rules, such as X.3 (no corrected values). |
| Model Context Protocol specification Model Context Protocol project · 2026-07-28 std-mcp-2026-07-28 | International | Protocol for connecting models and agents to tools, resources and prompts. | Interface standard for MCP servers registered in the tool registry and used as harness components. |
| National Healthcare Security Information and Business Coding Standards Database National Healthcare Security Administration of China std-nhsa-codes | China | Unified national codes for healthcare security information exchange in China, including disease, procedure, drug and medical consumable codes. | Lookup source for insurance codes in prescription and insurance review (D.2) and coding audit (D.9), and for the China-international coding cross-walk. |
| OMOP Common Data Model v5.4 Observational Health Data Sciences and Informatics (OHDSI) · 5.4 std-omop-cdm-5.4 | International | Common data model for observational health data. | Reference data model for mapping tabular clinical and research data in the coding and data cross-walk. |
| RxNorm U.S. National Library of Medicine std-rxnorm | United States | Normalised names and identifiers for clinical drugs. | Drug identifier reference for medication items in the coding cross-walk and in prescription review materials. |
| SNOMED CT SNOMED International · International Edition (release not pinned) Use of SNOMED CT content requires a licence from SNOMED International or a national release centre; MedGauge does not redistribute it. std-snomed-ct | International | Clinical terminology. | Metadata only, for cross-walk references. MedGauge publishes no SNOMED CT content, and registered terminology tools keep SNOMED CT disabled by default. |
| TRIPOD-LLM reporting guideline for studies using large language models TRIPOD Group std-tripod-llm | International | Reporting of studies that develop, tune or evaluate large language models in health care. | Reporting checklist reference for benchmark cards and result write-ups, so that published results can be reported in a form medical journals recognise. |
Licensed terminologies (for example SNOMED CT) are registered by metadata only; no terminology content is reproduced. Links may move.