WebMEM™

The Protocol for Structuring, Delivering, and Conditioning Trust-Scored AI Memory on the Open Web

ProvenanceMeta Best Practices

Part of the WebMEM Protocol
Location: /specification/sdt/yaml-in-html/provenance/
Last Updated: 2025-07-28

Overview

The ProvenanceMeta block anchors every SDT fragment to its source, license, scope, and trust lineage. While required fields ensure basic traceability, enhanced provenance declarations improve downstream retrievability, auditability, and trust scoring.

This section defines best practices for structuring a complete ProvenanceMeta block in YAML-in-HTML SDT fragments.

Scope

The Scope field clarifies the structural origin of the fragment’s content. Use controlled vocabulary values to maintain consistency and interoperability:

Scope Value Description
single-dataset Derived directly from one source dataset (e.g., CMS MA Landscape ZIP)
multi-dataset Harmonized from multiple distinct datasets
derived Computed or aggregated from source values (e.g., averages, rankings)
editorial Authored manually or guided by policy; used in curated content like FAQs or glossary entries

Guidance: When in doubt, favor derived or editorial to clarify transformation depth for downstream agents.

Format

Use clear, specific labels for the structure or archive type of the original source data.

Examples:

  • ZIP (XLSX)
  • CSV
  • JSON
  • Markdown
  • PDF

Avoid vague formats like spreadsheet or file. If a file is inside an archive, list the archive format (e.g., ZIP) and clarify the filename in the Description field.

Version

Use this to indicate structured versioning for datasets or glossary files.

Examples:

  • v2025.0 — Baseline CMS release
  • v2025.1 — Patch or post-correction file
  • v1.2.3 — Internal editorial or glossary revision

Only include if the dataset is expected to be referenced or evolved over time.

Tags

Use short, lowercase tags to categorize and cluster fragments.

Example:

Tags:
  - cms
  - landscape
  - 2025
  - pbp

Tags are flat, non-hierarchical, and non-canonical — intended for internal filtering and export behavior.

Guidelines

When a dataset or content source is governed by regulation or editorial policy, cite the URL or document name explicitly.

Examples:

  • CMS Medicare Marketing Guidelines
  • HHS Data Governance Framework
  • Internal Trust Publishing Editorial Policy

Note: Fragments without a declared Guidelines: field may be scored lower in regulated publishing domains.

Other Notes

  • Use ISO 8601 format for all dates (e.g., 2025-07-28)
  • Use SHA-256 or MD5 for Checksum: field when declaring raw file integrity
  • The Digest: field must match across all fragments that belong to the same exportable set

Learn More