Metadata Standard – LlmWikis.org

Metadata is how an LLM Wiki tells humans, crawlers, and AI agents what a page is, where its canonical URL lives, who owns it, whether it is current, what sources support it, and what an agent may do with it.

Recommended Frontmatter

---
title: System Overview
slug: system-overview
canonical_url: https://example.com/reference/system-overview/
description: Current system boundary, owners, components, source traces, and review state.
page_type: reference
section: architecture
audience: both
owner: Platform Team
status: current
quality_class: reviewed
source_status: source-linked
last_reviewed: 2026-05-11
review_cycle: quarterly
sensitivity: internal
agent_use: allowed-with-citation
parent: /reference/
children: []
aliases:
  - system map
redirects_from:
  - /old-system-overview/
primary_topic: system architecture
tags:
  - architecture
trust_level: authoritative
source_trace:
  - raw/architecture/2026-05-platform-review.md
claim_status: reviewed
freshness_rule: review every quarter and after major architecture changes
reviewer: Architecture owner
updated: 2026-05-11
created: 2026-05-01
noindex: false
sitemap_include: true
jsonld_type: TechArticle
social_title: System Overview
social_description: Current reviewed system boundary and source-backed architecture context.
related:
  - title: Services
    url: /reference/services/
    relation: has_part
typed_relations:
  depends_on: []
  prerequisite_for: []
  part_of:
    - /reference/structure/
  has_part:
    - /reference/services/
  example_of: []
  implemented_by: []
  supersedes: []
  superseded_by: []
  contradicts: []
  source_for: []
  related_to:
    - /reference/navigation/
  see_also:
    - /reference/metadata/
---

Required Fields

Field	Why it matters
`title`	Stable page identity for humans, search, and citations.
`slug`	Lowercase kebab-case route component.
`canonical_url`	One durable public URL per important concept, entity, workflow, standard, source type, or task.
`description`	Answer-first summary for search snippets, agents, cards, and llms.txt.
`page_type`	Allowed values: concept, guide, reference, tool, template, example, comparison, glossary, changelog, evidence, or report.
`section`	Navigation parent or handbook section.
`audience`	Allowed values: human, agent, or both.
`status`	Allowed values: draft, reviewed, current, stale, deprecated, archived, or blocked.
`quality_class`	Editorial quality level such as stub, start, reviewed, canonical, or needs-repair.
`source_status`	Whether the page is source-linked, source-needed, canonical-linked, derived, or unsupported.
`owner`	Prevents ownerless authority.
`last_reviewed`	Makes freshness visible.
`review_cycle`	Defines when stale detection should trigger.
`sensitivity`	Prevents accidental exposure or unsafe agent use.
`agent_use`	States whether agents may read, cite, summarize, propose edits, or must avoid the page.
`parent` and `children`	Power breadcrumbs, hub pages, and local hierarchy.
`aliases` and `redirects_from`	Preserve alternate names without creating duplicate canonicals.
`related`	Keeps navigation explicit and retrieval-ready.

Strongly Recommended Fields

Use primary_topic, tags, trust_level, source_trace, claim_status, freshness_rule, reviewer, updated, created, noindex, sitemap_include, jsonld_type, social_title, social_description, social_image, and typed_relations when the page is public, source-backed, or agent-facing.

Search result fields

Field	Where it appears	Search behavior
`title`	Result heading, route card, citation label.	Should name the durable concept, not a vague report title.
`description`	Search excerpt, llms.txt summary, social metadata.	Should answer what the page is for in one sentence.
`canonical_url`	Result target and citation path.	Aliases and old URLs should redirect here instead of competing.
`page_type` and `section`	Result grouping and route shortcuts.	Lets broad searches recover into guide, reference, operation, example, or comparison lanes.
`status`, `source_status`, and `last_reviewed`	Freshness and trust signals.	Agents should treat stale, draft, source-needed, and contradicted results differently from current reviewed pages.
`aliases` and `redirects_from`	Query recall and migration compatibility.	Search can recognize old names without creating duplicate canonical answers.

Validation Rules

Missing title, invalid status, invalid page_type, duplicate canonical_url, or missing public canonical_url fails lint.
Missing public description, breadcrumb data, or local related link warns or fails depending on publication policy.
Duplicate titles warn; duplicate canonical URLs fail.
Stale pages without review status warn.
Contradicted pages without a contradicts relation fail.
Public pages with noindex: true need an archive, evidence, duplicate, or privacy reason.