Glossary
Define a shared vocabulary — one canonical term per entry, with a concise definition, synonyms, related terms, and a stewarding owner.
Create this collectionThe policies it ships with
People and their agents follow these when they record in a collection made from this template. Its owners can change them.
- A node belongs in a glossary when it defines exactly one term. Each Reference is one entry: its prose is the headword (the word being defined) and its `definition` gives the meaning. PASS when the candidate names and defines a single concept a reader of this domain would look up. FAIL only when it bundles several unrelated terms into one entry, is a passing note, task, or decision record rather than a definable term, or carries no definable concept at all.
- Only Reference and Principal belong in a glossary. Each term is a Reference — its prose is the headword and its `definition` carries the meaning; Principals are the stewards who own terms. A process step, decision record, or free-form note belongs in its own collection.
- Only these edge types may be used in a glossary: `relates_to` (see-also between related terms), `has_parent` (place a narrower term under its broader term or category), `replaces` (a preferred term supersedes a deprecated one), and `attributed_to` (name the Principal who stewards the term). A term cites its source inline in its `locator`, not via an edge.
- Every committed (`queued` or `active`) term carries a `definition` — a headword with no meaning is not yet an entry. A `drafting` stub may capture the word first and fill in the definition later.
- Check the term's `definition`. PASS when it is a concise, self-contained explanation of the concept (roughly one to three sentences) in plain language: it states what the concept *is* — its category and what distinguishes it — and could stand in for the headword in a sentence. FAIL with a reason when the definition is circular (it defines the term using the term itself), is merely an example or a bare synonym rather than an explanation, is empty or a placeholder, or rambles well beyond a few sentences.
- Keep one entry per concept. Give each concept a single canonical entry under its preferred term, and do not create a second entry for the same idea. When two words mean the same thing, keep the preferred one as the entry and record the others as synonyms on it.
- Record synonyms and deprecated variants as `alternatives` on the canonical entry, not as separate terms. Each alternative is `{ name, note }` (or `{ name, rejected_because }` to mark a form readers should stop using), so a reader who looks up a synonym still lands on the one real definition.
- Keep the headword bare: put only the term in the prose and the meaning in `definition`. Don't restate the term inside its own definition (no circular definitions), and don't fold the part of speech, pronunciation, or source into the headword.
- Cross-reference related concepts with `relates_to`, and place a narrower term under its broader term or category with `has_parent`, so the glossary reads as a connected vocabulary rather than a flat list of isolated words.
- When a new preferred term replaces an old one, retire the old entry (lifecycle `retired`) and link the new term to it with a `replaces` edge (the `supersede` changeset op creates the replacement and the edge together), so a reader who looks up the old word is redirected to the current term instead of finding two live definitions.
- Cite where a definition comes from when it is drawn from an external standard, contract, or document: put the citation in the term's own `locator` (a URL, standard number, or document id). Source provenance is a property of the entry, not a separate node.
- Name a steward by attributing a term — or the glossary's anchor terms — to a Principal with an `attributed_to` edge, so there is a clear owner accountable for reviewing and approving changes. A glossary is a living document; it needs someone to keep it current.
- Walk a term through the lifecycle as it matures: `drafting` while you are still capturing or wording it, `queued` once it is ready for review, `active` when it is the approved, in-force definition, and `retired` when it is deprecated or superseded. Completeness and quality rules apply once a term is committed (`queued` or `active`); a `drafting` stub may be incomplete.
- Agents: read `GET /<handle>/api/authoring-contract.json` and write entries with `POST /<handle>/api/changesets.json` — create each term as a Reference with its `definition` (and any `alternatives`, plus a `locator` when citing a source), and add its `relates_to` / `has_parent` / `attributed_to` edges in the same changeset rather than as disconnected nodes.