glossary

glossary

Source of truth: Notion Glossary database under Marketing foundations. The database has slim properties; the rich content lives inside each row's page body under structured H2 headings. Always fetch both.

Notion source (pinned)

Resource	Notion ID	URL
Glossary database	bc7da8d1-b5a1-48b1-9e80-9d2f28730c62	https://www.notion.so/bc7da8d1b5a148b19e809d2f28730c62
Data source (collection)	85e15553-a0ef-42a6-80f9-c984c535f8c7	collection://85e15553-a0ef-42a6-80f9-c984c535f8c7
Parent: Marketing foundations	35cf15e1-b3e7-8168-a618-de718878bdf3	https://www.notion.so/35cf15e1b3e78168a618de718878bdf3

When to load this

• Output skills generating drafts that contain technical or domain terminology.
• pre-publish-qa (Phase 2) when checking term consistency.
• localize-mechanical (Phase 3) and per-language ToV skills (Phase 4) when verifying term equivalence across languages.
• When the user asks "how do we refer to X" or "is term Y on or off."

How to fetch

Two-step fetch:

1. Query the database via query_data_sources against the pinned data source. Filter by Status = Live by default. Filter by Category if the caller specifies one. Filter by Term if the caller asks about a specific term. Returns row metadata: term, category, status, last reviewed, page URL.
2. Fetch each row's page body via notion-fetch on the page URL from step 1. The page body contains the rich detail under structured H2 headings.

If the pinned ID returns an error, fall back to notion-search for Glossary in the Marketing teamspace, then update this skill's pinned ID.

Database properties (slim, by design)

Property	Type	Use
Term	Title	The English form of the term
Status	Select (Draft / In review / Live / Archived)	Plugin filters to Live by default
Category	Select (Cybersecurity / Cyberinsurance / Product / Competitor / Regulatory / Internal)	Filter by content type
Last reviewed	Date	Staleness check

Everything else is in the page body.

Page body structure

Each row's page body uses these H2 headings (set by the team's standard template):

• Definition
• Preferred usage
• Avoid / use instead
• Translations (with FR / DE / ES / NL bullet sub-items)
• Notes

When parsing the page body, extract content under each heading. Empty sections mean "not yet defined" — surface to the caller. For Translations, extract the FR / DE / ES / NL sub-items individually.

Output shape

terms:
  - term: <text>
    category: <text>
    status: <text>
    last_reviewed: <date>
    notion_source: <row url>
    body:
      definition: <text>
      preferred_usage: <text>
      avoid: <text>
      translations:
        fr: <text>
        de: <text>
        es: <text>
        nl: <text>
      notes: <text>
last_fetched: <timestamp>
gaps_flagged: <list>

Categories of particular care

• Regulatory and standards terms — spelling, scope, jurisdiction matter for compliance reasons.
• Insurance terms — jurisdiction-specific. A term that means one thing in French insurance may mean something different in German. The translation sub-items exist for exactly this reason.
• Product names — never translate unless the row explicitly says so. Capitalization is fixed.

Note: competitor info is not stored in this glossary by design. The research-competitor skill fetches competitor data live each time it's invoked. The Competitor category remains in the schema but is not populated.

Empty database state (v1)

In v1 the database is created but not yet populated. When invoked against an empty database, return:

• status: empty
• note: "Glossary database exists at <url> but has no rows yet. The marketing team is in the process of populating it. Output skills should fall back to general usage and flag terminology choices in the QA notes."

This behavior should change as the team adds entries.

Note on use

This is not a writing skill. It returns reference data. Output skills consult it the way a translator consults a termbase: check, do not paraphrase.

Drift detection

When fetching glossary rows, watch for: stale Last reviewed (>6 months), Draft rows that should be Live, empty page-body sections, missing translations for high-stakes terms, contradictions across the FR/DE/ES/NL sub-items. Flag concrete suggestions to the team via the conversation per the standing drift-detection rule.