classical-lexicon

Classical Lexicon

Check Risk Map terminology against established security terms of art. The governing rule: never invent a term when an established term of art already exists. This skill grounds a proposed term, regrounds an invented one, or confirms an existing term is already sound — it does not rewrite content wholesale.

When to use

• Authoring a new Control / Risk / Component / Persona title or description.
• Critiquing a draft for coined or product-specific terminology.
• Deciding what to call a capability, role, mechanism, or locus — including a persona/role name, not only a control or risk mechanism term.

When not to use

• Framework mapping IDs (MITRE ATLAS, NIST AI RMF, etc.) — that is the framework-mappings audit tooling's job.
• Prose style/format rules (bold, sentinels, length) — that is the prose-subset linter's job.

Procedure

1. Extract candidate terms from the input: the nouns and noun-phrases that name a capability, security measure, role, mechanism, or architectural locus. These are where terminology matters.
2. Look each up in the lexicon: references/lexicon.md.
3. Classify each candidate:
   • Grounded — already an established term of art. Accept; name the source.
   • Reground — an invented or informal term with an established equivalent. Return the canonical term and the mapping.
   • Generalize — the term encodes a specific product or protocol (e.g. an MCP-prefixed name). The identity is the role/locus; the product is an attribute. Return the role-grain term; keep the product only as an example.
   • Novel — flag — no established term of art found. Do not coin one silently (see the maintainer-flag vocabulary below). "novel-flag" applies only when no established term of art exists in the field, not merely when the term is absent from this pinned lexicon snapshot; when you can ground a term via general NIST-first sourcing knowledge that the snapshot lacks, ground it and propose the lexicon addition rather than flagging it novel. 3b. Record the international/standards equivalent — unconditionally, for grounding verdicts only. This obligation applies to grounded and reground candidates — verdicts that ground a term in standards. For every such candidate, record its international/standards equivalent(s) (ISO/IEC, EU AI Act, ENISA, XACML, or other recognized non-US/international body). Do this every time a term is grounded or regrounded — not only when NIST is silent or the term is contested. A maintainer flag fires only when this equivalence check turns up a problem: the equivalent is missing (D3b-parochialism) or the equivalent conflicts with the NIST-first term (naming-conflict or substantive-conflict, see below). When the equivalent is present and agrees, record it and raise no maintainer flag.