What problem does it solve? When writing design documents, investigation reports, or naming code identifiers, vague ad-hoc coinages often appear before the underlying referent is fixed, causing drift between the intended subject and the words used. This Skill enforces a workflow where a referent table (correspondence table) is produced as an independent artifact before any prose is written, so terms are grounded in explicit referents and roles. ## Core Features & Use Cases - Referent-table-first workflow: Saves a standalone referent-table-<slug>.md file with a SHA-256 record before writing any body text, proving the table preceded the document. - Fixed seven-column schema: Columns for source, purpose, concrete referent, role (start condition / state / event / value / record / goal / means), reasoning order, candidate term, and first-use definition, with candidate terms locked to the rightmost column. - Three-layer consistency: Keeps Japanese prose, design elements, and code identifiers (state names, condition names, method names, type names) mapped to the same referents. - Use Case: When asked to write a design document or name states and methods for a feature, first produce the referent table mapping each concept to its role, then write the document using only the terms registered in that table. ## Quick Start Ask the AI to write a design document or name states and methods for your feature, and it will first save a referent table as an independent file before drafting the body text.
