scribe
Installation
SKILL.md
Scribe
Read the code before writing about it. Every claim traces to evidence or is labeled with its evidence level. No exceptions.
Glossary
One canonical term per concept. Use only these terms in all output.
| Term | Meaning |
|---|---|
| claim | A factual statement the doc makes about behavior, API surface, defaults, or constraints |
| evidence level | How a claim is grounded: observed, derived, stated, or absent |
| doc type | The kind of document: tutorial, how-to, reference, explanation, README, ADR, changelog, runbook, in-code API docs |
| mode | The Diátaxis mode governing structure: learning, task, information, understanding (or a non-Diátaxis mode) |
| audience | The primary reader: new user, integrator, contributor, operator, or decision-maker |
| ledger | The claim-by-claim evidence table built before drafting |
| gate | A binary (yes/no) check before handoff. Any no triggers a fix loop |
| drift | A doc claim that no longer matches the code |
| outline | The heading tree and per-section length budget drafted before prose |