futurefin-docs-and-writing

CHANGELOG entries follow Keep a Changelog 1.1.0 with SemVer, newest release first under an [Unreleased] bucket. Fixes must be forensic: observable symptom, exact root cause with code identifiers in backticks, the fix, and why it does not regress. Use bold-lead-then-detail bullets and Spanish narrative with English technical terms.

New or changed routes go in .claude/api-routes.md plus an MCP parity evaluation. New or changed env vars go in .claude/env-and-config.md, the README env table if it is a production variable, and .env.example. The Skill's change-type checklist maps every change to its owning docs.

Documentation must be updated in the same PR as the code change, never after. A PR that adds a route without touching api-routes.md is incomplete. A doc may lag the code temporarily but must never describe behavior the code does not yet have.

Code identifiers, API fields, SQL, and commands are always English. CHANGELOG narrative is Spanish with English technical terms embedded untranslated. Reference docs mix Spanish and English; match the surrounding section. Skills are written in English keeping Spanish project vocabulary.

It covers only the writing half of the release ritual. Deciding whether a change may ship belongs to futurefin-change-control, debugging to futurefin-debugging-playbook, test authoring to futurefin-validation-and-qa, and FIRE math content to futurefin-fire-domain-reference.

Known doc/code disagreements are recorded in a standing errata table with the document location, what it says, the verified reality, and a verification date. When a disagreement is found, verify against the code as ground truth, fix the doc in the same change, or register the errata if the fix is out of scope.