Skip to main content
Language: English

README quality blueprint

RC1 planned public routes cover en and zh-Hant-TW; the 41-locale source catalog is private. This editorial blueprint remains canonical in English.

A Better Workflows README is a landing page, not a compressed reference manual. Its job is to help a reader answer five questions in order:

  1. What is this, and is it for me?
  2. What problem does it solve?
  3. Why should I trust its claims?
  4. What is the shortest path to a first success?
  5. Where should I go next?

This blueprint defines the narrative, visual, localization, and validation contract for every repository README. The machine-readable source is readme-quality-v1.json.

Start from the reader's decision

GitHub surfaces a README before most repository content. The first screen must therefore establish the product promise, intended audience, and a bounded next action. It must not begin with internal architecture, a complete command reference, or release-recovery detail.

Write for these reader jobs:

Use a cause-and-effect narrative

The five landing READMEs use the same eight-part semantic sequence. Headings may be idiomatic in each language, but the reader journey does not change.

SectionReader question and narrative role
Promise and audienceWhat is Better Workflows, why does it exist, and who is it for?
Problem to outcomeWhat goes wrong when intent, authority, evidence, and provider outcome are conflated?
Proof and boundariesWhich guarantees make the proposed outcome credible?
First successWhat is the shortest complete install-to-result path?
Choose the next pathWhich workflow or document matches the reader's goal?
LifecycleHow does a goal become a reconciled completion—or stop safely?
Trust and limitsWhat can the system never infer, authorize, or claim?
Learn, help, contributeWhere are the deep docs, support, governance, development, and license?

This order provides a practical arc:

Separate landing content from deep documentation

Use the README for decision-relevant information. Route depth by purpose:

Do not duplicate cache recovery, lock ownership, full provider transport semantics, exhaustive commands, or implementation change history in the landing page. A concise safety claim stays local; its auditable depth belongs in the canonical guide.

This separation follows the Diátaxis distinction between tutorials, how-to guides, explanation, and reference. A single page cannot optimize for all four reader needs at once.

Make every visual earn its place

Use a visual only when relationships, hierarchy, or state transitions are materially easier to understand than prose.

The landing pages permit two visuals:

  1. Authority-boundary architecture: answers which layers shape intent, current facts, tool authority, bounded retries, and read-only state.
  2. Goal-to-completion lifecycle: answers where evidence is checked, where side effects are authorized, and where unknown state stops progress.

Every visual must include:

Do not add decorative screenshots, text-heavy images, or a diagram that merely duplicates a short list. Keep selector tables to two concise columns so they remain usable on narrow screens.

Preserve meaning across languages

English is the semantic reference, not a line-count target. Traditional Chinese, Simplified Chinese, Japanese, and Korean should sound natural to a native reader while preserving the same contract.

The following items must remain equivalent:

Headings, sentence boundaries, punctuation, examples, and calls to action may be idiomatic. Never translate commands, selectors, evidence identifiers, or security semantics.

Write for scanning and translation

Validate semantics, not decoration

The documentation tests must detect more than matching headings. They verify:

Research basis