Spec overview

The ContentRX content model is a small, opinionated, executable specification. It exists so a single piece of UI copy can be evaluated the same way in three different surfaces (Figma plugin, CLI, CI pipeline) without anyone arguing about whose linter is right.

The three primitives

The model has three layers and they compose:

1. Content type (8)

What surface does the string live on? A button label is judged differently than a help-article paragraph. Picking the right type narrows the relevant rules.

See all content types.

2. Moment (13)

What is the user trying to do at the moment of contact? Browsing? Recovering from an error? Onboarding? Moments shape which standards get applied with what weight.

See all moments.

3. Standards (47)

Atomic, citable rules — the things that can be true or false about a string. Each one names the rule, gives a pass and fail example, and lists the content types it applies to.

See all standards.

How it composes

For each string, the evaluator picks a content type and moment, then runs only the standards relevant to that pairing. Mechanical rules (length caps, list parallelism, sentence-case checks) run deterministically; nuanced rules (voice, empathy, jargon) go through an LLM with the standards library injected as the system prompt. The merge step de-dupes and prioritizes by severity.

Categories

• Clarity — 5 standards
• Voice and tone — 5 standards
• Consistency — 5 standards
• Accessibility — 7 standards
• Action-oriented writing — 4 standards
• Content structure — 6 standards
• Grammar and mechanics — 6 standards
• Inclusive language — 2 standards
• Translation readiness — 7 standards