Perceptual Protocols

Perceptual Protocols

This project is all about how things should feel.

We've spent the last two years building increasingly sophisticated ways to give AI agents procedural knowledge. How to write CSS. How to follow a design system. How to obey our rules. Owl-Listener contains a lot of procedural design skills and methods that are all machine readbale and made for producing great design with LLMs. This collection is more about telling the LLM how things should feel with great accuracy so you can get rich and meaningful outcomes.

Designers, musicians, writers, architects, and every creative human spend their career cultivating, is **perceptual**. Taste. Judgement. The ability to look at something and know it is wrong before you can articulate why. We've been encoding the **how** and ignoring the **what it should feel like**.

This is a family of protocols for this.

---

What's in here

Each protocol does one thing and stops. Each one outputs a plain markdown file that any agent — Claude, Gemini, ChatGPT, Cursor, Copilot — can read. They work alone, but they compose into a complete workflow when used together.

Protocol	What it does	Output	Status
mood-protocol	Declare what the product should feel like, from visual references	mood.md	Released (v0.1)
vocab-protocol	Share a vocabulary for the static qualities you mean	vocab.md	Released (v0.1)
motion-protocol	Share a vocabulary for how the product should move	motion.md	Released (v0.1)
voice-protocol	Specify how the brand writes — personality, register, mechanics	voice.md	Released (v0.1)
listen-protocol	Author a brief from audio references; share sound vocabulary	listen.md + sound.md	Released (v0.1)
situation-protocol	Weight the qualities differently by context	situation.md	Released (v0.1)