agentic brand

role
Originator, engineer
date
Apr 2026–present
team
Brand, Creative, Product

brand judgment before generation

Agents now make screens, emails, and copy. Without a brand’s stance, they fall back on the model’s defaults. I designed and built Ghost to give them that judgment before they start working.

Teams record their voice, principles, and decisions about trust alongside concrete materials in a .ghost/ package. Any compatible agent can read it. Designers can deepen that guidance without putting the whole brand into every task.

give the next agent the decision

Correcting one output doesn’t teach the next agent. I made the package a place to record the underlying judgment: when to celebrate, how to deliver bad news, or when brand presence should recede.

Authors write those decisions in Markdown, describe when they apply, and reference existing components, assets, or examples. The package travels with the repo.

a package and a selection protocol

I separated discovering the brand from loading its guidance. Ghost handles retrieval; the host agent makes the decisions.

Authors maintain

.ghost/ package

Brand stance · guidance · applicability · materials

User supplies

The task

A screen, email, sentence, or presentation to make.

  1. 01 / Ghost CLI

    gather

    Expose the menu

    Every entry and when it applies. Full guidance stays in the package.

  2. 02 / Host agent

    select

    Match the task

    The agent reads the menu and chooses the applicable entries.

  3. 03 / Ghost CLI

    pull

    Read selected guidance

    Selected entries, material references, and the brand’s overall stance.

  4. 04 / Host agent

    make

    Apply the judgment

    The agent interprets the guidance while creating the work.

Ghost makes no model calls. Unselected guidance stays available in the package, outside the active working context.

what agents make with it

Each brief below went through that selection flow against the same package. The outputs explore different situations and interactions within a shared design language.

less guidance to load

37%

less brand-guidance text in the Square comparison

Ghost’s selected guidance averaged 99.4 KB against a 156.9 KB DESIGN.md containing the full guidance. Across both brands and models, the reduction averaged 30.4%. Human designers reviewed all results and confirmed they met the quality bar.

Project site · Source on GitHub · Development log

Open-source development preview.