design-loop

Reviews visual deliverables with independent fresh-context critics against a design system until they pass.

Updated Aug 20, 2026
One-click install
npx skills add https://github.com/rhorba/RestoLedger --skill design-loop-rhorba
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: design-loop
Source: https://github.com/rhorba/RestoLedger/tree/main/skills/design-loop
Command: npx skills add https://github.com/rhorba/RestoLedger --skill design-loop-rhorba

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? AI-generated designs often look generic because the same context that built the artifact also grades it, leading to endless fix-it loops that never converge. This Skill separates the builder from the judges by running independent, fresh-context critics against a written design system until the rendered result passes. ## Core Features & Use Cases - Design System First: Codifies colors, type scale, spacing, motion, and explicit out-of-bounds rules into a reusable brief before any design work begins. - Three Independent Critics: Runs a Brief critic (requirements), a System critic (token consistency), and a Craft critic (visual taste), each with fresh context and no knowledge of how the artifact was built. - Iterative Fix Rounds: Applies numbered, concrete findings and re-reviews with fresh critics until all pass or diminishing returns are reached. - Use Case: Before shipping a landing page or pitch deck, run the loop so the first delivered version passes brand consistency and craft review instead of the tenth draft. ## Quick Start Run the design loop on this landing page screenshot against our design system brief and tell me what to fix.

Frequently Asked Questions about design-loop

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I stop AI-generated designs from looking generic?

Define a design system up front with color roles, type scale, spacing, and an explicit out-of-bounds list, then have independent fresh-context critics review the rendered result. Separating who builds from who judges prevents the builder from grading its own work generously.

How do I run a design critique loop on a landing page?

Write a one-paragraph brief, build the candidate against your design tokens, then launch three fresh-context critics: Brief, System, and Craft. Apply their numbered findings and re-run all three critics fresh until all pass or new rounds stop surfacing issues.

When should I not use a multi-round design review loop?

Skip the loop for throwaway internal tooling, single icons, one-off emails, or anything nobody will judge on sight and that will not be reused. A full multi-round pass can cost millions of tokens, so reserve it for reusable templates and user-facing assets.

Why must design critics use fresh context instead of the builder conversation?

A context that built the artifact rates its own work generously, which is why fix-it loops never converge. Fresh-context critics see only the rendered artifact, the brief, and the system tokens, with no attachment to the effort spent building it.

How many review rounds does a design critique loop typically need?

Observed convergence is roughly 4 to 10 rounds for a genuinely new artifact, and fewer for tweaks to an already system-conformant one. Stop when all three critics pass or when new rounds stop surfacing new issues.