← Back to skills
Agent-installable Meta Source review 1 minute

Design Eval Loop

No dead controls, no fake 5/5.

Iterate an existing UI artifact to a verified 5/5 design score with per-view dead-item audits, rubric scoring, fixes, and fresh re-score receipts.

ClawHub download
This skill is mirrored to the h-mascot ClawHub account for install-from-the-page convenience.
The ClawHub package is regenerated by scripts/sync-superada-to-clawhub.mjs on every push to main.
Open on ClawHub →
Install contract
Source
enterprise-crew-skills/design-eval-loop
Entrypoint
SKILL.md
Installability
Ready for agent install
Source spec
https://superada.ai/skills/design-eval-loop/SKILL.md

This page is designed to be the thing you send to another agent. It tells the agent where the source of truth lives, whether install is real today, and what command or manual path to follow.

Canonical source
https://github.com/h-mascot/Enterprise-Crew-skills/tree/main/design-eval-loop
Install command
curl -sSf https://superada.ai/install/design-eval-loop | sh
Prompt to hand another agent
Install the skill described on this page from https://superada.ai/skills/design-eval-loop/SKILL.md. Review the source first, then run: curl -sSf https://superada.ai/install/design-eval-loop | sh

Install steps

Step 1
Review the skill contract and rubric before running it.
Step 2
Install with the SuperAda curl installer or copy the bundle into your skills directory.
Step 3
Use it only after a UI artifact exists; build-from-scratch belongs to a separate design skill.
Step 4
Keep public scorecards free of private URLs, screenshots, and customer data.

What it covers

  • Skill contract
  • Design rubric
  • Sub-agent prompt templates
  • Curl installer

Good use cases

  • Score a finished prototype against a current design rubric
  • Audit every clickable item for dead controls
  • Iterate each view until a fresh reviewer confirms 5/5
  • Produce evidence instead of claiming a vibes-based pass

Published artifacts

Skill contract
https://superada.ai/skills/design-eval-loop/SKILL.md
Agent-facing loop for per-view design scoring and dead-item audits.
Rubric
https://superada.ai/skills/design-eval-loop/references/rubric.md
Weighted 2025/2026 design evaluation rubric.
Sub-agent prompts
https://superada.ai/skills/design-eval-loop/references/sub-agent-prompts.md
Read-only audit and scoring prompt templates.
Curl installer
https://superada.ai/install/design-eval-loop
Shell installer for local agent runtimes.

Current limits

  • Does not create the initial UI artifact.
  • Requires actual browser or file-grounded inspection to make a 5/5 claim.
  • If a structural blocker remains after the iteration cap, report the true score instead of rounding up.