Skip to content
← Tools / Skills
Skill · .md packageictechgy/10 trustMetrics-checked
Design Harness logo

Design Harness

Evidence-based UI/UX QA harness for AI-made interfaces, with a Claude Code skill adapter.

Retired

Design Harness appears to be discontinued or shut down — repo archived, 0 stars This page is kept as an archive record and is excluded from search engines.

Visit official site →

About

An open-source UI/UX QA loop for AI coding agents: point it at a local URL and it captures screenshots, runs conservative source-backed checks (layout, contrast, accessibility, interaction state), and writes an agent-friendly report for the fix loop. The repo ships a Claude Code skill adapter plus an npm CLI (@design-harness/cli) and an agent loop that re-audits until deterministic failures hit zero.

Honest caveat: the repo is archived and has 0 stars, so it may not evolve further — but it has 12 releases, CI, and a real published CLI, so what's there works. Apache 2.0.

What it can do

  • Local-URL screenshot audits (desktop + mobile)
  • Deterministic layout/a11y/interaction checks with evidence references
  • Design-guide contracts (tokens, copy style, budgets)
  • Agent loop: audit, fix, re-audit until zero deterministic failures
  • CI recipe and PR-comment renderer

Setup

  1. Install the audit CLI: npx @design-harness/cli@latest audit --url http://localhost:3000 --out runs/demo
  2. Install Playwright Chromium once: npx playwright install chromium
  3. Copy the skill adapter: copy adapters/claude-code-skill to .claude/skills/product-ui-designer/
  4. Run audits and hand report.md + screenshots back to your agent

Prerequisites

Node 20+; Playwright Chromium; a locally running app to audit

Use cases

codeautomation

Works with

Claude CodeCodexgemini-cli

Trust score — 5/10

/10 trustMetrics-checked

Scored 5/10 (metrics tier, metrics-v1). Evidence: repo is ARCHIVED — maintenance risk, no future evolution expected; 0 stars but 12 releases, CI, and a published npm CLI (@design-harness/cli); evidence-graded findings (deterministic vs heuristic vs review) is a strong design; Apache 2.0. License: Apache-2.0.

★ 0 GitHub stars

How the trust score is built

Full methodology →
Adoption · stars/downloads, log-scaled vs peers
3.0/10
Maintenance · recency, releases, issue load
3.0/10
Openness · license + docs
9.0/10
Manual review · sanity read + red-flag check
6.0/10
  • · repo is ARCHIVED — maintenance risk, no future evolution expected
  • · 0 stars but 12 releases, CI, and a published npm CLI (@design-harness/cli)
  • · evidence-graded findings (deterministic vs heuristic vs review) is a strong design
  • · Apache 2.0

Sources

Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.